Browse the corpus

Introduction When left untreated, approximately 5% of patients with transient ischemic attack (TIA) or minor stroke have a major stroke within 24 hours, comprising more than 40% of all recurrent strokes within 30 days.1 Urgent investigation and medical treatment substantially reduce risk of early recurrent stroke,2,3 as could initial self-medication with aspirin alone.4 Consequently, guidelines recommend that patients with high-risk TIA should be assessed urgently.5,6,7 However, patients frequently fail to recognize or act on TIA symptoms, either delaying seeking medical attention8,9 or not seeking medical attention at all.10,11 The number of potentially preventable early recurrent strokes that consequently go unprevented is unknown, although recent public education campaigns designed to increase recognition of major stroke symptoms might change behavior after TIA. The Face, Arm, Speech, Time (FAST) test was adopted as a tool to improve symptom recognition after stroke12,13 and has formed the basis of public education in many countries, including the United Kingdom, Ireland, United States, Australia, and New Zealand, with variants in several non–English-speaking countries. The FAST test was used in an ongoing television public awareness campaign in the United Kingdom from 2009 onward. It appears to have improved the response after major stroke,14,15,16,17 but the association of the campaign with patient behavior after TIA and minor stroke has not been determined and may well differ given differences in event duration, severity, and coverage by the FAST acronym.

n the United Kingdom from 2009 onward. It appears to have improved the response after major stroke,14,15,16,17 but the association of the campaign with patient behavior after TIA and minor stroke has not been determined and may well differ given differences in event duration, severity, and coverage by the FAST acronym. We prospectively studied patient perception and behavior after TIA and stroke in a population-based study before and during the ongoing FAST campaign. We also investigated the number of early strokes after a TIA for which no medical attention is sought. Methods Ethical Approval The Oxford Vascular Study (OxVasc) was approved by the Oxfordshire Research Ethics Committee and included use of routinely collected health care data for investigating incidence rates without consent. Written informed consent or assent was obtained from all participants for additional data. Study Design The OxVasc is a population-based study of all acute vascular events, including TIA and stroke in 92 728 individuals of all ages registered with 100 collaborating primary care physicians at 9 general practices in Oxfordshire, United Kingdom. The OxVasc study methods have been described previously18 and are recounted in the online methods (eFigure in the Supplement). The present article includes all consecutive incident TIA and stroke cases, with the exception of subarachnoid hemorrhages, occurring outside the hospital between April 1, 2002, and March 31, 2014. Data analysis took place from July 1, 2013, to March 2, 2015.

recounted in the online methods (eFigure in the Supplement). The present article includes all consecutive incident TIA and stroke cases, with the exception of subarachnoid hemorrhages, occurring outside the hospital between April 1, 2002, and March 31, 2014. Data analysis took place from July 1, 2013, to March 2, 2015. Most patients were seen in the dedicated study TIA and stroke clinics or were admitted to the acute stroke service at the principal center (John Radcliffe Hospital, Oxford, United Kingdom) serving the study population. Patients provided informed consent (or assent was obtained from relatives) and were seen by study physicians (among whom were F.J.W., L.L., and P.M.R.) for structured interview using a standard questionnaire as soon as possible after initial presentation to assess their perception about the event and immediate response to symptoms, including date and time of symptom onset, when medical attention was sought and by whom, the first contact with emergency medical services (EMS), and why they did not seek medical attention straightaway in case of delay to medical attention exceeding 3 hours. Patients were routinely questioned about any neurological symptoms within 90 days before their presenting event (ie, unheeded TIAs) and were followed up for recurrent cerebrovascular events (eFigure in the Supplement).18 Baseline characteristics, including demographic data, self-reported race/ethnicity, and risk factors for stroke (based on prior diagnosis and current medication use), were recorded, and assessments were made for severity of the event using the National Institutes of Health Stroke Scale. Major stroke was defined as a National Institutes of Health Stroke Scale score exceeding 3. Socioeconomic status was assessed according to the United Kingdom’s 2007 indexes of deprivation.19 Based on these indexes, the electoral districts covering our population are less deprived than the rest of the United Kingdom, but still 22% of our districts rank in the lower one-third nationally. Further data were acquired from medical records, ambulance sheets, general practitioner referral letters, and consultation notes. Consent for access to this information was obtained from all participating patients.

an the rest of the United Kingdom, but still 22% of our districts rank in the lower one-third nationally. Further data were acquired from medical records, ambulance sheets, general practitioner referral letters, and consultation notes. Consent for access to this information was obtained from all participating patients. To best reflect the response of patients to symptoms, time of symptom onset was defined as time of awaking if onset of stroke was during sleep. In patients unable to call for help, time of symptom onset was considered the moment another person noted their symptoms. Events were classified as FAST positive when at least one symptom from the FAST campaign (ie, facial weakness, arm weakness, or speech disturbance) was present at symptom onset.

ke was during sleep. In patients unable to call for help, time of symptom onset was considered the moment another person noted their symptoms. Events were classified as FAST positive when at least one symptom from the FAST campaign (ie, facial weakness, arm weakness, or speech disturbance) was present at symptom onset. FAST Campaign in the United Kingdom The initial FAST public education television campaign in the United Kingdom ran from February through April 2009, with 8 weeks of national television broadcasts. With initiation of the 2009 campaign, the T in FAST was redesignated “Time to call 999” (ie, EMS) rather than “Test all 3.” Repeated television campaigns ran intermittently for several weeks from October 2009 to early 2010, for 6 weeks in March and April 2011, and for 4 weeks in March 2012, March 2013, and March 2014 and are now continued on a yearly basis. The total campaign investment from 2009 to 2013 was £10.2 million (US $13.6 million). The impact of the campaign as measured by television viewer ratings (1 television viewer rating equals 1% of the target audience of all adults, including multiple views) was 274 in 2012 and 317 in 2013. In October 2005, there was a preceding small-scale 18-month public transport poster campaign.

ion (US $13.6 million). The impact of the campaign as measured by television viewer ratings (1 television viewer rating equals 1% of the target audience of all adults, including multiple views) was 274 in 2012 and 317 in 2013. In October 2005, there was a preceding small-scale 18-month public transport poster campaign. Statistical Analysis Analyses included all first TIA and first stroke occurring outside of the hospital during the study period, with the exception of analyses of patient perception, which excluded patients with reduced consciousness, event-related confusion, or dysphasia. We analyzed patient behavior before April 1, 2009, and after April 1, 2009 (ie, the end of the first major 2-month television campaign) and stratified the analyses (ie, assessed behavior per year of the study) into study years. To assess any association of the FAST campaign with the number of strokes in our study population that could potentially have been prevented by urgent patient behavior after initial symptoms of a TIA, we assessed the number of all 90-day recurrent ischemic strokes preceded by an unheeded TIA. Together with all 90-day recurrences after heeded events, these constitute all early recurrent strokes. We compared the unheeded TIAs with the TIAs in patients who sought medical attention and extrapolated the OxVasc population rates (number of unheeded TIAs divided by the OxVasc source population per 10 years) to the general population to estimate the number of potentially preventable strokes after unheeded TIAs annually.

compared the unheeded TIAs with the TIAs in patients who sought medical attention and extrapolated the OxVasc population rates (number of unheeded TIAs divided by the OxVasc source population per 10 years) to the general population to estimate the number of potentially preventable strokes after unheeded TIAs annually. Time from TIA or stroke symptom onset to first seeking medical attention and the nature of the first medical attention sought—defined as emergency (ie, direct contact with ambulance services or presentation to an emergency department) vs nonemergency (ie, the first contact with a general practitioner or other local health care professional)—were also compared before vs after April 1, 2009, within clinically relevant timing cutoffs of 3 and 24 hours by performing χ2 test and computing Mantel-Haenszel test odds ratios (ORs). To further assess time trends, the first contact with EMS vs non-EMS was analyzed per year of the study. Subsequent adjustment was made for time trends and age, sex, race/ethnicity, socioeconomic status, cohabitation, National Institutes of Health Stroke Scale and ABCD2 (age, blood pressure, clinical features of the TIA, duration of symptoms, and history of diabetes) scores, onset during sleep, and occurrence during the weekend20 by means of segmented regression analysis,21 breaking down time to months and again using April 1, 2009, as the change point. Missing data for covariates in this model were imputed (eTable 1 in the Supplement). We formally assessed for differential associations of the FAST campaign by event type (TIA and minor stroke vs major stroke) by testing for multiplicative interaction in the fully adjusted time-series regression model.

he change point. Missing data for covariates in this model were imputed (eTable 1 in the Supplement). We formally assessed for differential associations of the FAST campaign by event type (TIA and minor stroke vs major stroke) by testing for multiplicative interaction in the fully adjusted time-series regression model. We assessed the association of perception (classified as correct [ie, TIA, stroke, or ministroke] vs incorrect) with subsequent behavior, as well as the proportion of patients with correct initial perception of symptoms before vs after April 1, 2009, in relation to presence of FAST symptoms. Finally, we repeated the trends analyses, comparing only the 5 years immediately before April 1, 2009, vs the subsequent 5 years. All analyses were performed using a software program (SPSS Statistics, version 21.0; IBM). Two-sided α level (type I error) was set at .05.

1, 2009, in relation to presence of FAST symptoms. Finally, we repeated the trends analyses, comparing only the 5 years immediately before April 1, 2009, vs the subsequent 5 years. All analyses were performed using a software program (SPSS Statistics, version 21.0; IBM). Two-sided α level (type I error) was set at .05. Results Among 2243 consecutive patients with first TIA or stroke in the study period (mean [SD] age, 73.6 [13.4] years; 1126 [50.2%] female; 96.3% of white race/ethnicity), 825 (36.8%) were initially seen with TIA, 831 (37.0%) with minor stroke, and 587 (26.2%) with major stroke. Baseline characteristics of patients (1231 pre-FAST and 1012 post-FAST) are listed in eTable 1 in the Supplement. Data on who sought medical attention were available in 2003 patients (89.3%), and data on first clinician were available in 2196 patients (97.9%). Time from event to call for medical attention and hospital arrival was unknown in 209 patients (9.3%), most commonly because of unconsciousness, dysphasia, dementia, or early death and unavailability of an alternative informant.

2003 patients (89.3%), and data on first clinician were available in 2196 patients (97.9%). Time from event to call for medical attention and hospital arrival was unknown in 209 patients (9.3%), most commonly because of unconsciousness, dysphasia, dementia, or early death and unavailability of an alternative informant. Presence of FAST Symptoms Of 504 major strokes without collapse or loss of consciousness, 452 (89.7%) had one or more FAST symptoms, as opposed to only 504 of 799 (63.1%) patients initially seen with TIA and 498 of 811 (61.4%) patients with minor stroke. Among patients with minor stroke and TIA, facial weakness was reported in 178 of 780 (22.8%) and 125 of 771 (16.2%), respectively; upper limb motor symptoms in 307 of 785 (39.1%) and 224 of 770 (29.1%), respectively; and speech disturbance in 307 of 810 (37.9%) and 351 of 800 (43.9%), respectively. Among patients with major stroke, 288 of 471 (61.1%) had facial weakness, 377 of 481 (78.4%) had upper limb weakness, and 336 of 458 (73.4%) had speech symptoms. Weekend Presentation Patients with TIA or minor stroke were seen less often during the weekend, with higher numbers initially seen directly after the weekend (Figure 1). For major stroke, presentation was similar throughout the week. Weekend presentation of TIA or minor stroke vs major stroke was 267 of 1537 (17.4%) vs 155 of 575 (27.0%) (OR, 0.57; 95% CI, 0.45-0.71; P < .001). These findings were similar before and after April 1, 2009.

ially seen directly after the weekend (Figure 1). For major stroke, presentation was similar throughout the week. Weekend presentation of TIA or minor stroke vs major stroke was 267 of 1537 (17.4%) vs 155 of 575 (27.0%) (OR, 0.57; 95% CI, 0.45-0.71; P < .001). These findings were similar before and after April 1, 2009. Figure 1. Occurrence of and Presentation With Transient Ischemic Attack (TIA), Minor Stroke, and Major Stroke by Day of the Week Shown is the occurrence of cerebrovascular events, stratified by event severity and day of the week (A), contrasted against the day of the week when medical attention is sought for these events (B).

1. Occurrence of and Presentation With Transient Ischemic Attack (TIA), Minor Stroke, and Major Stroke by Day of the Week Shown is the occurrence of cerebrovascular events, stratified by event severity and day of the week (A), contrasted against the day of the week when medical attention is sought for these events (B). Patient Presentation Before and After the FAST Campaign For major stroke, use of nonemergency services (chiefly a general practitioner) declined steeply from 2009 onward (Figure 2B). Use of EMS was 58.8% before April 1, 2009, vs 78.9% after April 1, 2009 (OR, 2.63; 95% CI, 1.80-3.82; P < .001) (Table 1). Moreover, first medical attention was sought more quickly after April 1, 2009, than before April 1, 2009 (Figure 2D). First medical attention was obtained within 3 hours in 67.6% (198 of 293) before April 1, 2009, vs in 81.3% (204 of 251) after April 1, 2009 (OR, 2.08; 95% CI, 1.40-3.11; P < .001). Taking into account time trends and potential confounding variables, these changes largely coincided with initiation of the FAST campaign (Table 2). For TIA and minor stroke, medical attention was also sought more often directly via EMS after April 1, 2009, occurring in 21.9% before April 1, 2009, vs in 30.9% after April 1, 2009 (OR, 1.60; 95% CI, 1.28-2.00; P < .001) (Table 1). However, this increase was not significantly associated with the FAST campaign in a time-series analysis (Figure 2A). The change was attributable to a baseline trend (OR, 1.01; 95% CI, 1.00-1.02; P = .009) rather than to the campaign (adjusted OR, 0.79; 95% CI, 0.50-1.23; P = .29) (Table 2). Time to first seeking medical attention after TIA and minor stroke was similar before and after April 1, 2009. Medical attention was obtained within 3 hours in 42.1% (384 of 912) before April 1, 2009, vs in 40.4% (298 of 737) after April 1, 2009 (OR, 0.93; 95% CI, 0.77-1.14; P = .49), and was obtained within 24 hours in 70.4% (612 of 869) before April 1, 2009, vs in 70.6% (493 of 698) after April 1, 2009 (OR, 1.01; 95% CI, 0.81-1.26; P = .93), and did not change substantially during the course of the study (Table 2 and Figure 2C). The observed association of the FAST campaign with response to TIA and minor stroke differed significantly from the association with response to major stroke for both use of EMS (P for interaction in the time-series analysis = .03 vs major stroke) and time to first seeking medical attention within 24 hours (P for interaction in the time-series analysis = .006 vs major stroke).

to TIA and minor stroke differed significantly from the association with response to major stroke for both use of EMS (P for interaction in the time-series analysis = .03 vs major stroke) and time to first seeking medical attention within 24 hours (P for interaction in the time-series analysis = .006 vs major stroke). Results were similar for TIA and minor stroke and when restricting analyses to 5 years before and 5 years after initiation of the campaign. Figure 2. Nonemergency Presentation (Chiefly to a General Practitioner) and Time to Seeking Medical Attention per Year Within the Study Period Shown is nonemergency presentation and time to seeking medical attention for transient ischemic attack (TIA) and minor stroke (A and C) and for major stroke (B and D). The vertical blue lines indicate televised Face, Arm, Speech, Time (FAST) campaigns. The first 2 lines represent 3-month time periods; the third line, 6-week time periods; and the subsequent lines, 4-week time periods.

ention for transient ischemic attack (TIA) and minor stroke (A and C) and for major stroke (B and D). The vertical blue lines indicate televised Face, Arm, Speech, Time (FAST) campaigns. The first 2 lines represent 3-month time periods; the third line, 6-week time periods; and the subsequent lines, 4-week time periods. Table 1. First Health Care Professional Contacted Before and After the 2009 FAST Campaign for TIA and Minor Stroke and for Major Stroke Variable No. (%) OR (95% CI) P Value Pre-FAST Post-FAST TIA and Minor Stroke (n = 897) (n = 735) Nonemergency 701 (78.1) 508 (69.1) 0.63 (0.50-0.78) <.001 GP 643 (71.7) 447 (60.8) 0.61 (0.50-0.75) <.001 NHS Directa 21 (2.3) 29 (3.9) 1.71 (0.97-3.03) .06 Otherb 37 (4.1) 32 (4.4) 1.06 (0.65-1.72) .82 Emergency 196 (21.9) 227 (30.9) 1.60 (1.28-2.00) <.001 A&E 34 (3.8) 51 (6.9) 1.89 (1.21-2.95) .004 999c 143 (15.9) 164 (22.3) 1.51 (1.18-1.94) .001 Eye hospital 19 (2.1) 12 (1.6) 0.77 (0.37-1.59) .48 Major Stroke (n = 308) (n = 256) Nonemergency 127 (41.2) 54 (21.1) 0.38 (0.26-0.56) <.001 GP 124 (40.3) 47 (18.4) 0.33 (0.23-0.49) <.001 NHS Directa 1 (0.3) 2 (0.8) 2.42 (0.22-26.81) .46 Otherb 2 (0.6) 5 (2.0) 3.05 (0.59-15.84) .16 Emergency 181 (58.8) 202 (78.9) 2.63 (1.80-3.82) <.001 A&E 5 (1.6) 9 (3.5) 2.21 (0.73-6.67) .15 999c 176 (57.1) 193 (75.4) 2.30 (1.60-3.30) <.001 Abbreviations: A&E, accident and emergency department; FAST, Face, Arm, Speech, Time; GP, general practitioner; NHS, National Health Service; OR, odds ratio; TIA, transient ischemic attack. a Was dissolved in March 2014.

Table 1. First Health Care Professional Contacted Before and After the 2009 FAST Campaign for TIA and Minor Stroke and for Major Stroke Variable No. (%) OR (95% CI) P Value Pre-FAST Post-FAST TIA and Minor Stroke (n = 897) (n = 735) Nonemergency 701 (78.1) 508 (69.1) 0.63 (0.50-0.78) <.001 GP 643 (71.7) 447 (60.8) 0.61 (0.50-0.75) <.001 NHS Directa 21 (2.3) 29 (3.9) 1.71 (0.97-3.03) .06 Otherb 37 (4.1) 32 (4.4) 1.06 (0.65-1.72) .82 Emergency 196 (21.9) 227 (30.9) 1.60 (1.28-2.00) <.001 A&E 34 (3.8) 51 (6.9) 1.89 (1.21-2.95) .004 999c 143 (15.9) 164 (22.3) 1.51 (1.18-1.94) .001 Eye hospital 19 (2.1) 12 (1.6) 0.77 (0.37-1.59) .48 Major Stroke (n = 308) (n = 256) Nonemergency 127 (41.2) 54 (21.1) 0.38 (0.26-0.56) <.001 GP 124 (40.3) 47 (18.4) 0.33 (0.23-0.49) <.001 NHS Directa 1 (0.3) 2 (0.8) 2.42 (0.22-26.81) .46 Otherb 2 (0.6) 5 (2.0) 3.05 (0.59-15.84) .16 Emergency 181 (58.8) 202 (78.9) 2.63 (1.80-3.82) <.001 A&E 5 (1.6) 9 (3.5) 2.21 (0.73-6.67) .15 999c 176 (57.1) 193 (75.4) 2.30 (1.60-3.30) <.001 Abbreviations: A&E, accident and emergency department; FAST, Face, Arm, Speech, Time; GP, general practitioner; NHS, National Health Service; OR, odds ratio; TIA, transient ischemic attack. a Was dissolved in March 2014. b Includes optician, eye hospital, private physician, medical staff while traveling (eg, airport staff, ship physician, or hotel physician), research physician, or mentioning of symptoms during routine consultant review. Statistical tests comparing pre-FAST vs post-FAST for nonemergency and emergency presentation are interchangeable. c Emergency medical services. Table 2.

b Includes optician, eye hospital, private physician, medical staff while traveling (eg, airport staff, ship physician, or hotel physician), research physician, or mentioning of symptoms during routine consultant review. Statistical tests comparing pre-FAST vs post-FAST for nonemergency and emergency presentation are interchangeable. c Emergency medical services. Table 2. Determinants of Urgent Response to TIA and Minor Stroke and to Major Stroke in Segmented Time-Series Analysis Variable Use of Emergency Medical Services Seeking Medical Attention Within 3 h Seeking Medical Attention Within 24 h OR (95% CI) P Value OR (95% CI) P Value OR (95% CI) P Value Determinants of Urgent Response to TIA and Minor Stroke Constant 0.06 (0.02-0.14) <.001 0.25 (0.10-0.64) .004 0.59 (0.23-1.53) .28 Baseline trend 1.01 (1.00-1.02) .009 1.00 (0.99-1.01) .87 1.01 (1.00-1.01) .03 Change at interventiona 0.79 (0.50-1.23) .29 0.87 (0.58-1.29) .48 0.75 (0.48-1.19) .22 Trend after intervention 1.01 (0.99-1.02) .42 1.00 (0.99-1.01) .59 1.00 (0.99-1.01) .51 Age 1.00 (0.99-1.01) .64 1.01 (1.00-1.02) .03 1.01 (1.00-1.02) .22 Female sex 0.89 (0.70-1.12) .31 0.89 (0.72-1.09) .26 0.84 (0.67-1.06) .13 Nonwhite race/ethnicity 1.28 (0.64-2.56) .47 0.76 (0.38-1.52) .43 0.85 (0.46-1.57) .60 IMD 1.01 (0.99-1.02) .50 1.01 (0.99-1.02) .42 1.01 (0.99-1.03) .40 Living alone 1.17 (0.89-1.52) .26 0.81 (0.64-1.03) .08 0.92 (0.70-1.19) .51 ABCD2 score 1.22 (1.13-1.33) <.001 1.22 (1.13-1.30) <.001 1.31 (1.21-1.42) <.001 Symptoms on awaking 0.78 (0.59-1.04) .09 0.72 (0.56-0.93) .01 0.83 (0.63-1.11) .20 Weekend occurrence 0.99 (0.76-1.27) .90 0.80 (0.64-1.00) .05 0.53 (0.42-0.68) <.001 Determinants of Urgent Response to Major Stroke Constant 0.60 (0.09-3.68) .43 0.28 (0.03-2.52) .26 NA NA Baseline trend 1.01 (1.00-1.02) .15 0.99 (0.98-1.00) .17 NA NA Change at interventiona 1.68 (0.76-3.75) .15 2.56 (1.11-5.90) .03 NA NA Trend after intervention 1.00 (0.98-1.02) .82 1.01 (0.99-1.04) .31 NA NA Age 0.98 (0.97-1.00) .07 1.01 (0.99-1.03) .30 NA NA Female sex 0.86 (0.57-1.29) .47 1.04 (0.68-1.60) .85 NA NA Nonwhite race/ethnicity 1.27 (0.42-3.82) .67 3.28 (0.66-16.39) .15 NA NA IMD 1.00 (0.97-1.03) .82 0.98 (0.95-1.02) .31 NA NA Living alone 1.58 (1.01-2.46) .05 0.60 (0.36-1.00) .05 NA NA Stroke severity on NIHSS 1.16 (1.11-1.22) <.001 1.10 (1.06-1.15) <.001 NA NA Symptoms on awaking 0.79 (0.50-1.24) .31 0.52 (0.30-0.88) .02 NA NA Abbreviations: ABCD, age, blood pressure, clinical features of the TIA, duration of symptoms, and history of diabetes; FAST, Face, Arm, Speech, Time; IMD

(0.36-1.00) .05 NA NA Stroke severity on NIHSS 1.16 (1.11-1.22) <.001 1.10 (1.06-1.15) <.001 NA NA Symptoms on awaking 0.79 (0.50-1.24) .31 0.52 (0.30-0.88) .02 NA NA Abbreviations: ABCD, age, blood pressure, clinical features of the TIA, duration of symptoms, and history of diabetes; FAST, Face, Arm, Speech, Time; IMD , Index of Multiple Deprivation (indicating socioeconomic status, with higher indexes indicating more deprived); NA, not applicable; NIHSS, National Institutes of Health Stroke Scale; OR, odds ratio; TIA, transient ischemic attack. a The odds ratios for change at intervention reflect the change at the infliction point (ie, the start of the FAST campaign) in use of emergency medical services or time to seeking medical attention, adjusted for all other variables, including trends before and after campaign initiation. For example, in the penultimate right column, the odds of seeking medical attention within 24 hours were 0.75 times lower after initiation of the campaign than before.

mergency medical services or time to seeking medical attention, adjusted for all other variables, including trends before and after campaign initiation. For example, in the penultimate right column, the odds of seeking medical attention within 24 hours were 0.75 times lower after initiation of the campaign than before. Stroke After Unheeded TIAs Ninety-five patients who had sought attention for an initial TIA or stroke had a first or recurrent stroke by 90-day follow-up. In addition, 93 patients with stroke reported having symptoms of a TIA for which they had not sought medical attention during the 90 days before their presenting stroke. Therefore, there were 188 early strokes after initial TIA or stroke in all patients, in whom 93 (49.5%) occurred after a TIA for which no medical attention was sought. This number of strokes preceded by an unheeded TIA was similar before and after the FAST campaign (43 of 538 [8.0%] before vs 50 of 615 [8.1%] after, P = .93). The median interval from the first unheeded TIA to a stroke was 6 days (interquartile range [IQR], 1-17 days). The median duration of unheeded TIAs was 20 minutes (IQR, 5-60 minutes). Thirty-two of 93 patients (34.4%) with unheeded TIAs preceding their stroke had more than one TIA before stroke occurrence. Among all 93 patients with unheeded TIAs preceding stroke, 69 (74.2%) were not taking antiplatelet medication or oral anticoagulants at the time of their stroke. Compared with TIAs for which medical attention was sought, unheeded TIAs were shorter, more often consisted of a single symptom only, and were less likely to disturb speech (Table 3). Only 34.8% of those with unheeded TIAs had any symptom covered by the FAST acronym. Patients with unheeded TIAs were also younger, were more often male, resided in more socially deprived neighborhoods, and lived in a shared household. Most recurrent stroke after heeded TIA occurred despite specialized hospital care within 24 hours (67.4% [58 of 86 for whom data were available]) or 72 hours (87.2% [75 of 86]) of the initial event.

ere also younger, were more often male, resided in more socially deprived neighborhoods, and lived in a shared household. Most recurrent stroke after heeded TIA occurred despite specialized hospital care within 24 hours (67.4% [58 of 86 for whom data were available]) or 72 hours (87.2% [75 of 86]) of the initial event. Table 3. Patient and Event Characteristics of Those Initially Seen With TIA vs Those Initially Seen With Stroke After an Unheeded TIAa Variable All TIAs (N = 825) Unheeded TIAs (n = 93) P Value Age, mean (SD), y 72.7 (13.4) 67.7 (14.9) <.001 Female sex, No. (%) 426 (51.6) 38 (40.9) .05 Nonwhite race/ethnicity, No. (%) 22 (3.2) 2 (2.2) .63 IMD, median, (IQR) 7.2 (4.8-13.1) 9.4 (5.0-13.0) .08 Living alone, No. (%) 221 (28.2) 16 (17.8) .04 Increased cardiovascular risk, No. (%)b 26 (28.6) 236 (28.8) .97 Duration of TIA, median (IQR), min 30 (10-180) 20 (5-60) .03 Isolated symptoms, No. (%)c 401 (50.2) 69 (75.8) <.001 Symptom type, No. (%)d Motor 340 (42.7) 33 (35.9) .35 Facial weakness 125 (16.2) 3 (3.4) .05 Arm weakness 224 (29.1) 18 (19.4) .72 Sensory 209 (26.3) 19 (20.7) .37 Speech 349 (43.7) 16 (17.2) <.001 Visual 231 (29.1) 28 (30.1) .92 Vertigo 48 (6.0) 8 (8.6) .15 FAST positive, No. (%) 504 (63.1) 32 (34.8) <.001 Abbreviations: FAST, Face, Arm, Speech, Time; IMD, Index of Multiple Deprivation (indicating socioeconomic status, with higher indexes indicating more deprived); IQR, interquartile range; TIA, transient ischemic attack. a For the percentages, some denominators vary from the heading totals because of missing data.

Table 3. Patient and Event Characteristics of Those Initially Seen With TIA vs Those Initially Seen With Stroke After an Unheeded TIAa Variable All TIAs (N = 825) Unheeded TIAs (n = 93) P Value Age, mean (SD), y 72.7 (13.4) 67.7 (14.9) <.001 Female sex, No. (%) 426 (51.6) 38 (40.9) .05 Nonwhite race/ethnicity, No. (%) 22 (3.2) 2 (2.2) .63 IMD, median, (IQR) 7.2 (4.8-13.1) 9.4 (5.0-13.0) .08 Living alone, No. (%) 221 (28.2) 16 (17.8) .04 Increased cardiovascular risk, No. (%)b 26 (28.6) 236 (28.8) .97 Duration of TIA, median (IQR), min 30 (10-180) 20 (5-60) .03 Isolated symptoms, No. (%)c 401 (50.2) 69 (75.8) <.001 Symptom type, No. (%)d Motor 340 (42.7) 33 (35.9) .35 Facial weakness 125 (16.2) 3 (3.4) .05 Arm weakness 224 (29.1) 18 (19.4) .72 Sensory 209 (26.3) 19 (20.7) .37 Speech 349 (43.7) 16 (17.2) <.001 Visual 231 (29.1) 28 (30.1) .92 Vertigo 48 (6.0) 8 (8.6) .15 FAST positive, No. (%) 504 (63.1) 32 (34.8) <.001 Abbreviations: FAST, Face, Arm, Speech, Time; IMD, Index of Multiple Deprivation (indicating socioeconomic status, with higher indexes indicating more deprived); IQR, interquartile range; TIA, transient ischemic attack. a For the percentages, some denominators vary from the heading totals because of missing data. b Presence of at least 2 vascular risk factors (ie, hypertension, diabetes, hypercholesterolemia, or current smoking). c One type of symptoms present (eg, sensory symptoms only). d P values adjusted for co-occurrence of any other symptoms.

a For the percentages, some denominators vary from the heading totals because of missing data. b Presence of at least 2 vascular risk factors (ie, hypertension, diabetes, hypercholesterolemia, or current smoking). c One type of symptoms present (eg, sensory symptoms only). d P values adjusted for co-occurrence of any other symptoms. Patient Perception of Symptoms In 1419 patients without severe speech disturbance, cognitive impairment, or event-related confusion after TIA or minor stroke who provided their initial perception of symptoms, 467 (32.9%) correctly attributed these to TIA or stroke. Patients with correct perception sought medical attention more quickly (241 of 438 [55.0%] when correct vs 337 of 887 [38.0%] when incorrect within 3 hours; OR, 1.45; 95% CI, 1.18-1.77; P < .001) and were more likely to directly contact EMS (130 of 458 [28.4%] when correct vs 193 of 946 [20.4%] when incorrect; OR, 1.55; 95% CI, 1.20-2.00; P = .001) (eTable 2 in the Supplement). Correct perception of symptoms after TIA and minor stroke was less common after April 1, 2009, compared with before April 1, 2009 (from 37.3% [289 of 774] to 27.6% [178 of 645]; OR, 0.64; 95% CI, 0.51-0.80; P < .001), particularly in FAST-positive cases (eTable 3 in the Supplement). Decline in correct perception was seen for both TIA and minor stroke and was similar for low-risk and high-risk TIA. Among 245 of 588 patients (41.7%) with major stroke who provided their initial perception of symptoms, perception was correct in 116 patients (47.3%) but was not associated with the proportion contacting EMS (54.5% [61 of 112] when correct vs 53.6% [67 of 125] when incorrect; OR, 1.04; 95% CI, 0.62-1.73; P = .89) or time to seeking first medical attention (median, 60 minutes [IQR, 10-173 minutes] when correct vs 60 minutes [IQR, 15-537 minutes] when incorrect; P = .12). In most major strokes (483 of 543 [89.0%]), first medical attention was sought by someone other than the patient. For TIA and minor stroke, 52.5% (767 of 1460) of patients did not call for medical aid themselves but relied on their spouse (373 [25.5%]), another relative or close acquaintance (262 [17.9%]), a bystander (63 [4.3%]), or staff at a care home (61 [4.2%]). Seventy-five of 1460 patients (5.1%) did not report symptoms to a health care professional until a routine appointment with a physician. The proportion of patients seeking medical attention themselves was similar for TIA and minor stroke and was unaltered before vs after April 1, 2009.

staff at a care home (61 [4.2%]). Seventy-five of 1460 patients (5.1%) did not report symptoms to a health care professional until a routine appointment with a physician. The proportion of patients seeking medical attention themselves was similar for TIA and minor stroke and was unaltered before vs after April 1, 2009. Data on the initial diagnostic impression of relatives or bystanders were not collected. Reasons for Delay in Seeking Medical Attention Of 864 patients with TIA or minor stroke who delayed seeking medical attention more than 3 hours, 654 (75.7%) provided a reason, usually attributing symptoms to another (less harmful) cause (157 [24.0%]), being reassured by or awaiting improvement of symptoms (153 [23.4%]), not being worried about initial symptoms (145 [22.2%]), or not seeking medical attention because symptom onset occurred outside of office hours (50 [7.6%]) (eTable 4 in the Supplement). Reasons for delay were similar for TIA and minor stroke (eTable 4 in the Supplement) and were essentially unchanged before and after April 1, 2009. Discussion In contrast to major stroke, we found in this study that for TIA and minor stroke the UK nationwide televised FAST campaign has not improved patient use of EMS or patient delay in seeking medical attention. Moreover, the percentage of strokes that followed shortly after an initial TIA for which no medical attention was sought remained unchanged after the FAST campaign, representing approximately 100 potentially preventable strokes per 1 million inhabitants annually.

se of EMS or patient delay in seeking medical attention. Moreover, the percentage of strokes that followed shortly after an initial TIA for which no medical attention was sought remained unchanged after the FAST campaign, representing approximately 100 potentially preventable strokes per 1 million inhabitants annually. It is likely that many individuals experience transient neurological symptoms for which they do not seek medical attention and do not have stroke. The results of large randomly dialed telephone studies10,11 have suggested that approximately 15% of people in the general population had experienced transient neurological symptoms during the previous few years, such as weakness, numbness, or visual problems. Although we cannot determine the total number of unheeded TIAs in our study population, we found that the share of strokes preceded by symptoms suggestive of TIA in the community for which no medical attention is sought has remained unchanged after the FAST campaign, affirming that these ignored events are an important target for stroke prevention. The large number of patients with TIA during the weekend who seek medical attention only days later during office hours emphasizes that less severe or shorter-lasting symptoms are not thought to require immediate medical attention, with no reduction since the FAST campaign. Seventy-five percent (70 of 93) of patients with stroke after unheeded TIAs were not taking antithrombotic medication, highlighting the missed opportunity for secondary prevention.2,3,4 Given the high effectiveness of preventive strategies within the first hours and days after TIA,4 the influence of public education will depend largely on its ability to convince patients to take action within this crucial window of time.

ic medication, highlighting the missed opportunity for secondary prevention.2,3,4 Given the high effectiveness of preventive strategies within the first hours and days after TIA,4 the influence of public education will depend largely on its ability to convince patients to take action within this crucial window of time. Use of EMS has repeatedly been shown to be one of the most important factors in early hospital arrival after TIA and minor stroke.8,9,22,23 We observed no positive association of the FAST television campaign with presentation via EMS after TIA and minor stroke. Even when patients attributed their symptoms to stroke, still less than one-third (130 of 458) contacted EMS, which is in line with the findings of a 2012 UK survey.24 Studies14,15,16,17 have indicated a moderate influence of public education on presentation after stroke, but no published studies to date have assessed the effect of large public education campaigns on presentation after TIA and minor stroke. Compared with major stroke, one of the main differences we found in the response to TIA and minor stroke is the high association of symptom recognition with subsequent behavior in the latter. While severe symptoms require medical aid regardless of their cause, transient or minor symptoms may leave room for deliberation and attribution of symptoms to other (less worrisome) causes. A population-based survey in the United Kingdom recently showed that, despite recall of the FAST campaign, recognition and the response to hypothetical stroke scenarios did not improve.25 The decline in correct recognition of symptoms that we observed after extensive FAST-based public education campaigns suggests that patients may be falsely reassured when their symptoms do not match the more severe symptoms depicted in public education advertisements.26 The limited sensitivity of the FAST acronym for TIA and minor stroke (approximately 60% vs 95% for major stroke based on the Results section herein) may further explain the lack of positive influence of the campaign in this group of patients and emphasizes the need for adapted forms of public education.

vertisements.26 The limited sensitivity of the FAST acronym for TIA and minor stroke (approximately 60% vs 95% for major stroke based on the Results section herein) may further explain the lack of positive influence of the campaign in this group of patients and emphasizes the need for adapted forms of public education. Limitations Although we believe that our findings are valid and should guide future public education, there are some limitations. First, the retrospective diagnosis of unheeded TIAs preceding stroke is inevitably subjective and may be subject to recall bias. However, we applied the same diagnostic criteria that we apply when making a diagnosis acutely after a TIA, and we will have underestimated unheeded TIAs preceding major stroke in patients who were unable to report these because of drowsiness, dysphasia, or dementia. Moreover, although we allowed a 90-day cutoff for preceding TIA to limit recall bias of distant events, most TIAs occurred within the week before the stroke and mirrored previous prospective investigations of the natural history of stroke risk after TIA.1 Second, we did not record the diagnostic perception of bystanders or relatives, nor did we directly ask about individual exposure to and awareness of the campaign. Third, increased but slow presentation of otherwise nonpresenters after the FAST campaign may have biased time trends of TIA and minor stroke toward the null. Fourth, in view of the small fraction of people in our study who were of nonwhite race/ethnicity, our results may not be fully applicable to racial/ethnic minorities, who have previously been reported to delay to presentation longer.8

he FAST campaign may have biased time trends of TIA and minor stroke toward the null. Fourth, in view of the small fraction of people in our study who were of nonwhite race/ethnicity, our results may not be fully applicable to racial/ethnic minorities, who have previously been reported to delay to presentation longer.8 Conclusions In our study, a lack of improvement in patient response to TIA and minor stroke was found after extensive FAST-based public education television campaigns in the United Kingdom. This highlights the need for effective public education to be tailored to transient and minor stroke symptoms, as well as major stroke. Supplement. eFigure. Flow Chart of Case Ascertainment in OXVASC eTable 1. Baseline Characteristics of Patients With a First in Study TIA or Stroke eTable 2. Delay in Presentation and Choice of First Contact With Healthcare Services in Relation to Patients’ Perception of Symptoms in TIA and Minor Stroke eTable 3. Patient Perception of TIA and Minor Stroke Symptoms Before and After Initiation of the FAST Campaign eTable 4. Reasons for More Than 3 Hour Delay in Seeking Medical Attention Click here for additional data file.

Introduction Cerebral palsy (CP) is the umbrella term used for a heterogeneous group of etiologies that occur to a developing fetal or infant brain.1 People diagnosed as having CP have similar hallmark symptoms including issues with movement, coordination, posture, and balance.1,2 These motor disturbances can also be comorbid with other issues such as behavioral disturbance, cognitive difficulties, communication difficulties, sensory impairments, epilepsy, and intellectual disability (ID).1It is estimated that CP affects 2 to 3 in 1000 live births, and approximately 1 in 400 people in the United Kingdom live with CP.3,4 Because CP develops and is diagnosed in early childhood,5 it is often considered a pediatric condition. However, CP is a lifelong condition with the majority of children living into adulthood, depending on the severity of the condition and associated physical comorbidities.6,7 There is evidence that as individuals with CP transition into and throughout adulthood, there can be deterioration in physical functioning and a rise in secondary health conditions.8,9 The experience of aging with CP is therefore likely to be linked with different psychological, social, and medical issues than those experienced through having CP as a child. However, most evidence on mental health and CP is focused on children,10 which cannot be generalized to adult populations.

health conditions.8,9 The experience of aging with CP is therefore likely to be linked with different psychological, social, and medical issues than those experienced through having CP as a child. However, most evidence on mental health and CP is focused on children,10 which cannot be generalized to adult populations. Depression and anxiety are 2 of the most common mental illnesses in the general population,11 and there is substantial evidence that living with a long-term condition or disability is associated with a 2- to 3-fold increase in the likelihood of being diagnosed as having depression12,13 or anxiety.14 However, there is relatively little research specifically examining mental health outcomes in adults with CP, to our knowledge. Existing evidence indicates that 20% to 25% of adults with CP have clinically significant levels of depressive symptoms.15,16 Furthermore, a recent clinical study in 501 adults with CP from a US clinic found that 39% of patients met criteria for a diagnosis of anxiety disorder, and 31% met criteria for a diagnosis of major depression.17 A recent article that compared the prevalence of depression and anxiety in adults with CP to the general population found that adults with CP were more likely to experience depression, but not anxiety, than the general population.18 However, this work was cross sectional, and there is a need for longitudinal work to systematically investigate whether CP is associated with an increased risk of anxiety or depression in adulthood.

tion found that adults with CP were more likely to experience depression, but not anxiety, than the general population.18 However, this work was cross sectional, and there is a need for longitudinal work to systematically investigate whether CP is associated with an increased risk of anxiety or depression in adulthood. However, the association between depression, anxiety, and CP could be modified by comorbid ID. Approximately one-third of individuals with CP also experience comorbid ID2; however, ID has been associated with difficulties correctly identifying common mental illness owing to diagnostic overshadowing.19 Diagnostic overshadowing can happen in individuals with ID because distress (including anxiety and depression) can present as challenging behaviors instead of the symptoms we typically associate with depression and anxiety.20,21 In other words, the challenging behaviors overshadow the correct diagnosis of mental illness. This could mean that although we might expect the incidence of anxiety and depression to be higher in adults with CP, it is also important to account for the presence of ID. The aim of this article was to determine the incidence of depression and anxiety in adults with CP compared with age-, sex-, and general practice–matched controls using primary care data from the United Kingdom. We also sought to determine whether the presence of comorbid ID impacted on the associations between CP and incident depression and anxiety.

etermine the incidence of depression and anxiety in adults with CP compared with age-, sex-, and general practice–matched controls using primary care data from the United Kingdom. We also sought to determine whether the presence of comorbid ID impacted on the associations between CP and incident depression and anxiety. Methods Clinical Practice Research Datalink Database Data for this study were taken from the Clinical Practice Research Datalink (CPRD) primary care database. This database reflects the collection of consultation data from consenting general practices throughout the United Kingdom, and it covers 6.9% of the UK population with active data available for 4.4 million people.22 Previous work has shown the population contained within the CPRD database are representative of the UK population.22 Collected data include clinical events, prescriptions, referrals, and hospital admissions. Formal data collection commenced in 1987, and data for this study were collected from January 1, 1987, to November 30, 2015. Data of entry into the study ranged from January 1987 to September 2015. Clinical Practice Research Datalink obtained ethical approval from a National Research Ethics Service Committee, which allows researchers to access anonymized data for observational studies on the approval of a protocol to an Independent Scientific Advisory Committee. Patient consent was waived.

1987 to September 2015. Clinical Practice Research Datalink obtained ethical approval from a National Research Ethics Service Committee, which allows researchers to access anonymized data for observational studies on the approval of a protocol to an Independent Scientific Advisory Committee. Patient consent was waived. Participants Any patient 18 years or older who had at least 1 record of CP during the study period, and during a period in which their data were considered research standard (ie, their data were determined by CPRD to be of sufficient quality for research assessment), was included as a case of CP. Clinical Practice Research Datalink checks data to ensure they are research standard by determining whether the patient-level data consist of valid registration dates and that the data provided by the practice have been continuous.22

RD to be of sufficient quality for research assessment), was included as a case of CP. Clinical Practice Research Datalink checks data to ensure they are research standard by determining whether the patient-level data consist of valid registration dates and that the data provided by the practice have been continuous.22 Diagnoses were identified using Read codes, which are alphanumeric codes used in UK health care to reference a Read term that captures the reason for consultation (eg, the Read code F2B..00 refers to the Read term cerebral palsy) A diagnosis of CP was identified by 1 of 22 Read codes for CP that were created by the senior investigator (J.M.R.) and checked by other CP experts (eTable 1 in the Supplement). The index date, ie, the start of follow-up, was defined as the latest of1 the date that the patient registered with their general practitioner (GP),2 the date that their data became research standard,3 or the year in which they turned age 18 years. Initially, 14 788 patients with CP were identified. Following exclusions based on age (<18 years, n = 2510), a Read code for CP not occurring within the study period and/or a period when data were research standard (n = 10 038), and potentially inaccurate codes (n = 535), a sample of 1705 patients with CP were included within the main analysis.

with CP were identified. Following exclusions based on age (<18 years, n = 2510), a Read code for CP not occurring within the study period and/or a period when data were research standard (n = 10 038), and potentially inaccurate codes (n = 535), a sample of 1705 patients with CP were included within the main analysis. Each patient with CP was matched to 3 patients without CP for age (within ±3 years), sex, and practice. Practice was used as an indicator of area-level socioeconomic status, as CPRD uses area-level deprivation as an indicator of socioeconomic position.23 In total, 5115 matched controls were identified, and these matched controls acted as the reference (comparison) group for our main analyses. The index date for each patient without CP was set as the same date as their respective matched CP case. In additional analyses, we examined the association of ID comorbidity with the incidence of anxiety and depression in adults with CP. These cases were identified using a list of Read codes provided by the Cambridge primary care unit.24 For these additional analyses, we split the CP group into patients with CP with no ID comorbidity (CPnoID) and patients with CP and ID (CP+ID).

omorbidity with the incidence of anxiety and depression in adults with CP. These cases were identified using a list of Read codes provided by the Cambridge primary care unit.24 For these additional analyses, we split the CP group into patients with CP with no ID comorbidity (CPnoID) and patients with CP and ID (CP+ID). Identification of Depression and Anxiety Cases of depression and anxiety were identified using the Read codes developed by John et al25 and the Cambridge primary care unit.24 Where the codes could refer to a possible case of depression (eg, mood disorders, depression-related symptoms) or anxiety (eg, worrying), these were only considered a case of depression in which the patient was also given antidepressants and/or anxiolytic medication (eTable 1 in the Supplement includes a list of Read codes). We excluded those codes that referred to a history of depression or anxiety, depression, or anxiety remission, interim reviews, or medication reviews (as we wanted to ascertain the first event of depression and anxiety after the index date). The date of the first event of depression was identified following the index date. Where no event of depression occurred, participants were followed up until the earliest of the following and treated as censored: transfer out of CPRD, death, or end of follow-up period. This was repeated for anxiety.

er the index date). The date of the first event of depression was identified following the index date. Where no event of depression occurred, participants were followed up until the earliest of the following and treated as censored: transfer out of CPRD, death, or end of follow-up period. This was repeated for anxiety. Identification of Intellectual Disability Read codes developed by the Cambridge primary care unit24 were used to identify individuals with ID. These Read codes included specific conditions associated with ID such as fragile X syndrome and Down syndrome alongside other synonyms for “learning disability” and ”intellectual disability.”24

f Intellectual Disability Read codes developed by the Cambridge primary care unit24 were used to identify individuals with ID. These Read codes included specific conditions associated with ID such as fragile X syndrome and Down syndrome alongside other synonyms for “learning disability” and ”intellectual disability.”24 Potential Confounders As other chronic conditions and GP visits could be linked with an increased likelihood of detecting incident anxiety and depression, we adjusted our analyses for these confounders. We used the Read codes proposed by the Cambridge primary care unit24 to identify the following chronic conditions: heart disease (myocardial infarction, coronary heart disease, and/or arrhythmia), lung disease (chronic obstructive pulmonary disease, chronic bronchitis, and/or asthma), pain conditions (4 or more prescriptions of pain medication), epilepsy, and diabetes. We also identified Read codes for osteoarthritis based on International Classification of Diseases, Ninth Revision, and International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, Read terms. Finally, we identified the mean number of GP visits per year, (0-2 visits per year, 2-11.9 visits per year, or ≥12 visits per year). We used 12 visits per year as an indicator for frequent GP consultations as has been done in previous work.26 For each potential confounder, we identified only those cases that occurred before the event date (ie, depression diagnosis or anxiety diagnosis) or the date of censoring.

its per year, or ≥12 visits per year). We used 12 visits per year as an indicator for frequent GP consultations as has been done in previous work.26 For each potential confounder, we identified only those cases that occurred before the event date (ie, depression diagnosis or anxiety diagnosis) or the date of censoring. Statistical Analysis To ascertain risk, we used stratified Cox proportional hazards regression for depression and anxiety as outcomes for the group with CP compared with the matched reference group. These were first run unadjusted and then adjusted for potential confounders. In our stratified analysis, we then reran all analyses with the CP group stratified by the presence of comorbid ID. For these analyses, we compared CPnoID and CP+ID with their respective matched reference groups. Prior to running our analyses, we checked the assumption of proportional hazards by plotting scaled Schoenfeld residuals against time for all models and found this assumption was satisfied. All analyses were conducted using Stata, version 14.0 (StataCorp). Results Descriptive Data We present descriptive data pertaining to the sociodemographic and health-related characteristics of the sample in Table 1. The mean (SD) age of the sample was 33.3 (15.5) years, and 798 individuals (46.8%) in both the group with CP and matched controls were women. The data also show that individuals with CP had a higher frequency of attending the physician, a higher frequency of epilepsy, and a higher frequency of pain conditions than the matched controls.

of the sample was 33.3 (15.5) years, and 798 individuals (46.8%) in both the group with CP and matched controls were women. The data also show that individuals with CP had a higher frequency of attending the physician, a higher frequency of epilepsy, and a higher frequency of pain conditions than the matched controls. Table 1. Characteristics of Participants Variable No. (%) No CP (n = 5115) CP (n = 1705) CPnoID (n = 1342) CP+ID (n = 363) Age, y <30 2631 (51.4) 877 (51.4) 691 (51.5) 186 (51.2) 30-39 1008 (19.7) 335 (19.7) 252 (18.8) 84 (23.1) 40-49 669 (13.1) 223 (13.1) 172 (12.8) 51 (14.1) 50-59 405 (7.9) 135 (7.9) 108 (8.1) 27 (7.4) ≥60 402 (7.9) 134 (7.9) 119 (8.9) 15 (4.1) Sex Male 2721 (53.2) 907 (53.2) 703 (52.8) 204 (56.2) Female 2394 (46.8) 798 (46.8) 639 (47.6) 159 (43.8) Region North 1419 (27.7) 473 (27.7) 349 (26.0) 124 (34.2) Midlands 1809 (35.4) 603 (35.4) 498 (37.1) 105 (28.9) South 1761 (34.4) 587 (34.4) 459 (34.0) 128 (35.3) Northern Ireland 126 (2.5) 42 (2.5) 36 (2.7) 6 (1.7) GP visits per year, mean (SD) 0-1.9 715 (14.0) 133 (7.8) 123 (9.2) 10 (2.8) 2-11.9 4035 (79.0) 1178 (69.3) 956 (71.5) 222 (61.3) ≥12 358 (7.0) 388 (22.8) 258 (19.3) 130 (35.9) Presence of Depression 867 (17.0) 312 (18.3) 263 (19.6) 49 (13.5) Anxiety 697 (13.6) 261 (15.3) 216 (16.1) 45 (12.4) Diabetes 218 (4.3) 55 (3.2) 47 (3.5) 14 (3.9) Heart disease 584 (11.4) 160 (9.4) 165 (12.3) 20 (5.5) Osteoarthritis 320 (6.3) 87 (5.1) 90 (6.7) 15 (4.1) Epilepsy 49 (1.0) 354 (20.8) 216 (16.1) 155 (42.7) Lung disease 379 (7.4) 147 (8.6) 143 (10.7) 26 (7.2) Pain conditions 238 (4.7) 166 (9.7) 179 (13.3) 23 (6.3) Abbreviations: CP, cerebral palsy; CPnoID, cerebral palsy with no comorbid ID; CP+ID: cerebral palsy with comorbid ID; GP, general practitioner; ID, intellectual disability.

1.0) 354 (20.8) 216 (16.1) 155 (42.7) Lung disease 379 (7.4) 147 (8.6) 143 (10.7) 26 (7.2) Pain conditions 238 (4.7) 166 (9.7) 179 (13.3) 23 (6.3) Abbreviations: CP, cerebral palsy; CPnoID, cerebral palsy with no comorbid ID; CP+ID: cerebral palsy with comorbid ID; GP, general practitioner; ID, intellectual disability. Risk of Depression During the follow-up period, there were 1179 new events of depression following the index date. In total, 867 people (17.0%) from the reference group had a new event of depression during a median of 9.1 (range, 0.01-28.01) years of follow-up. In total, 312 patients (18.3%) with CP had a new event of depression during a median of 5.7 (range, 0.001-27.9) years of follow-up (Table 2). The unadjusted Cox model indicated that patients with CP had an increased hazard of depression compared with matched patients without CP (hazard ratio [HR], 1.43; 95% CI, 1.24-1.64). This association remained statistically significant (HR, 1.28; 95% CI, 1.09-1.51) after controlling for other chronic conditions and the mean number of GP visits (Table 2).

that patients with CP had an increased hazard of depression compared with matched patients without CP (hazard ratio [HR], 1.43; 95% CI, 1.24-1.64). This association remained statistically significant (HR, 1.28; 95% CI, 1.09-1.51) after controlling for other chronic conditions and the mean number of GP visits (Table 2). Table 2. Incidence of Depression and Anxiety in 1705 Individuals With CP Compared With 5115 Age-, Sex-, and Practice-Matched Controls Variable Events No. (%) Person-Years in 10 000s Incidence Per 10 000 Person-Years (95% CI) Unadjusted Adjusted Hazards Ratio (95% CI) P Value HR (95% CI)a P Value Depression No CP 867 (17.0) 49.93 0.017 (0.016-0.019) 1 [Reference] NA 1 [Reference] NA CP 312 (18.3) 12.64 0.025 (0.022-0.028) 1.43 (1.24-1.64) <.001 1.28 (1.09-1.51) .003 Anxiety No CP 697 (13.6) 51.67 0.013 (0.013-0.015) 1 [Reference] NA 1 [Reference] NA CP 261 (15.3) 12.93 0.020 (0.018-0.023) 1.40 (1.21-1.63) <.001 1.38 (1.15-1.64) <.001 Abbreviations: CP, cerebral palsy; NA, not applicable. a Adjusted for baseline (ie, predepression or preanxiety diagnosis) diagnosis of diabetes, heart disease, lung disease, osteoarthritis, epilepsy, pain conditions, and general practitioner visits per year.

Table 2. Incidence of Depression and Anxiety in 1705 Individuals With CP Compared With 5115 Age-, Sex-, and Practice-Matched Controls Variable Events No. (%) Person-Years in 10 000s Incidence Per 10 000 Person-Years (95% CI) Unadjusted Adjusted Hazards Ratio (95% CI) P Value HR (95% CI)a P Value Depression No CP 867 (17.0) 49.93 0.017 (0.016-0.019) 1 [Reference] NA 1 [Reference] NA CP 312 (18.3) 12.64 0.025 (0.022-0.028) 1.43 (1.24-1.64) <.001 1.28 (1.09-1.51) .003 Anxiety No CP 697 (13.6) 51.67 0.013 (0.013-0.015) 1 [Reference] NA 1 [Reference] NA CP 261 (15.3) 12.93 0.020 (0.018-0.023) 1.40 (1.21-1.63) <.001 1.38 (1.15-1.64) <.001 Abbreviations: CP, cerebral palsy; NA, not applicable. a Adjusted for baseline (ie, predepression or preanxiety diagnosis) diagnosis of diabetes, heart disease, lung disease, osteoarthritis, epilepsy, pain conditions, and general practitioner visits per year. After stratifying for the presence of comorbid ID in individuals with CP, we found that there were 264 new cases (19.6%) of depression in the CPnoID group during a median of 5.2 (range, 0.05-27.9) years of follow-up. However, there were 48 incident cases (13.3%) of depression in the CP+ID group during a median of 7.4 (range, 0.01-26.1) years of follow-up. When we conducted Cox regression analysis, stratified according to presence of ID, we found that the CPnoID group had an increased adjusted hazards of incident depression compared with their matched reference group (adjusted HR, 1.44; 95% CI, 1.20-1.72) (Table 3). For both unadjusted and adjusted analyses, the CP+ID group had no difference in their hazard of incident depression compared with their matched reference group (Table 3). For Kaplan-Meier plots, see the eFigure in the Supplement.

pared with their matched reference group (adjusted HR, 1.44; 95% CI, 1.20-1.72) (Table 3). For both unadjusted and adjusted analyses, the CP+ID group had no difference in their hazard of incident depression compared with their matched reference group (Table 3). For Kaplan-Meier plots, see the eFigure in the Supplement. Risk of Anxiety During the follow-up period, there were 958 new events of anxiety after the index date. In total, 697 people (13.6%) from the reference group had a new event of anxiety during a median of 9.5 (range, 0.01-28.0) years of follow-up and 261 patients (15.3%) with CP had a new event of anxiety during a median of 6.0 (range, 0.003-27.9) years of follow-up (Table 3). The unadjusted HR indicated an increased risk of anxiety for individuals with CP compared with the matched reference group (HR, 1.40; 95% CI, 1.21-1.63). This increased risk persisted after controlling for other chronic conditions and the mean number of GP visits (HR, 1.38; 95% CI, 1.15-1.64) (Table 3).

w-up (Table 3). The unadjusted HR indicated an increased risk of anxiety for individuals with CP compared with the matched reference group (HR, 1.40; 95% CI, 1.21-1.63). This increased risk persisted after controlling for other chronic conditions and the mean number of GP visits (HR, 1.38; 95% CI, 1.15-1.64) (Table 3). Table 3. Incidence of Depression and Anxiety in Individuals With CP With and Without Comorbid IDa Variableb Events, No. (%) Person-Years in 10 000s Incidence Per 10 000 Person-Years (95% CI) Unadjusted Adjusted Hazard Ratio (95% CI) P Value Hazard Ratio (95% CI)c P Value Depression Matched reference group 687 (17.0) 39.32 0.017 (0.016-0.019) 1 [Reference] NA 1 [Reference] NA CPnoID 264 (19.6) 9.55 0.028 (0.025-0.031) 1.59 (1.36-1.85) <.001 1.44 (1.20-1.72) <.001 Depression Matched reference group 180 (16.6) 10.61 0.017 (0.015-0.020) 1 [Reference] NA 1 [Reference] NA CP+ID 48 (13.3) 3.10 0.015 (0.012-0.021) 0.92 (0.66-1.29) .66 0.68 (0.43-1.07) .09 Anxiety Matched reference group 542 (13.4) 40.72 0.013 (0.012-0.014) 1 [Reference] NA 1 [Reference] NA CPnoID 217 (16.2) 9.79 0.022 (0.019-0.025) 1.57 (1.32-1.85) <.001 1.55 (1.28-1.87) <.001 Anxiety Matched reference group 155 (14.3) 10.94 0.014 (0.012-0.017) 1 [Reference] NA 1 [Reference] NA CP+ID 44 (12.2) 3.15 0.014 (0.010-0.019) 0.92 (0.65-1.30) .65 0.77 (0.48-1.25) .29 Abbreviations: CP, cerebral palsy; CPnoID, cerebral palsy with no comorbid ID; CP+ID, cerebral palsy with comorbid ID; ID, intellectual disability; NA, not applicable.

55 (14.3) 10.94 0.014 (0.012-0.017) 1 [Reference] NA 1 [Reference] NA CP+ID 44 (12.2) 3.15 0.014 (0.010-0.019) 0.92 (0.65-1.30) .65 0.77 (0.48-1.25) .29 Abbreviations: CP, cerebral palsy; CPnoID, cerebral palsy with no comorbid ID; CP+ID, cerebral palsy with comorbid ID; ID, intellectual disability; NA, not applicable. a Each CP group was compared with their respective age-, sex-, and practice-matched reference group. There were 1342 individuals in the CPnoID group and 4026 in their respective matched group. There were 363 individuals in the CP+ID group and 1089 in their respective matched group. b CPnoID compared with respective matched controls and CP+ID compared with respective matched controls. c Adjusted for baseline (ie, predepression or preanxiety diagnosis) diagnosis of diabetes, heart disease, lung disease, osteoarthritis, epilepsy, pain conditions, and general practitioner visits per year.

a Each CP group was compared with their respective age-, sex-, and practice-matched reference group. There were 1342 individuals in the CPnoID group and 4026 in their respective matched group. There were 363 individuals in the CP+ID group and 1089 in their respective matched group. b CPnoID compared with respective matched controls and CP+ID compared with respective matched controls. c Adjusted for baseline (ie, predepression or preanxiety diagnosis) diagnosis of diabetes, heart disease, lung disease, osteoarthritis, epilepsy, pain conditions, and general practitioner visits per year. After stratifying for comorbid ID, we found that in the CPnoID group, there were 217 new cases (16.2%) of incident anxiety during a median of 5.6 (range, 0.003-27.9) years of follow-up (Table 3). In the CP+ID group, there were 44 cases (12.2%) of incident anxiety during a median of 7.7 (range, 0.06-26.1) years of follow-up (Table 3). The Cox regression analysis indicated that the CPnoID group had an increased adjusted hazards of incident anxiety compared with their matched reference group (HR, 1.55; 95% CI, 1.28-1.87) (Table 3). However, the CPnoID group had no difference in their unadjusted or adjusted hazards of experiencing incident anxiety compared with their matched reference group (Table 3). For Kaplan-Meier plots, see the eFigure in the Supplement.

iety compared with their matched reference group (HR, 1.55; 95% CI, 1.28-1.87) (Table 3). However, the CPnoID group had no difference in their unadjusted or adjusted hazards of experiencing incident anxiety compared with their matched reference group (Table 3). For Kaplan-Meier plots, see the eFigure in the Supplement. Sensitivity Analysis As a total of 24 participants in the no-CP group also had ID, which could affect estimates, we reran all analyses excluding these participants from our ID-stratified analyses (eTable 2 in the Supplement). Removal of individuals with ID from the control group did not have a substantial effect on the results (eTable 2 in the Supplement).

participants in the no-CP group also had ID, which could affect estimates, we reran all analyses excluding these participants from our ID-stratified analyses (eTable 2 in the Supplement). Removal of individuals with ID from the control group did not have a substantial effect on the results (eTable 2 in the Supplement). Discussion Results from this study indicated that individuals with CP had an increased risk of being diagnosed as having depression or anxiety, compared with a matched control group of adults without CP. These results could have been observed because adults with CP present with many physiological, psychological, social, and health-related risk factors that have been shown to be associated with depression and anxiety in the general population such as multimorbidity,27,28,29 increased pain,30,31,32 functional limitations,2,33 noncommunicable diseases,12,14,34,35 difficulties with social relationships36,37 and poorer sleep.38,39 Furthermore, when we examine work that has been conducted in adults with CP, depressive symptoms are associated with fatigue15 and pain.15,40 However, we need more research to determine why individuals with CP may have a higher risk of depression and anxiety so that we may develop the evidence base for mental health interventions in this population. Our results also indicate that ID comorbidity should be considered when assessing the mental health of adults with CP. We found the risk of depression and anxiety was higher in the adults with CP who did not have ID, compared with the matched reference group. Furthermore, adults with CP and ID had similar hazards of developing depression and anxiety to the matched reference group. We could have observed these results as previous work suggests that diagnostic overshadowing may lead to an underdiagnosis of mental illness among people who have ID19 because distress can present as challenging behaviors.20,21 Thus, it is possible that GPs may not be as well trained in diagnosing depression and anxiety in these individuals. However, there is also work that has been conducted in populations with ID, indicating that the prevalence of anxiety and depression in individuals with ID is no different from the general population.21

1 Thus, it is possible that GPs may not be as well trained in diagnosing depression and anxiety in these individuals. However, there is also work that has been conducted in populations with ID, indicating that the prevalence of anxiety and depression in individuals with ID is no different from the general population.21 While our evidence suggests an increased risk of developing anxiety and depression in adults with CP compared with adults who do not have CP, it should be noted that there was little difference in absolute risk for developing anxiety or depression over the total follow-up period. This is the first study, to our knowledge, to compare the risk of depression and anxiety in adults with CP to adults who do not have CP using a population-based cohort. The results support previous work that shows a relatively high frequency of depression and anxiety in individuals with CP.15,17

While our evidence suggests an increased risk of developing anxiety and depression in adults with CP compared with adults who do not have CP, it should be noted that there was little difference in absolute risk for developing anxiety or depression over the total follow-up period. This is the first study, to our knowledge, to compare the risk of depression and anxiety in adults with CP to adults who do not have CP using a population-based cohort. The results support previous work that shows a relatively high frequency of depression and anxiety in individuals with CP.15,17 Limitations Owing to the nature of the population examined and the reliance on the presence of diagnostic codes to define outcomes, there is a possibility that any observed associations are underestimated as previous work suggests depression and anxiety diagnoses in primary care may be underestimated.41,42 This could also explain why the rates of depression and anxiety we observed in individuals with CP for this study were lower than other studies have previously reported.17,18 Furthermore, CP is an umbrella term used to describe heterogeneous etiologies; however, we could not account for the severity of issues associated with CP. These include gross motor function, communication issues, subtypes of motor abnormality, and fatigue. Future work should provide more insight into how CP-specific issues might be associated with mental health. In addition, the measure of pain conditions that we had within this study relied only on medications.

with CP. These include gross motor function, communication issues, subtypes of motor abnormality, and fatigue. Future work should provide more insight into how CP-specific issues might be associated with mental health. In addition, the measure of pain conditions that we had within this study relied only on medications. There are also additional caveats that should be borne in mind when interpreting these data. Depression and anxiety are considered chronic conditions that typically have their first onset in adolescence or early adulthood43; therefore, it is plausible to assume that some people may have had diagnoses of depression and anxiety prior to the index date for this study. In future work, it could be interesting to look at the life course of mental illness in individuals with CP, looking at people from adolescence through to older age. Conclusions This work provides evidence that adults with CP have an increased risk of developing depression and anxiety. Furthermore, comorbidity of ID is an important effect modifier that should be considered when examining the mental health of adults with CP. There is a need for more work to elucidate the causal mechanisms of poor mental health in adults with CP so that we may develop targeted interventions. Supplement. eTable 1. Read codes and associated Read terms used to define Cerebral Palsy eTable 2. Risk of depression and anxiety in individuals with CP with and without co-morbid ID (excluding all individuals without CP who had a diagnosis of ID, n=24) eFigure. Kaplan Meier survival plots Click here for additional data file.

Introduction Women who undergo bilateral salpingo-oophorectomy (BSO) before the onset of menopause have an accelerated accumulation of multimorbidity, with an increased risk of aging-associated neurological diseases, including dementia.1,2,3 Furthermore, surgical menopause at an early age was associated with Alzheimer disease (AD) pathology at autopsy.4 Because imaging biomarkers associated with cognitive impairment and dementia precede the clinical symptoms, and may provide insight into the underlying causative mechanisms of cognitive impairment and dementia later in life, we investigated β-amyloid deposition (primary outcome), magnetic resonance imaging (MRI)–based biomarkers of medial temporal lobe neurodegeneration, and white matter hyperintensity volume in women who underwent BSO before age 50 years and before reaching natural menopause.

gnitive impairment and dementia later in life, we investigated β-amyloid deposition (primary outcome), magnetic resonance imaging (MRI)–based biomarkers of medial temporal lobe neurodegeneration, and white matter hyperintensity volume in women who underwent BSO before age 50 years and before reaching natural menopause. Methods The Mayo Clinic Cohort Study of Oophorectomy and Aging-2 (MOA-2) is a population-based cohort study that includes women who underwent BSO before age 50 years and before reaching natural menopause from 1988 through 2007 and an age-matched control group who had not undergone bilateral oophorectomy before age 50 years.5 The Mayo Clinic Study of Aging (MCSA) is another population-based cohort study, which includes participants with normal cognitive aging, mild cognitive impairment, and dementia.6 All MCSA participants are invited to undergo MRI and Pittsburgh compound B (PiB) positron emission tomography (PET). Both study cohorts are representative of the geographically defined population of Olmsted County, Minnesota. In the current study, women with BSO and control-participant women from the MOA-2 cohort who later were enrolled in the MCSA and underwent a neuropsychological evaluation, MRI, and PiB-PET scan were included (Figure 1). This study was approved by the Mayo Clinic institutional review board, and written informed consent was obtained from all participants.

O and control-participant women from the MOA-2 cohort who later were enrolled in the MCSA and underwent a neuropsychological evaluation, MRI, and PiB-PET scan were included (Figure 1). This study was approved by the Mayo Clinic institutional review board, and written informed consent was obtained from all participants. Figure 1. Study Sample BSO indicates bilateral salpingo-oophorectomy; MCSA, Mayo Clinic Study of Aging; MOA-2, Mayo Clinic Cohort Study of Oophorectomy and Aging-2; MRI, magnetic resonance imaging; PET, positron emission tomography; PiB, Pittsburgh compound B.

O and control-participant women from the MOA-2 cohort who later were enrolled in the MCSA and underwent a neuropsychological evaluation, MRI, and PiB-PET scan were included (Figure 1). This study was approved by the Mayo Clinic institutional review board, and written informed consent was obtained from all participants. Figure 1. Study Sample BSO indicates bilateral salpingo-oophorectomy; MCSA, Mayo Clinic Study of Aging; MOA-2, Mayo Clinic Cohort Study of Oophorectomy and Aging-2; MRI, magnetic resonance imaging; PET, positron emission tomography; PiB, Pittsburgh compound B. Global cognitive status was assessed using the short test of mental status, and combining 9 tests into a global cognitive status score.6 The MRI studies were performed at 3-T (GE Healthcare). Previously described and validated MRI analysis methods were used.7,8 Hippocampal and amygdala volumes were adjusted for total intracranial volume. Parahippocampal-entorhinal cortical thickness was measured using FreeSurfer version 5.3 (Athinoula A. Martinos Center for Biomedical Imaging at Massachusetts General Hospital).9 A semi-automated segmentation10 of fluid-attenuated inversion recovery (FLAIR) images was used for white matter hyperintensity volume quantification and adjusted for total intracranial volume. Cerebral infarcts were also evaluated. Previously described and validated methods were used to process diffusion tensor imaging scans.11 Entorhinal white matter fractional anisotropy, which includes the perforant pathway, was quantified using the Johns Hopkins University atlas.12 A PET and computed tomography scanner with 3-dimensional mode (GE Healthcare) was used for PET imaging, and image analysis was performed using an automated image processing pipeline.8 Cortical β-amyloid deposition on PiB-PET scan was calculated using standard uptake value ratio.

ng the Johns Hopkins University atlas.12 A PET and computed tomography scanner with 3-dimensional mode (GE Healthcare) was used for PET imaging, and image analysis was performed using an automated image processing pipeline.8 Cortical β-amyloid deposition on PiB-PET scan was calculated using standard uptake value ratio. Statistical Analysis Wilcoxon rank sum tests, rank regression tests adjusted for total intracranial volume, and Fisher exact tests were used for comparisons of the BSO and control groups. Correlations between imaging biomarkers and global cognitive status scores were assessed using Spearman correlations, and the P values were adjusted for multiple comparisons using the false-discovery rate. Statistical significance was considered at the 2-sided α level of .05.

comparisons of the BSO and control groups. Correlations between imaging biomarkers and global cognitive status scores were assessed using Spearman correlations, and the P values were adjusted for multiple comparisons using the false-discovery rate. Statistical significance was considered at the 2-sided α level of .05. Results A total of 43 women fulfilled the inclusion criteria, including 23 who had undergone BSO and 20 control participants. The median age at oophorectomy in the BSO group was 46 (interquartile range [IQR], 45-48) years. Age at imaging and frequency of APOE ε4 carriers did not differ between the groups. Among the 23 women who underwent BSO with hysterectomy, 22 (96%) took unopposed estrogen for a median duration of 10 (IQR, 5-13) years after surgery. Oral conjugated equine estrogen was the most common type used (n = 15 of 22 [68%]), typically at a dosage of 0.625 mg/d. Among the 20 control-participant women, 19 (95%) reached menopause during the study follow-up, 17 (85%) had natural menopause, and 10 of 19 (53%) took estrogen for a median duration of 10 (IQR, 6-16) years. The most common type used was oral conjugated equine estrogen at a dosage of 0.625 mg/d (n = 9 of 10 [90%]) with progestin (oral medroxyprogesterone acetate at 2.5 mg/d; n = 8 of 19 [42%]) for a median duration of 9 (IQR, 4-13) years. Although the frequency of mild cognitive impairment diagnosis was slightly higher in the BSO group (n = 3 [13%]) compared with the control group (n = 1 [5%]; P = .40), the short test of mental status, global cognitive status scores, and the Beck depression and anxiety inventory scores did not differ between the groups (Table).

the frequency of mild cognitive impairment diagnosis was slightly higher in the BSO group (n = 3 [13%]) compared with the control group (n = 1 [5%]; P = .40), the short test of mental status, global cognitive status scores, and the Beck depression and anxiety inventory scores did not differ between the groups (Table). Table. Demographics, Global Cognitive Status, and Imaging Characteristics of the Study Samplea Characteristic No. (%) P Value Bilateral Salpingo-oophorectomy (n = 23) Control Group (n = 20) Demographic and clinical characteristic Age at oophorectomy, median (IQR), y 46 (45-48) NA NA Time from oophorectomy to imaging, median (IQR), y 19 (17-22) NA NA Age at imaging, median (IQR), y 65 (62-68) 63 (60-66) .23 Length of Olmsted County residence at the time of MCSA baseline, median (IQR), y 41 (34-44) 36 (34-41) .56 Women who reached menopause 23 (100) 19 (95) .47 Hormone therapy with estrogenb 22 (96) 10 (53) .002 Estrogen therapy, median (IQR), yc 10 (5-13) 10 (6-16) .76 Oral conjugated equine estrogenc 15 (68) 9 (90) .38 Hormone therapy with progestinb NA 8 (42) NA Progestin therapy, median (IQR), yd NA 9 (4-13) NA Oral medroxyprogesterone acetated NA 8 (100) NA BMI, median (IQR) 30 (27-34) 29 (25-31) .22 Smokinge 5 (36) 9 (56) .30 Education, median (IQR), y 14 (13-16) 16 (13-18) .62 APOE ε4 carriers 2 (9) 4 (20) .40 Neuropsychological evaluation scores Short test of mental status, median (IQR) 37 (35-37) 37 (35-38) .20

Table. Demographics, Global Cognitive Status, and Imaging Characteristics of the Study Samplea Characteristic No. (%) P Value Bilateral Salpingo-oophorectomy (n = 23) Control Group (n = 20) Demographic and clinical characteristic Age at oophorectomy, median (IQR), y 46 (45-48) NA NA Time from oophorectomy to imaging, median (IQR), y 19 (17-22) NA NA Age at imaging, median (IQR), y 65 (62-68) 63 (60-66) .23 Length of Olmsted County residence at the time of MCSA baseline, median (IQR), y 41 (34-44) 36 (34-41) .56 Women who reached menopause 23 (100) 19 (95) .47 Hormone therapy with estrogenb 22 (96) 10 (53) .002 Estrogen therapy, median (IQR), yc 10 (5-13) 10 (6-16) .76 Oral conjugated equine estrogenc 15 (68) 9 (90) .38 Hormone therapy with progestinb NA 8 (42) NA Progestin therapy, median (IQR), yd NA 9 (4-13) NA Oral medroxyprogesterone acetated NA 8 (100) NA BMI, median (IQR) 30 (27-34) 29 (25-31) .22 Smokinge 5 (36) 9 (56) .30 Education, median (IQR), y 14 (13-16) 16 (13-18) .62 APOE ε4 carriers 2 (9) 4 (20) .40 Neuropsychological evaluation scores Short test of mental status, median (IQR) 37 (35-37) 37 (35-38) .20 z Scores, median (IQR) Global 1.51 (0.85-1.99) 1.59 (0.99-1.91) .78 Memory 1.57 (0.62-2.09) 1.53 (1.00-1.89) .73 Attention 1.43 (0.96-1.77) 1.37 (0.88-1.73) .79 Language 1.05 (0.17-1.34) 1.01 (0.73-1.52) .58 Visuospatial 0.73 (0.14-1.40) 1.19 (0.84-1.48) .28 Mild cognitive impairment 3 (13) 1 (5) .40 Beck Depression Inventory score, median (IQR) 4 (1-6) 4 (2-7) .52 Beck Anxiety Inventory score, median (IQR) 1 (0-4) 0 (0-3) .28 Neuroimaging biomarker measurements, median (IQR) White matter hyperintensity, cm3 5.46 (2.87-8.11) 4.13 (2.98-9.98) .52 Hippocampal volume, scoref −0.32 (−0.80 to 0.29) −0.05 (−0.29 to 0.30) .13 Amygdala volume, cm3 1.74 (1.59-1.91) 2.15 (2.05-2.37) <.001 Parahippocampal-entorhinal cortical thickness, mm 3.91 (3.64-4.00) 3.97 (3.89-4.28) .046 Entorhinal white matter fractional anisotropy 0.19 (0.18-0.22) 0.22 (0.20-0.23) .03 Global cortical Pittsburgh compound B standard uptake value ratio 1.29 (1.24-1.35) 1.25 (1.19-1.31) .17 Infarctions 1 (4) 1 (5) >.99 Abbreviations: BMI, body mass index (calculated as weight in kilograms divided by height in meters squared); IQR, interquartile range; MCSA, Mayo Clinic Study of Aging; NA, not applicable.

lobal cortical Pittsburgh compound B standard uptake value ratio 1.29 (1.24-1.35) 1.25 (1.19-1.31) .17 Infarctions 1 (4) 1 (5) >.99 Abbreviations: BMI, body mass index (calculated as weight in kilograms divided by height in meters squared); IQR, interquartile range; MCSA, Mayo Clinic Study of Aging; NA, not applicable. a Wilcoxon rank sum tests, rank regression tests adjusted for total intracranial volume, and Fisher exact tests were used to compare the bilateral salpingo-oophorectomy and control groups. b In 23 women who had undergone bilateral salpingo-oophorectomy and 19 control participants who reached menopause. c In 22 women who had undergone bilateral salpingo-oophorectomy and in 10 control participants who used hormone therapy with estrogen; the typical dosage was 0.625 mg/d. d In 8 control participants who used hormone therapy with progestin; the typical dosage was 2.5 mg/d. e In 14 women who had undergone bilateral salpingo-oophorectomy and in 16 control participants with known smoking status. f Hippocampal volume was calculated as raw right plus left hippocampal volumes, adjusted for total intracranial volume. To derive the total intracranial volume–adjusted hippocampal volume, we fit a linear regression model among 133 participants with normal cognitive function aged 30 to 59 years to predict hippocampal volume from total intracranial volume.13

raw right plus left hippocampal volumes, adjusted for total intracranial volume. To derive the total intracranial volume–adjusted hippocampal volume, we fit a linear regression model among 133 participants with normal cognitive function aged 30 to 59 years to predict hippocampal volume from total intracranial volume.13 On structural MRI, median (IQR) amygdala volume was smaller (BSO group, 1.74 [1.59-1.91] cm3; control group: 2.15 [2.05-2.37] cm3; P < .001), median (IQR) parahippocampal-entorhinal cortex was thinner (BSO group: 3.91 [3.64-4.00] mm; control group: 3.97 [3.89-4.28] mm; P = .046), and the entorhinal white matter fractional anisotropy on diffusion tensor imaging was lower (BSO group: 0.19 [0.18-0.22]; control group: 0.22 [0.20-0.23]; P = .03) in the BSO group compared with the control group (Table). Smaller hippocampal volume on MRI and higher cortical PiB standard uptake value ratio on PiB-PET scan were observed in the BSO group compared with the control group, but these results did not reach statistical significance (Figure 2). The results did not change noticeably in sensitivity analyses that added 4 women who had an MRI but not a PiB-PET scan, and in sensitivity analyses that removed 4 women who had mild cognitive impairment at the time of imaging testing (data not shown).

but these results did not reach statistical significance (Figure 2). The results did not change noticeably in sensitivity analyses that added 4 women who had an MRI but not a PiB-PET scan, and in sensitivity analyses that removed 4 women who had mild cognitive impairment at the time of imaging testing (data not shown). Figure 2. Imaging Characteristics in Women Who Underwent Bilateral Salpingo-Oophorectomy (BSO) vs Control Participants Box plots show median and interquartile ranges. Amygdala volume was smaller and parahippocampal-entorhinal cortex was thinner on structural magnetic resonance imaging, and entorhinal white matter fractional anisotrophy was lower on diffusion tensor imaging in women who had undergone BSO than in control participants. Hippocampal volume was calculated as raw right plus left hippocampal volumes, adjusted for total intracranial volume. To derive the total intracranial volume–adjusted hippocampal volume, we fit a linear regression model among 133 participants with normal cognitive function aged 30 to 59 years to predict hippocampal volume from total intracranial volume.13 There were no statistically significant differences in hippocampal volume, cortical Pittsburgh compound B (PiB) standard uptake value ratio, or white matter hyperintensity volume between women who had undergone BSO and control participants.

30 to 59 years to predict hippocampal volume from total intracranial volume.13 There were no statistically significant differences in hippocampal volume, cortical Pittsburgh compound B (PiB) standard uptake value ratio, or white matter hyperintensity volume between women who had undergone BSO and control participants. White matter hyperintensity volume and the frequency of infarctions did not differ between the groups. Imaging biomarkers were not associated with the global cognitive status score, and Beck Anxiety Inventory and Beck Depression Inventory scores after correcting for multiple comparisons using false-discovery rate. Discussion In this study, women who underwent BSO before age 50 years and before reaching natural menopause had smaller amygdala volumes, thinner parahippocampal-entorhinal cortices, and lower entorhinal white matter fractional anisotropy values compared with control-participant women.

White matter hyperintensity volume and the frequency of infarctions did not differ between the groups. Imaging biomarkers were not associated with the global cognitive status score, and Beck Anxiety Inventory and Beck Depression Inventory scores after correcting for multiple comparisons using false-discovery rate. Discussion In this study, women who underwent BSO before age 50 years and before reaching natural menopause had smaller amygdala volumes, thinner parahippocampal-entorhinal cortices, and lower entorhinal white matter fractional anisotropy values compared with control-participant women. There is an increased risk of cognitive impairment or dementia1 and presence of AD pathology4 in women who undergo BSO before menopause. However, imaging biomarkers associated with AD pathophysiology that precede cognitive impairment have not been studied in these women. Results of the present study suggest that women who underwent early BSO have biomarker abnormalities associated with neurodegeneration in the medial temporal lobe. In particular, the entorhinal cortex, which is one of the regions involved with neurofibrillary tangle pathology during the preclinical stages of AD, as well as primary age-associated taupathy.14 In addition, a thinner entorhinal cortex and lower fractional anisotropy on diffusion tensor imaging in the BSO group suggest a disruption in the entorhinal white matter microstructure that includes the perforant pathway carrying the connections between the entorhinal cortex and the hippocampus. Although not statistically significant in this small explorative sample, a difference in hippocampal volumes and cortical PiB uptake in the BSO group compared with the control group suggests the occurrence of early biomarker changes associated with AD pathophysiology.

between the entorhinal cortex and the hippocampus. Although not statistically significant in this small explorative sample, a difference in hippocampal volumes and cortical PiB uptake in the BSO group compared with the control group suggests the occurrence of early biomarker changes associated with AD pathophysiology. Premenopausal oophorectomy–induced estrogen deficiency is thought to be the primary cause of the increased risk of cognitive impairment or dementia in women with BSO before the onset of menopause.1 Furthermore, AD biomarker abnormalities have been observed more frequently in women undergoing menopause compared with premenopausal women, after controlling for age.15 In our study, 96% of the women with BSO were treated with estrogen (primarily oral conjugated equine estrogen), for a median of 10 years after BSO; however, this type and duration of hormonal treatment after BSO does not seem to be sufficient to prevent structural changes in the medial temporal lobe later in life. Further research into the type of estrogen treatments used, route of administration, dosing, and influence of the other ovarian hormones and hormones of the pituitary-ovarian axis is needed.

monal treatment after BSO does not seem to be sufficient to prevent structural changes in the medial temporal lobe later in life. Further research into the type of estrogen treatments used, route of administration, dosing, and influence of the other ovarian hormones and hormones of the pituitary-ovarian axis is needed. Limitations Lower hippocampal volume and higher cortical β-amyloid accumulation observed in the BSO group compared with the control group may have failed to reach statistical significance because of the small sample size. Because structural imaging biomarkers of medial temporal lobe neurodegeneration are associated with cognitive impairment and dementia later in life,16 findings of the current study support the association between BSO in premenopausal women and an increased risk of cognitive decline and dementia.1,4 Conclusions Abrupt hormonal changes because of BSO in premenopausal women may lead to medial temporal lobe structural abnormalities later in life. Because alterations in structural imaging biomarkers of neurodegeneration in the medial temporal lobe precede clinical symptoms of dementia, enlargement and longitudinal follow-up of this cohort is needed.

Introduction Down syndrome (DS) results from trisomy of chromosome 21 and is associated with multiple health and cognitive comorbidities including congenital heart defects and intellectual disability (ID).1 Fifty years ago, life expectancy for those with DS was just 10 years, with congenital heart defects responsible for most deaths within the first year.2 Medical advances have since increased mean survival to 63.5 years, yet people with DS still die a mean of 13 years before those without.3 Respiratory diseases are now the most frequently cited primary causes of death in adults with DS,2,3 with evidence that reduced mobility, poor vision, and epilepsy are each associated with reduced survival in later years.4 However, this increasing life expectancy in DS has also revealed an exceptional risk for developing dementia, driven by the near-universal neuropathology of Alzheimer disease (AD) by adulthood.5 Understanding how this disease burden is associated with mortality in this aging population is of primary importance for providing appropriate prognostic information, care, and research into potential AD treatments for those both with and without DS.

ear-universal neuropathology of Alzheimer disease (AD) by adulthood.5 Understanding how this disease burden is associated with mortality in this aging population is of primary importance for providing appropriate prognostic information, care, and research into potential AD treatments for those both with and without DS. Alzheimer disease neuropathology in DS stems from triplication of the amyloid precursor protein (APP) gene on chromosome 21.6,7 Recent mouse model research has found that triplication of other chromosome 21 genes can also increase amyloid-β deposition and worsen cognitive deficits,8 although these genes may also have different modulatory and protective roles in AD progression, as suggested by differences between DS-AD and non-DS familial early-onset AD caused by duplication of the APP locus alone.9 The dementia burden this causes in DS is striking: the mean age of dementia diagnosis is 55 years,10 and as many as 88% of this population can be expected to develop dementia by age 65 years.11 However, there is great variability in the age of dementia onset, with some individuals surviving past age 60 years with no clear signs of cognitive decline.12 Some dementia risk factors seen in the non-DS population similarly play a role in the variability of onset in those with DS. For example, possession of the apolipoprotein E (APOE) ε2 allele shows a protective influence, whereas APOE ε4 increases dementia risk.13 Demographic factors may also influence detection of cognitive decline by caregivers; there is evidence that those who remain living with their family are likely to receive a diagnosis earlier than those in other living situations.10

POE) ε2 allele shows a protective influence, whereas APOE ε4 increases dementia risk.13 Demographic factors may also influence detection of cognitive decline by caregivers; there is evidence that those who remain living with their family are likely to receive a diagnosis earlier than those in other living situations.10 Seizure development is closely linked with dementia in DS. Forty-three percent of those without a previous history of epilepsy develop seizures within a median of 2 years following dementia diagnosis, with most developing generalized tonic-clonic seizures or myoclonic jerks as dementia progresses.14 Long-standing epilepsy, present before dementia diagnosis, may shorten survival time after a diagnosis of dementia in individuals with DS,10 and there is also evidence that taking antidementia mediation can extend survival.10 Although it has been reported that 20 times more people with DS have dementia recorded as a contributory factor on their death certificate than those without DS,15 further studies are required to quantify the association that dementia has directly with mortality risk in those with DS, as well as exploring factors that may modify mortality and dementia risk in this population. Better information about factors associated with dementia onset and prognosis will also support the development of clinical trials of treatments.

the association that dementia has directly with mortality risk in those with DS, as well as exploring factors that may modify mortality and dementia risk in this population. Better information about factors associated with dementia onset and prognosis will also support the development of clinical trials of treatments. This study aimed to examine the effect of dementia on crude mortality rates (CMRs) in a large, representative cohort of older individuals with DS in the United Kingdom. Secondary analyses were used to evaluate the influence of additional health and demographic factors on age at death and at dementia diagnosis. Methods Study Design and Setting Data were acquired as part of a large, prospective longitudinal study of cognition and health in adults with DS in the United Kingdom.16 Ethical approval was secured from the North-West Wales Research Ethics Committee (13/WA/0194). Participation in the primary study was open to all adults with DS, regardless of capacity to consent. Capacity was assessed for each participant at each time, and written informed consent was obtained from all those who were able. A consultee (typically a family member or paid carer) was appointed for individuals without capacity. The consultee was asked to sign a form to indicate their decision about the individual’s inclusion based on their knowledge of the individual and their wishes, in accordance with the UK Mental Capacity Act 2005.

able. A consultee (typically a family member or paid carer) was appointed for individuals without capacity. The consultee was asked to sign a form to indicate their decision about the individual’s inclusion based on their knowledge of the individual and their wishes, in accordance with the UK Mental Capacity Act 2005. Participants Participants were recruited from DS support groups, care homes, existing participant databases, and National Health Services sites in England. Down syndrome status was confirmed genetically where possible (n = 193 of 211 successfully karyotyped). To be eligible for inclusion, participants were required to be 36 years or older at study entry, to have at least 2 data points (mean length of follow-up, 28.66 months; range, 1-65 months). and to have their clinical dementia status known to the informant. Data Sources/Measurements Data for all variables were collected as part of a prospective, longitudinal study. Medical history and demographic details were acquired through a semistructured interview with a carer who knew the participant well. APOE genotype was confirmed via blood or saliva sample using a Thermo Fisher Scientific Taqman assay for single-nucleotide polymorphisms rs7412 and rs429358.

spective, longitudinal study. Medical history and demographic details were acquired through a semistructured interview with a carer who knew the participant well. APOE genotype was confirmed via blood or saliva sample using a Thermo Fisher Scientific Taqman assay for single-nucleotide polymorphisms rs7412 and rs429358. Statistics Crude mortality rates were calculated using total months of follow-up time for the whole sample and split by dementia status. The Kaplan-Meier method was used to examine survival time for those with and without a dementia diagnosis. To explore factors predicting mortality, Cox proportional hazard models were computed separately for those with and without dementia, using age at exit (or death) as the time variable. Each predictor variable was entered into an independent predictor model in the first instance. Variables significantly associated with mortality were then combined in a final model, using the enter method. To explore factors associated with diagnosis of dementia, Cox regression models were computed using the same predictor variables but using age at diagnosis/exit from study as the timing variable.

rst instance. Variables significantly associated with mortality were then combined in a final model, using the enter method. To explore factors associated with diagnosis of dementia, Cox regression models were computed using the same predictor variables but using age at diagnosis/exit from study as the timing variable. Variables Time-to-event analyses were computed for death and dementia diagnosis. Dementia status was obtained through carer report, based on independent clinical diagnosis by participants’ regular clinicians after comprehensive clinical assessment. In the United Kingdom, individuals with DS are typically diagnosed as having dementia after specialist assessment in ID services; these expert clinical diagnoses have been shown to be reliable and valid.17 To confirm dementia status, 2 ID psychiatrists independently reviewed dementia symptoms for a sample of individuals blind to original clinical diagnoses using items mapping to International Statistical Classification of Diseases and Related Health Problems, Tenth Revision and Diagnostic and Statistical Manual of Mental Disorders (Fourth Edition) dementia criteria from the structured interview of the Cambridge Examination of Mental Disorders of Older People With Down Syndrome and Others With Intellectual Disabilities.18 The presence of significant cognitive decline owing to dementia was confirmed in 86% of individuals and the remaining 14% showing some degree of cognitive decline (possible dementia) (total n = 36). Because dementia is a progressive disease with a prodromal period spanning many years, individuals diagnosed during follow-up were included in the dementia group. Time variables included length in months from study entry to exit for CMR calculations and age in years at event (death or dementia diagnosis) or exit for hazard ratio calculations. The latest data collection point was used as the exit date for censored cases. Cases were censored on December 13, 2017.

e dementia group. Time variables included length in months from study entry to exit for CMR calculations and age in years at event (death or dementia diagnosis) or exit for hazard ratio calculations. The latest data collection point was used as the exit date for censored cases. Cases were censored on December 13, 2017. Predementia level of ID was obtained via carer report of the participants’ peak level of functioning, based on International Statistical Classification of Diseases and Related Health Problems, Tenth Revision characteristics of mild, moderate, severe, or profound ID. Categories were collapsed to produce a binary variable (mild/moderate vs severe/profound). For the APOE variable, participants were grouped for analysis such that those with 2 ε3 alleles formed the reference group, those with 1 or 2 ε2 alleles formed a second group, and those with 1 or 2 ε4 alleles formed the third group. Participants with ε2:ε4 genotype (n = 5) were excluded from APOE analyses owing to the opposing effects of alleles ε2 and ε4.

re grouped for analysis such that those with 2 ε3 alleles formed the reference group, those with 1 or 2 ε2 alleles formed a second group, and those with 1 or 2 ε4 alleles formed the third group. Participants with ε2:ε4 genotype (n = 5) were excluded from APOE analyses owing to the opposing effects of alleles ε2 and ε4. Living situation split those living with family from those in other living situations including supported accommodation, care homes, and residential homes. Additional factors of interest included sex, presence of early-onset (before age 20 years) and late-onset (older than 36 years) epilepsy, hypothyroidism, congenital heart defects, cataracts, dementia medication, antipsychotic medication, obesity (defined as having a body mass index greater than 30 [calculated as weight in kilograms divided by height in meters squared]), and a binary multimorbidity score (0 = none or 1 comorbid condition; 1 = 2 or more conditions). Health conditions included in this score and the list of drugs counted in the medication variables are listed in the eMethods of the Supplement. Results Sample Two hundred eleven people (96 women) were included in the final sample, giving 503.92 person-years of follow-up: 344.50 person-years from those without dementia and 159.42 person-years from those who received a clinical dementia diagnosis (n = 66). The mean (SD) age of dementia diagnosis overall was 51.98 (7.09) years (n = 65, data missing from 1 participant); 50.83 (5.72) years for women; and 53.41 (8.38) years for men. Table 1 displays participant characteristics by dementia status.

42 person-years from those who received a clinical dementia diagnosis (n = 66). The mean (SD) age of dementia diagnosis overall was 51.98 (7.09) years (n = 65, data missing from 1 participant); 50.83 (5.72) years for women; and 53.41 (8.38) years for men. Table 1 displays participant characteristics by dementia status. Table 1. Participant Demographics by Dementia Status Demographic No. (%) No Dementia Dementia Total No. (%) 145 (68.72) 66 (31.28) Female 60 (41.4) 36 (54.5) Level of ID Mild/moderate 119 (82.1) 50 (75.8) Severe/profound 25 (17.2) 10 (15.2) Missing data 1 (0.7) 6 (9.1) Living situation Home with family/partner 31 (21.4) 46 (69.7) Supported living/care home 113 (77.9) 19 (28.8) Missing 1 (0.7) 1 (1.5) Age at entry, mean (SD) [range], y 47.84 (7.29) [36-72] 53.62 (6.94) [38-67] Age at exit, mean (SD) [range], y 50.23 (7.30) [38-74] 56.05 (7.00) [40-70] Length of follow-up, mean (SD) [range] mo 28.51 (10.65) [3.0-65.0] 28.98 (12.65) [1.0-55.0] BMI, mean (SD) [range] 30.16 (6.99) [17.78-56.80 ] 30.79 (7.01) [20.40-54.00] Obesity (BMI >30) 56 (47.06) 25 (53.19) Missing data, No. 26 19 Late-onset epilepsy 7 (4.8) 19 (28.8) Receiving antiepilepsy medication 5 (71.42) 14 (73.68) Missing data, No. 1 2 Early-onset epilepsy 4 (2.8) 4 (6.1) Receiving antiepilepsy medication 4 (100) 4 (100) Receiving antipsychotics (all atypical) 15 (10.34) 10 (15.15) Hypothyroidism 63 (43.4) 26 (39.4) Cataracts 32 (22.1) 27 (40.9) Congenital heart condition 30 (20.7) 7 (10.6) APOE genotype ε2:ε2 or ε2:ε3 21 (14.5) 8 (12.1) ε3:ε3 89 (61.4) 31 (47.0) ε3:ε4 or ε4:ε4 25 (17.3) 20 (30.3) ε2:ε4 3 (2.1) 2 (3.0) Missing data 7 (4.8) 5 (7.6) ≥2 comorbid health conditions 76 (52.4) 34 (51.5) Receiving antidementia medication NA 33 (50.0) Abbreviations: APOE, apolipoprotein E; BMI, body mass index (calculated as weight in kilograms divided by height in meters squared); NA, not applicable.

25 (17.3) 20 (30.3) ε2:ε4 3 (2.1) 2 (3.0) Missing data 7 (4.8) 5 (7.6) ≥2 comorbid health conditions 76 (52.4) 34 (51.5) Receiving antidementia medication NA 33 (50.0) Abbreviations: APOE, apolipoprotein E; BMI, body mass index (calculated as weight in kilograms divided by height in meters squared); NA, not applicable. Crude Mortality Rates Figure 1 shows the Kaplan-Meier survival function for those with and without a diagnosis of dementia; estimated median survival times were 67 and 72 years, respectively. Twenty-seven participants (11 women) died during the follow-up period. The median age at death was 57 years (57 years for men and 54 years for women). Nineteen participants (70.37%) had a clinical diagnosis of dementia, 9 of whom were women (47.37%). Ten men (62.5%) and 9 women (81.8%) who died had a diagnosis of dementia. The median age at death was 57 years for those without dementia and 55 years for those with dementia. During the follow-up period, 28.78% of the dementia group (n = 19) died compared with 5.52% of those without dementia (n = 8). The CMR across the whole sample was 535.80 deaths per 10 000 person-years (95% CI, 529.30-550.30); for those with dementia, the CMR was 1191.85 deaths per 10 000 person-years (95% CI, 1168.49-1215.21); and for those without a dementia diagnosis, the CMR was 232.22 deaths per 10 000 person-years (95% CI, 227.67-236.77). Figure 1. Cumulative Survival by Dementia Status Kaplan-Meier survival curve for individuals with Down syndrome with dementia (n=66) and without dementia (n=145).

Crude Mortality Rates Figure 1 shows the Kaplan-Meier survival function for those with and without a diagnosis of dementia; estimated median survival times were 67 and 72 years, respectively. Twenty-seven participants (11 women) died during the follow-up period. The median age at death was 57 years (57 years for men and 54 years for women). Nineteen participants (70.37%) had a clinical diagnosis of dementia, 9 of whom were women (47.37%). Ten men (62.5%) and 9 women (81.8%) who died had a diagnosis of dementia. The median age at death was 57 years for those without dementia and 55 years for those with dementia. During the follow-up period, 28.78% of the dementia group (n = 19) died compared with 5.52% of those without dementia (n = 8). The CMR across the whole sample was 535.80 deaths per 10 000 person-years (95% CI, 529.30-550.30); for those with dementia, the CMR was 1191.85 deaths per 10 000 person-years (95% CI, 1168.49-1215.21); and for those without a dementia diagnosis, the CMR was 232.22 deaths per 10 000 person-years (95% CI, 227.67-236.77). Figure 1. Cumulative Survival by Dementia Status Kaplan-Meier survival curve for individuals with Down syndrome with dementia (n=66) and without dementia (n=145). Of the 8 participants who died without a clinical diagnosis of dementia, 2 had late-onset epilepsy, and 1 was reported to be showing early signs of cognitive decline. One died at aged 50 years of a possible underlying heart condition; 2 died of respiratory diseases with no signs of decline at age 63 years and 73 years, respectively; and for 2 participants the cause of death was unknown.

ia, 2 had late-onset epilepsy, and 1 was reported to be showing early signs of cognitive decline. One died at aged 50 years of a possible underlying heart condition; 2 died of respiratory diseases with no signs of decline at age 63 years and 73 years, respectively; and for 2 participants the cause of death was unknown. Factors Associated With Mortality by Dementia Status Table 2 shows the Cox regression results for each independent factor split by dementia status and the final combined model for those with dementia. For individuals without a dementia diagnosis, individual Cox regressions revealed that late-onset epilepsy was the only variable associated with mortality, with a near 10-fold increase in risk. For those with a clinical diagnosis of dementia, APOE genotype, multimorbidity, early-onset epilepsy, and dementia medication status were all significantly independently associated with mortality, such that presence of 1 or more APOE ε4 alleles, 2 or more health conditions, or early-onset epilepsy each were associated with increased mortality risk, and taking antidementia medication was associated with decreased risk. When entered into a combined model, including all significant factors, APOE genotype was the only factor to maintain an association at the P < .05 level. The presence of at least 1 APOE ε4 allele was associated with increased mortality risk nearly 7-fold compared with those with 2 APOE ε3 alleles. In our sample, sex was not statistically significantly associated with mortality for those with or without dementia.

ly factor to maintain an association at the P < .05 level. The presence of at least 1 APOE ε4 allele was associated with increased mortality risk nearly 7-fold compared with those with 2 APOE ε3 alleles. In our sample, sex was not statistically significantly associated with mortality for those with or without dementia. Table 2. Model Coefficients of Factors Associated With Mortality in Adults With DS Variable β Coefficient (SE) df P Value Hazard Ratio (95% CI) Adults with DS without dementia Independent factors Sex 0.8 (0.820) 1 .33 2.23 (0.45-1.11) Level of ID −3.28 (5.245) 1 .53 0.04 (0-1102.73) Multimorbidity status 1.12 (0.767) 1 .14 3.06 (0.68-13.77) APOE genotype 2 .69 APOE group 2 vs group 3 −0.94 (1.097) 1 .39 0.39 (0.045-3.35) APOE group 4 vs group 3 −12.49 (862.96) 1 .99 0 Early-onset epilepsy −3.02 (44.05) 1 .95 0.049 (0-1.54 × 1036) Late-onset epilepsy 2.27 (0.920) 1 .01 9.66 (1.59-58.56) Congenital heart defects 0.29 (0.856) 1 .73 1.34 (0.25-7.17) Antipsychotic medication −0.07 (1.085) 1 .95 0.93 (0.11-7.82) Obesity (BMI >30) 0.031 (0.925) 1 .97 1.03 (0.17-6.32) Hypothyroidism 0.61 (0.721) 1 .40 1.84 (0.45-7.56) Cataracts 0.06 (0.740) 1 .94 1.06 (0.25-4.52) Adults with DS and dementia Independent factors Sex −0.32 (0.495) 1 .52 0.73 (0.28-1.91) Level of ID −0.12 (0.772) 1 .89 0.90 (0.20-4.08) Multimorbidity status 1.24 (0.530) 1 .02 (1.23-9.80) APOE genotype 2 .01 APOE group 2 vs group 3 −0.28 (0.809) 1 .73 0.75 (0.15-3.68)

APOE group 4 vs group 3 −12.49 (862.96) 1 .99 0 Early-onset epilepsy −3.02 (44.05) 1 .95 0.049 (0-1.54 × 1036) Late-onset epilepsy 2.27 (0.920) 1 .01 9.66 (1.59-58.56) Congenital heart defects 0.29 (0.856) 1 .73 1.34 (0.25-7.17) Antipsychotic medication −0.07 (1.085) 1 .95 0.93 (0.11-7.82) Obesity (BMI >30) 0.031 (0.925) 1 .97 1.03 (0.17-6.32) Hypothyroidism 0.61 (0.721) 1 .40 1.84 (0.45-7.56) Cataracts 0.06 (0.740) 1 .94 1.06 (0.25-4.52) Adults with DS and dementia Independent factors Sex −0.32 (0.495) 1 .52 0.73 (0.28-1.91) Level of ID −0.12 (0.772) 1 .89 0.90 (0.20-4.08) Multimorbidity status 1.24 (0.530) 1 .02 (1.23-9.80) APOE genotype 2 .01 APOE group 2 vs group 3 −0.28 (0.809) 1 .73 0.75 (0.15-3.68) APOE group 4 vs group 3 1.75 (0.637) 1 .006 5.74 (1.65-19.99) Early-onset epilepsy 1.87 (0.805) 1 .02 6.50 (1.34-31.47) Late-onset epilepsy 0.59 (0.490) 1 .23 1.80 (0.69-4.69) Congenital heart defects −0.45 (1.036) 1 .67 0.64 (0.08-4.87) Antipsychotic medication 0.532 (0.586) 1 .36 1.70 (0.54-5.37) Obesity (BMI >30) −1.078 (0.692) 1 .12 0.34 (0.09-1.32) Hypothyroidism −0.06 (0.506) 1 .90 0.94 (0.35-2.53) Cataracts 0.44 (0.484) 1 .36 1.55 (0.60-4.01) Dementia medication status −1.47 (0.560) 1 .009 0.23 (0.08-0.69) Final model Multimorbidity status 1.27 (0.704) 1 .07 3.57 (0.90-14.20) APOE genotype 2 .009 APOE group 2 vs 3 −0.73 (0.857) 1 .39 0.48 (0.09-2.58)

APOE group 4 vs group 3 1.75 (0.637) 1 .006 5.74 (1.65-19.99) Early-onset epilepsy 1.87 (0.805) 1 .02 6.50 (1.34-31.47) Late-onset epilepsy 0.59 (0.490) 1 .23 1.80 (0.69-4.69) Congenital heart defects −0.45 (1.036) 1 .67 0.64 (0.08-4.87) Antipsychotic medication 0.532 (0.586) 1 .36 1.70 (0.54-5.37) Obesity (BMI >30) −1.078 (0.692) 1 .12 0.34 (0.09-1.32) Hypothyroidism −0.06 (0.506) 1 .90 0.94 (0.35-2.53) Cataracts 0.44 (0.484) 1 .36 1.55 (0.60-4.01) Dementia medication status −1.47 (0.560) 1 .009 0.23 (0.08-0.69) Final model Multimorbidity status 1.27 (0.704) 1 .07 3.57 (0.90-14.20) APOE genotype 2 .009 APOE group 2 vs 3 −0.73 (0.857) 1 .39 0.48 (0.09-2.58) APOE group 4 vs 3 1.93 (0.699) 1 .006 6.91 (1.76-27.20) Early-onset epilepsy 1.57 (0.896) 1 .08 4.79(0.83-27.69) Dementia medication status −0.97 (0.689 1 .16 0.38 (0.10-1.46) Abbreviations: APOE, apolipoprotein E; BMI, body mass index (calculated as weight in kilograms divided by height in meters squared); ID, intellectual disability.

1.93 (0.699) 1 .006 6.91 (1.76-27.20) Early-onset epilepsy 1.57 (0.896) 1 .08 4.79(0.83-27.69) Dementia medication status −0.97 (0.689 1 .16 0.38 (0.10-1.46) Abbreviations: APOE, apolipoprotein E; BMI, body mass index (calculated as weight in kilograms divided by height in meters squared); ID, intellectual disability. Factors Associated With Age at Dementia Diagnosis Sex (women diagnosed earlier), APOE genotype, multimorbidity, early-onset epilepsy, and living situation were found to be independently associated with age at dementia diagnosis. All but sex remained significantly associated in the combined model (Table 3, with hazard functions in Figure 2). Increased risk for developing dementia was seen for those carrying at least 1 APOE ε4 allele (5-fold increase compared with 2 APOE ε3 alleles), having 2 or more comorbid health conditions (2-fold increase), and having early-onset epilepsy (near 4-fold increase). Those living with family were diagnosed at an earlier age. In this sample, carrying an APOE ε2 allele was not found to be protective compared with those with 2 APOE ε3 alleles (P = .36). Table 3. Model Coefficients for Factors Associated With Dementia Variable β Coefficient (SE) df P Value Hazard Ratio (95% CI) Independent factors associated with dementia Sex (women vs men) −0.581 (0.253) 1 .02 0.56 (0.34-0.92) Level of ID 0.265 (0.352) 1 .45 1.30 (0.65-2.60) Early-onset epilepsy 1.716 (0.532) 1 .001 5.56 (1.96-15.79) Multimorbidity status 0.531 (0.254) 1 .04 1.70 (1.03-2.80) Congenital heart defect −0.294 (0.402 1 .46 0.75 (0.34-1.64) APOE genotype 2 <.001

ith dementia Sex (women vs men) −0.581 (0.253) 1 .02 0.56 (0.34-0.92) Level of ID 0.265 (0.352) 1 .45 1.30 (0.65-2.60) Early-onset epilepsy 1.716 (0.532) 1 .001 5.56 (1.96-15.79) Multimorbidity status 0.531 (0.254) 1 .04 1.70 (1.03-2.80) Congenital heart defect −0.294 (0.402 1 .46 0.75 (0.34-1.64) APOE genotype 2 <.001 APOE group 2 vs group 3 −0.245 (0.4) 1 .54 0.78 (0.36-1.71) APOE group 4 vs group 3 1.56 (0.324) 1 <.001 4.76 (2.52-8.97) Cataracts 0.339 (0.253) 1 .18 1.40 (0.85-2.30) Living situation (family vs other) 1.053 (0.29) 1 <.001 2.87 (1.62-5.06) Antipsychotic medication 0.061 (0.346) 1 .86 1.06 (0.54-2.10) Obesity (BMI >30) 0.308 (0.298) 1 .30 1.36 (0.76-2.44) Hypothyroidism 0.133 (0.255) 1 .60 1.14 (0.69-1.88) Final model Sex (women vs men) 0.391 (0.283) 1 .17 1.48 (0.85-2.58) Living situation (family vs other) 0.758 (0.346) 1 .03 2.14 (1.08-4.20) Early-onset epilepsy 1.284 (0.596) 1 .03 3.61 (1.12-11.60) Multimorbidity status 0.671 (0.3) 1 .03 1.96 (1.09-3.52) APOE genotype 2 <.001 APOE group 2 vs group 3 −0.543 (0.433) 1 .21 0.58 (0.25-1.36) APOE group 4 vs group 3 1.593 (0.339) 1 <.001 4.92 (2.53-9.56) Abbreviations: APOE, apolipoprotein E; BMI, body mass index (calculated as weight in kilograms divided by height in meters squared); ID, intellectual disability. Figure 2. Proportional Hazards of Dementia Predictors Hazard functions for variables associated with dementia diagnosis in Down syndrome.

APOE group 4 vs group 3 1.593 (0.339) 1 <.001 4.92 (2.53-9.56) Abbreviations: APOE, apolipoprotein E; BMI, body mass index (calculated as weight in kilograms divided by height in meters squared); ID, intellectual disability. Figure 2. Proportional Hazards of Dementia Predictors Hazard functions for variables associated with dementia diagnosis in Down syndrome. Discussion This study examined the effect of dementia diagnosis on mortality in a representative cohort of adults with DS in England. Dementia was the proximate cause of death in 70% of our sample overall: 10 men (62.5%) and 9 women (81.8%) had dementia when they died. At least 3 of 8 participants who died without a dementia diagnosis showed signs of cognitive decline and/or seizures; thus, this proportion may be even higher. These results compare strikingly with mortality statistics for England and Wales: dementia of any subtype is mentioned in 17.5% of death certificates for those 65 years and older; and older than 80 years, dementia is the leading cause of death for 14% and 22% of male and female deaths respectively.19,20 In our sample, crude mortality was increased 5-fold for those with dementia (CMR with dementia, 1191.85 deaths per 10 000 person-years vs CMR without dementia, 232.22 deaths per 10 000 person-years), giving a similar mortality rate for those with dementia to that reported for AD dementia in the non-DS population (1070 deaths per 10 000 person-years).21

increased 5-fold for those with dementia (CMR with dementia, 1191.85 deaths per 10 000 person-years vs CMR without dementia, 232.22 deaths per 10 000 person-years), giving a similar mortality rate for those with dementia to that reported for AD dementia in the non-DS population (1070 deaths per 10 000 person-years).21 In our sample, we found no clear differences in mortality between men and women, matching previous work showing similar ages at death in men and women with DS.2 Women were diagnosed as having dementia up to 3 years earlier than men. This pattern has been previously reported in adults with DS in the United Kingdom10; however, our findings only held when sex was considered as an independent factor. In the final model, including APOE genotype, multimorbidity, living situation and early-onset epilepsy, the influence of sex on dementia diagnosis was lost.

s pattern has been previously reported in adults with DS in the United Kingdom10; however, our findings only held when sex was considered as an independent factor. In the final model, including APOE genotype, multimorbidity, living situation and early-onset epilepsy, the influence of sex on dementia diagnosis was lost. Seizures are a common feature of AD in those with and without DS, occurring in a quarter of patients without DS with AD22 and 40% of patients with DS and AD.14 However, late-onset epilepsy was also noted in 7 people (4.8%) without a dementia diagnosis in our study, increasing mortality risk 10-fold. For those without dementia, late-onset epilepsy was the only factor associated with mortality. This raises the question of whether seizures can begin in the absence of other features of dementia in individuals with DS or whether these 7 individuals had significant AD pathology and neurological symptoms but had yet to receive a formal dementia diagnosis. While preexisting ID can make it challenging for decline to be identified in this population, detailed clinical assessments have been found to be robust and valid for those with DS,17 and a range of sensitive cognitive batteries have been developed within the past 10 years.16,23,24,25 Baseline assessments completed in early adulthood can help to serve as each individual’s reference point, allowing decline to be identified on an individual level in this highly variable population.26

those with DS,17 and a range of sensitive cognitive batteries have been developed within the past 10 years.16,23,24,25 Baseline assessments completed in early adulthood can help to serve as each individual’s reference point, allowing decline to be identified on an individual level in this highly variable population.26 For people with DS and dementia, carrying at least 1 APOE ε4 allele was associated with increased mortality risk 7-fold. These results suggest people with DS may be particularly vulnerable to the effects of APOE ε4 because APOE ε4 carriers with AD in the non-DS population show little difference to noncarriers with AD in disease progression or mortality.27,28 Our data confirm some of the other associations with mortality previously observed, including the deleterious effect of epilepsy and the potentially beneficial effect of currently available medication such as acetyl-choline esterase inhibitors.29 However, other associations were not observed: antipsychotics have been found to double mortality risk in people with dementia in the non-DS population,30 yet in our sample we did not find a statistically significant association between antipsychotic use and death in those with or without dementia. Similarly, obesity had no discernible association with death or dementia onset in this study. Our data are from a comparatively small group, with a maximum follow-up time of 65 months. Further research using larger samples over longer periods of time would be valuable to clarify whether these reflect genuine differences in risk in the DS population or simply reflect a lack of power for identifying multiple risk factors in this sample.

aratively small group, with a maximum follow-up time of 65 months. Further research using larger samples over longer periods of time would be valuable to clarify whether these reflect genuine differences in risk in the DS population or simply reflect a lack of power for identifying multiple risk factors in this sample. APOE ε4 genotype, early-onset epilepsy, multimorbidity, and living with family were all associated with earlier dementia diagnoses. Previous studies have shown that APOE genotype influences dementia risk in DS in much the same way as in the non-DS population. Similarly, epilepsy has been found to increase risk of AD in the general population, with adults with epilepsy younger than 65 years nearly 40 times more likely and those older than 65 years nearly 7 times more likely to be diagnosed as having AD.31 Aside from known vascular risk factors, combined health comorbidities may also increase dementia risk in those without DS, suggesting a role for poor general health in dementia risk.32 A further explanation could be that increased interaction with health care services for those with multiple health conditions and increased awareness of change for those living with family may make these groups more likely to receive a dementia assessment and subsequent diagnosis rather than increasing risk of dementia per se.

r explanation could be that increased interaction with health care services for those with multiple health conditions and increased awareness of change for those living with family may make these groups more likely to receive a dementia assessment and subsequent diagnosis rather than increasing risk of dementia per se. Because multimorbidity was associated with increased dementia risk and mortality in those who received dementia diagnoses, our results also highlight the need for effective recognition and treatment of common health comorbidities in DS. Individuals with ID experience significant health inequalities,33 and evidence suggests that incentivizing general practitioners to offer comprehensive ID health checks increases the number of specific health assessments completed and may thus reduce said health inequalities.34 Given that several of the comorbidities we included are treatable, such health checks could have longer-term positive effects than have previously been assessed. Limitations Our data were collected as part of a prospective, longitudinal study of adults with DS, providing extensive health information and cognitive assessments for the individuals. While our sample is large for a study of such detail, we acknowledge that the numbers included are relatively small for an epidemiologic study. Health data were collected via informant report, which may be influenced by reporter bias, their memory, and the relationship between the informer and the individual with Down syndrome.

sample is large for a study of such detail, we acknowledge that the numbers included are relatively small for an epidemiologic study. Health data were collected via informant report, which may be influenced by reporter bias, their memory, and the relationship between the informer and the individual with Down syndrome. Conclusions Our study shows that most adults with DS now have dementia when they die and are affected by some of the same factors associated with dementia (such as APOE genotype) as we see in the non-DS population. These findings support the urgent need for clinical trials of treatments to prevent or delay dementia in those with DS. Finally, we hope that our findings can improve clinical care by identifying factors associated with increased risk for dementia and mortality risk in this population, suggesting the potentially beneficial effects of existing medication options and helping clinicians provide prognostic information for their patients with DS. Supplement. eMethods. Click here for additional data file.

Conclusions From a practical perspective, we believe that the most advantageous future use of optimized blood Aβ assays is as a screening tool for identifying subjects at a higher risk of being Aβ positive. They could, for example, be applied as an initial test together with other noninvasive, cost-efficient tools that aid the decision about whom a general practitioner should refer for further investigation at memory clinics where CSF or PET and more extensive clinical assessment could be used to support the AD diagnosis. Another useful setting for the blood biomarkers are clinical AD trials enrolling Aβ-positive participants, where they can be used for prescreening to minimize the number of unnecessary (Aβ-negative) lumbar punctures and Aβ PET scans, as well as lowering the costs for the examinations up to 30% to 50% depending on the cutoff (eFigure 5 in the Supplement).48 Although further validation studies are needed, this illustrates the potential usefulness blood assays might have, especially considering the ongoing great need to recruit large cohorts for AD drug trials in preclinical and prodromal stages. Supplement. eMethods. eResults. eTable 1. Performance Characteristics of the Plasma Aβ42, Aβ40 and Tau Elecsys Assays eTable 2. Performance Characteristics of CSF and Plasma NFH Assays eTable 3. Associations Between Plasma and CSF Biomarkers. eTable 4. Area Under the Curves From Logistic Regression Models for Prediction of Aβ Positivity eTable 5. Plasma NfH as Additional Predictor for Aβ Positivity.

eTable 1. Performance Characteristics of the Plasma Aβ42, Aβ40 and Tau Elecsys Assays eTable 2. Performance Characteristics of CSF and Plasma NFH Assays eTable 3. Associations Between Plasma and CSF Biomarkers. eTable 4. Area Under the Curves From Logistic Regression Models for Prediction of Aβ Positivity eTable 5. Plasma NfH as Additional Predictor for Aβ Positivity. eTable 6. Area Under the Curves From Logistic Regression Models for Prediction of Aβ Positivity in the Younger and Older Half of the BioFINDER Cohort. eTable 7. Demographic and Clinical Data of the German Validation Cohort eFigure 1. Correlations Between Plasma and CSF Biomarkers. eFigure 2. Plasma Biomarkers in Diagnostic Groups. eFigure 3. ROC Analysis of Plasma Biomarkers Using the Ratio of CSF P-tau/Aβ42 as Reference Standard in BioFINDER. eFigure 4. ROC Analysis of Plasma Biomarkers Using the Ratio of CSF P-tau/Aβ42 as Reference Standard in the Independent Validation Cohort. eFigure 5. Implementation of Plasma Aβ42, Aβ40 and APOE Genotype in an AD Trial Screening Scenario. eReferences. Click here for additional data file.

Introduction A key hallmark of Alzheimer disease (AD) is the gradual accumulation of β-amyloid (Aβ) in the brain, which starts decades before the onset of cognitive symptoms. Detection of abnormal Aβ accumulation (Aβ positivity) may support the clinical diagnosis of AD1,2 and is essential for including participants in clinical AD trials targeting Aβ.3 β-Amyloid can be detected in vivo using positron emission tomography (PET) with ligands that bind to Aβ fibrils or by measuring the levels of the peptide Aβ1-42 (Aβ42) in cerebrospinal fluid (CSF).4 Alzheimer disease affects 1 in 10 persons aged 65 years and older and is expected to affect more than 100 million people by 2050.5,6 The costs and limited access to PET or CSF analysis may restrict their use to a minority of cases. There is thus a great need for readily available methods that can detect brain Aβ, and perhaps the most desirable goal has been to establish blood-based biomarkers of Aβ. Many candidate blood biomarkers have failed in replication studies,7,8 but somewhat promising results have been seen for plasma tau, neurofilament light chain (NFL), and combinations of Aβ42 and Aβ40.9,10,11,12,13,14,15,16,17 Although there are diagnostic inconsistencies regarding the plasma Aβ42/Aβ40 ratio in older studies,18 more recent studies have demonstrated that it correlates with brain Aβ and can differentiate patients with AD from healthy control participants.12,13 Most recently, 2 independent groups demonstrated improved accuracy for plasma Aβ42/Aβ40 using immunoprecipitation–mass spectrometry assays.19,20 Although these studies are promising and show the potential of plasma Aβ as a true AD biomarker, they are costly and need extensive development before they can be implemented in primary care or in large screenings where cost-effective, fully automated, high-throughput, and highly reliable analysis methods are needed.

0 Although these studies are promising and show the potential of plasma Aβ as a true AD biomarker, they are costly and need extensive development before they can be implemented in primary care or in large screenings where cost-effective, fully automated, high-throughput, and highly reliable analysis methods are needed. Measuring plasma Aβ presents the same challenges as measuring CSF Aβ in that several analysis methods exist and unified cutoffs have been difficult to establish, even using the same assay, owing to high variability between laboratories and assay batches.21,22 Recently, fully automated immunoassays have been developed by several different vendors with improved reliability and precision for CSF Aβ and tau species.23,24,25 For example, for the Elecsys immunoassays (Roche Diagnostics), it has been shown that CSF cutoffs established in one European cohort could be applied to another independent cohort in the United States to determine amyloid PET status with high accuracy.24 Using these newly developed Elecsys assays for detection of Aβ42, Aβ40, and tau, our aims were to examine the accuracy of plasma Aβ42, Aβ40, and tau to estimate Aβ positivity, whether the accuracy could be improved by adding plasma neurofilament (light and heavy chain) and APOE genotype to the models, and how the Elecsys assays perform in an independent validation cohort.

on of Aβ42, Aβ40, and tau, our aims were to examine the accuracy of plasma Aβ42, Aβ40, and tau to estimate Aβ positivity, whether the accuracy could be improved by adding plasma neurofilament (light and heavy chain) and APOE genotype to the models, and how the Elecsys assays perform in an independent validation cohort. Methods Participants The study population was included from the prospective Swedish BioFINDER Study, which enrolled participants between July 6, 2009, to February 11, 2015, from the southern part of Sweden. Of all 892 participants in BioFINDER’s control, mild cognitive symptoms, and AD cohorts, plasma samples were hemolyzed or not available in sufficient amount for 50 individuals. Thus, 842 participants could be included in the present study. They were classified as cognitively unimpaired26 (CU; 513 participants, of whom 195 had subjective cognitive decline)27; mild cognitive impairment28 (MCI; 265 participants); or AD dementia2 (64 participants). In subsample analyses, we grouped the population into CU and cognitively impaired (MCI + AD), because all participants with AD were Aβ positive and therefore could not be examined separately using Aβ status as outcome. Study design and specific inclusion and exclusion criteria are described elsewhere29 (eMethods in the Supplement). The study was approved by the Regional Ethics Committee in Lund, Sweden, and all participants gave their written informed consent to participate in the study. For the independent validation cohort, the study was approved by the ethical committee of the Medizinische Hochschule in Hannover and the ethical committee of the University of Ulm, and all participants gave written informed consent.

and all participants gave their written informed consent to participate in the study. For the independent validation cohort, the study was approved by the ethical committee of the Medizinische Hochschule in Hannover and the ethical committee of the University of Ulm, and all participants gave written informed consent. Plasma and CSF Procedures Blood samples were collected at the same time as CSF samples, and the collection was performed in the morning with participants not fasting. Blood samples were collected and analyzed according to a standardized protocol. For each study participant, blood was collected in 6 EDTA-plasma tubes (Vacutainer K2EDTA tube; BD Diagnostics) and centrifuged (2000 g, 4°C) for 10 minutes. After centrifugation, plasma from all 6 tubes was transferred into one 50-mL tube (62.547.254, Sarstedt), mixed, and 1 mL was aliquoted into polypropylene tubes (72.694.100; Sarstedt) and stored at –80°C within 30 to 60 minutes of collection. All plasma samples went through 1 freeze-thaw cycle before the analysis, when 300 μL was further aliquoted into Lobind tubes (72.704.600; Sarstedt). The current standardized protocol is consistent with recent findings that blood must be centrifuged within 1 hour and frozen shortly thereafter; however, up to 3 freeze-thaw cycles and 5 tube transfers do not affect plasma Aβ and tau values.30 Lumbar puncture and CSF handling followed a structured protocol.31 Plasma and CSF Aβ42, Aβ40, total tau (tau), and phosphorylated tau (P-tau; only in CSF) were analyzed using the Elecsys immunoassays on a cobas e 601 analyzer (Roche Diagnostics) at the Clinical Neurochemistry Laboratory, University of Gothenburg, Sweden. Additional assay data (also including NFL and neurofilament heavy chain [NFH] analyses) can be found in the eMethods and eTables 1 and 2 in the Supplement.

e analyzed using the Elecsys immunoassays on a cobas e 601 analyzer (Roche Diagnostics) at the Clinical Neurochemistry Laboratory, University of Gothenburg, Sweden. Additional assay data (also including NFL and neurofilament heavy chain [NFH] analyses) can be found in the eMethods and eTables 1 and 2 in the Supplement. Reference Standard for Aβ Status β-Amyloid status was determined using the Elecsys CSF Aβ42/Aβ40 ratio, which is a ratio that has been validated against amyloid PET status with more than 90% agreement.32,33,34 An unbiased cutoff of less than 0.059 was used to define Aβ positivity based on mixture modeling statistics, which previously has proved to provide robust and accurate thresholds.35,36 In a secondary analysis (eFigure 3 and eFigure 4 in the Supplement), we used the Elecsys CSF P-tau/Aβ42 ratio to define Aβ positivity, using the predefined cutoff of 0.022 or greater.24

define Aβ positivity based on mixture modeling statistics, which previously has proved to provide robust and accurate thresholds.35,36 In a secondary analysis (eFigure 3 and eFigure 4 in the Supplement), we used the Elecsys CSF P-tau/Aβ42 ratio to define Aβ positivity, using the predefined cutoff of 0.022 or greater.24 Independent Validation Cohort All 237 participants of this study were enrolled between January 29, 2000, and October 11, 2006, at 2 clinical sites in Germany, Ulm and Hannover, as part of a prospective validation study of new biomarkers for the early diagnosis of AD. The participants were classified as having CU (n = 34), MCI (n = 109),37 or AD mild dementia38 (Mini-Mental State Examination score >22; n = 94). Specific inclusion/exclusion criteria and CSF and blood collection procedures30 are described in the eMethods in the Supplement. The cutoff of CSF Aβ42/Aβ40 of less than 0.059 established in BioFINDER to define Aβ positivity was also used in the validation cohort after a thorough assessment of the CSF Aβ42/Aβ40 distribution. As in BioFINDER, the previously published cutoff of CSF P-tau/Aβ42 0.022 or greater24 was used as a secondary reference standard for Aβ status also in the validation cohort.

blished in BioFINDER to define Aβ positivity was also used in the validation cohort after a thorough assessment of the CSF Aβ42/Aβ40 distribution. As in BioFINDER, the previously published cutoff of CSF P-tau/Aβ42 0.022 or greater24 was used as a secondary reference standard for Aβ status also in the validation cohort. Statistical Analysis According to previous publications39,40 and present analyses (eResults in the Supplement), APOE (OMIM:107741) genotype analyzed from blood was grouped into (A) ε2/ε2 or ε2/ε3; (B) ε3/ε3; (C) ε2/ε4 or ε3/ε4; and (D) ε4/ε4. APOE ε3/ε3 was the reference category in the statistical models. β-Amyloid status was predicted in logistic regression models to produce estimates of the predictors, probabilities for Aβ positivity, and resulting area under the receiver operating characteristic curve (AUC). The examined predictors in BioFINDER were the plasma biomarkers Aβ42, Aβ40, tau, NFH, and NFL and APOE genotype. The models were built using the Akaike information criterion (AIC) to evaluate the model fit. A predictor was kept in the model if AIC improved significantly (a decrease in AIC of at least 2, noted as “ΔAIC -2”).41 Differences in AUCs were compared using DeLong statistics.42 In the replication analysis, the models (intercepts and estimates) established in BioFINDER were applied to the validation cohort. The resulting probabilities from the validation cohort were used to calculate the AUCs (only plasma Aβ42, Aβ40, and tau were available in this cohort). Additional statistical methods are described in the eMethods in the Supplement. SPSS version 24 (IBM) and R version 3.4 (R Foundation for Statistical Computing) were used for the statistical analyses. Two-sided P < .05 indicated statistical significance.

(only plasma Aβ42, Aβ40, and tau were available in this cohort). Additional statistical methods are described in the eMethods in the Supplement. SPSS version 24 (IBM) and R version 3.4 (R Foundation for Statistical Computing) were used for the statistical analyses. Two-sided P < .05 indicated statistical significance. Results Among the 842 study participants in BioFINDER, mean (SD) age was 72.0 (5.6) years, and 446 (52.5%) were female. Demographic and clinical data for the study participants in BioFINDER are shown in Table 1. In the total BioFINDER population of 842, 368 were positive for Aβ (prevalence, 44%); 147 of 513 with CU (29%) were positive; 157 of 265 (60%) with MCI; and, by definition, all 64 (100%) with AD dementia.

46 (52.5%) were female. Demographic and clinical data for the study participants in BioFINDER are shown in Table 1. In the total BioFINDER population of 842, 368 were positive for Aβ (prevalence, 44%); 147 of 513 with CU (29%) were positive; 157 of 265 (60%) with MCI; and, by definition, all 64 (100%) with AD dementia. Table 1. Demographic and Clinical Dataa Characteristic CU Aβ− (n = 366) CU Aβ+ (n = 147) MCI Aβ− (n = 108) MCI Aβ+ (n = 157) AD Aβ+ (n = 64) Sex Male 152 69 76b 79 25 Female 214 78 32 78 39 Age, y 72 (5) 73 (5) 69 (6)b 72 (5) 76 (5)b APOE genotype, % 1 or 2 ε4 alleles, 19 63b 24 70b 69b MMSE 28.9 (1.1) 28.6 (1.3) 27.5 (1.8)b 26.7 (1.8)b 21.8 (3.7)b Delayed recall (ADAS-cog; errors)c 2.2 (1.9) 3.2 (2.3)b 5.7 (2.4)b 7.0 (2.1)b 8.6 (1.6)b CSF Aβ42, pg/mL 1665 (596) 819 (303)b 1572 (605) 706 (256)b 671 (315)b Aβ40, ng/mL 18.2 (5.2) 19.5 (5.9)d 17.3 (5.7) 17.8 (5.0) 17.9 (6.2) Aβ42/Aβ40 0.091 (0.016) 0.042 (0.009)b 0.090 (0.014) 0.040 (0.098)b 0.037 (0.009)b T-tau, pg/mL 209 (62) 309 (112)b 209 (76) 341 (136)b 384 (143)b P-tau, pg/mL 17.5 (5.3) 28.5 (12.0)b 16.9 (6.4) 32.2 (14.5) b 36.3 (16.3)e NFL, pg/mL 918 (490) 1216 (842)b 1648 (1517)b 1531 (1195)b 2002 (1835)b NFH, pg/mL 504 (190) 584 (241)b 641 (463)b 637 (303) b 821 (687)b Plasma Aβ42, pg/mL 32.8 (4.9) 29.6 (4.3)b 33.1 (5.2) 30.3 (4.5)b 23.3 (8.2)b Aβ40, pg/mL 482 (63.3) 479 (67.5) 495 (83.2) 492 (75.4) 380 (131.7)e T-tau, pg/mL 16.6 (4.7) 17.9 (5.4)e 18.7 (6.1)b 19.1 (5.2)b 16.7 (6.0) Aβ42/Aβ40 0.068 (0.007) 0.062 (0.007)b 0.067 (0.007) 0.062 (0.006)b 0.062 (0.010)b NFL, pg/mL 21.0 (11.8) 29.1 (59.6)e 28.3 (28.4)b 29.0 (17.9)b 43.8 (28.7)b NFH, pg/mLf 51.4 (68.2) 53.7 (48.7) 59.7 (55.1) 65.9 (56.6)b 79.8.4 (77.0)b Abbreviations: Aβ, β-amyloid; Aβ+, Aβ positive; Aβ−, Aβ negative; CSF, cerebrospinal fluid; MMSE, Mini-Mental State Examination; NHL, neurofilament heavy chain; NFL, neurofilament light chain.

29.1 (59.6)e 28.3 (28.4)b 29.0 (17.9)b 43.8 (28.7)b NFH, pg/mLf 51.4 (68.2) 53.7 (48.7) 59.7 (55.1) 65.9 (56.6)b 79.8.4 (77.0)b Abbreviations: Aβ, β-amyloid; Aβ+, Aβ positive; Aβ−, Aβ negative; CSF, cerebrospinal fluid; MMSE, Mini-Mental State Examination; NHL, neurofilament heavy chain; NFL, neurofilament light chain. a β-Amyloid status was defined based on a CSF Aβ42/Aβ40 cutoff of ≤0.059. Data are shown as mean (SD) unless otherwise specified. Demographic factors, clinical characteristics, and biomarkers levels were compared using χ2 test and 1-way analysis of variance (not adjusted for multiple comparisons). Neurofilament light chain and NFH values were ln-transformed before the analysis. In the receiver operating characteristic subanalyses, the mild cognitive impairment and Alzheimer disease cohorts are combined as cognitively impaired (Figure 3A and B; eFigure 3C and D; eTable 4 in the Supplement). When calculating the Aβ42/Aβ40 ratio, picomolar per milliliter was used for both peptides. b P < .001 compared with CU Aβ−. c Data were missing for 1 CU Aβ−, 1 MCI Aβ−, 8 MCI Aβ+ and 5 AD Aβ+ individuals. d P < .05. e P < .01. f Data were missing for 6 CU Aβ−, 3 CU Aβ+, 2 MCI Aβ−, 5 MCI Aβ+ and 5 AD Aβ+ individuals. Correlations Between Plasma and CSF Biomarkers In the whole BioFINDER population, there were statistically significant positive correlations between all plasma and corresponding CSF biomarkers (eTable 3 in the Supplement). The correlations were similar within diagnostic subgroups (eFigure 1 and eTable 3 in the Supplement).

elations Between Plasma and CSF Biomarkers In the whole BioFINDER population, there were statistically significant positive correlations between all plasma and corresponding CSF biomarkers (eTable 3 in the Supplement). The correlations were similar within diagnostic subgroups (eFigure 1 and eTable 3 in the Supplement). Plasma Aβ and tau Levels in Diagnostic Groups In BioFINDER, plasma levels of Aβ42, Aβ40, and Aβ42/Aβ40 were decreased in Aβ-positive (CSF Aβ42/Aβ40 ≤ 0.059) compared with Aβ-negative (CSF Aβ42/Aβ40 > 0.059) participants (Aβ42, P < .001; Aβ40 P = .003, Aβ42/Aβ42, P < .001; Figure 1A-C). When comparing Aβ groups stratified by diagnostic subgroup, plasma levels of Aβ42 were lower in the CU Aβ-positive, MCI Aβ-positive and AD Aβ-positive dementia groups compared with the CU Aβ-negative and MCI Aβ-negative groups (P < .001 for all; Figure 1D). The decrease in plasma Aβ42 was more pronounced in AD Aβ-positive dementia compared with CU Aβ-positive and MCI Aβ positive groups. Plasma Aβ40 levels were lower in the AD Aβ-positive dementia group compared with all other groups (P < .001 for all), but there were no differences between CU Aβ-negative, CU Aβ-positive, and MCI Aβ-positive participants (Figure 1E). The plasma Aβ42/Aβ40 ratio was lower in the CU Aβ-positive, MCI Aβ-positive, and AD Aβ-positive dementia groups than in the CU Aβ-negative and MCI Aβ-negative groups with no differences across the Aβ-positive groups (Figure 1F). The significant findings were very similar when adjusting for age and sex (data not shown). Comparisons of plasma tau, NFL, and NFH are shown in eFigure 2 in the Supplement.

D Aβ-positive dementia groups than in the CU Aβ-negative and MCI Aβ-negative groups with no differences across the Aβ-positive groups (Figure 1F). The significant findings were very similar when adjusting for age and sex (data not shown). Comparisons of plasma tau, NFL, and NFH are shown in eFigure 2 in the Supplement. Figure 1. Levels of Plasma β-Amyloid (Aβ) Biomarkers Plasma Aβ42 (A), Aβ40 (C), and the plasma Aβ42/Aβ40 ratio (E) in the Aβ-positive (Aβ+) (CSF Aβ42/Aβ40 ≤ 0.059) and Aβ-negative (Aβ−) (CSF Aβ42/Aβ40 > 0.059) groups. Plasma Aβ42 (B), Aβ40 (D), and the plasma Aβ42/Aβ40 ratio (F) in the CU, MCI, and AD participant groups stratified by Aβ status. The dotted lines indicate median levels in the CU Aβ-negative group. P values are calculated from t test (A, C, E) or 1-way analysis of variance and post hoc tests with the statistical significance set to P < .005 (.05/10.00) to account for the Bonferroni correction (B, D, F). The significant findings were similar when adjusting for age and sex (data not shown). Group comparisons of plasma tau, NFH, and NFL are shown in eFigure 2 in the Supplement. AD, Alzheimer disease; CSF, cerebrospinal fluid; CU, cognitively unimpaired; MCI, mild cognitive impairment; NFH, neurofilament heavy chain; and NFL, neurofilament light chain.

imilar when adjusting for age and sex (data not shown). Group comparisons of plasma tau, NFH, and NFL are shown in eFigure 2 in the Supplement. AD, Alzheimer disease; CSF, cerebrospinal fluid; CU, cognitively unimpaired; MCI, mild cognitive impairment; NFH, neurofilament heavy chain; and NFL, neurofilament light chain. Accuracy of Plasma Aβ42 and Aβ40 for Predicting Brain Aβ Positivity The results from the logistic regression models in BioFINDER of all tested single and combined biomarkers are shown in eTables 4, 5, and 6 in the Supplement (including AUC and AIC values). The receiver operating characteristic curves and AUCs of selected biomarkers for predicting Aβ positivity are shown in Figure 2A and B (sensitivity, specificity, and cutoffs are shown in Table 2). The plasma Aβ42/Aβ40 ratio predicted Aβ positivity with an AUC of 0.77 (95% CI, 0.74-0.81) in the whole BioFINDER population. Using plasma Aβ42 and Aβ40 as separate predictors in a logistic regression resulted in a slightly but significantly better AUC (0.80; 95% CI, 0.77-0.83; P = .01) and a better model fit (ΔAIC, –66). We also tested the accuracy of the biomarkers in different age groups and in those with and without cognitive impairment (Figure 3; eTable 4 and eTable 6 in the Supplement), with similar results (AUC ±0.02 compared with the total population).

er AUC (0.80; 95% CI, 0.77-0.83; P = .01) and a better model fit (ΔAIC, –66). We also tested the accuracy of the biomarkers in different age groups and in those with and without cognitive impairment (Figure 3; eTable 4 and eTable 6 in the Supplement), with similar results (AUC ±0.02 compared with the total population). Figure 2. Receiver Operating Characteristic (ROC) Analysis of Plasma Biomarkers in the BioFINDER and Validation Cohorts Optimized ROC curves and corresponding areas under the curve (AUCs) for plasma Aβ together with the additional predictors, APOE, plasma tau, and neurofilament light chain (NFL) to assess accuracy when predicting Aβ positivity (crebrospinal fluid Aβ42/Aβ40 ≤ 0.059) in the BioFINDER (A and B, n = 842); and the replication of these models (C and D, n = 237) in the validation cohort using the estimates and intercepts established in BioFINDER. APOE genotype and NFL were not available in the validation cohort. Error bars indicate 95% CIs. ROC analyses in subpopulations can be found in Figure 3 and eTable 4 and 6 in the Supplement. Sensitivities and specificities are shown in Table 2. ROC analyses using the alternative reference standard for Aβ positivity (CSF P-tau/Aβ42 ≥ 0.022) are shown in eFigures 3 and 4 in the Supplement.

ror bars indicate 95% CIs. ROC analyses in subpopulations can be found in Figure 3 and eTable 4 and 6 in the Supplement. Sensitivities and specificities are shown in Table 2. ROC analyses using the alternative reference standard for Aβ positivity (CSF P-tau/Aβ42 ≥ 0.022) are shown in eFigures 3 and 4 in the Supplement. Table 2. Sensitivity and Specificity for Aβ Status in the BioFINDER Cohort and the Validation Cohort Plasma Biomarkers Cutoffa % (95% CI) Sensitivity Specificity BioFINDER cohort Aβ42/Aβ40 ratio 0.065 75 (68-80) 72 (65-77) Aβ42, Aβ40 0.45 73 (65-78) 76 (68-80) Aβ42, Aβ40, tau 0.36 86 (78-90) 68 (61-72) Aβ42, Aβ40, NFL 0.38 84 (76-88) 70 (62-74) Aβ42, Aβ40, APOE 0.29 88 (82-92) 68 (58-72) Aβ42, Aβ40, tau, NFL, APOE 0.52 73 (64-78) 86 (77-89) Validation cohort Aβ42/Aβ40 ratio 0.065 70 (61-80) 73 (61-81) Aβ42, Aβ40 0.45 89 (80-95) 69 (54-81) Aβ42, Aβ40, tau 0.36 89 (74-94) 64 (49-74) Abbreviations: Aβ, β-amyloid; NFL, neurofilament light chain. a Cutoffs were determined based on the highest Youden index (sensitivity + specificity – 1) for Aβ positivity in the BioFINDER cohort. The cutoffs were then replicated in the validation cohort. Cutoffs are from the probabilities from the corresponding logistic regression models, except for Aβ42/Aβ40 where the actual ratio of the biomarker levels constitute the cutoff. Aβ status (reference standard) was determined using the cerebrospinal fluid Aβ42/40 ratio (<0.059). The 95% CIs were computed using 2000 stratified bootstrap replicates. Neurofilament light chain and APOE genotype were not available in the validation cohort.

e actual ratio of the biomarker levels constitute the cutoff. Aβ status (reference standard) was determined using the cerebrospinal fluid Aβ42/40 ratio (<0.059). The 95% CIs were computed using 2000 stratified bootstrap replicates. Neurofilament light chain and APOE genotype were not available in the validation cohort. Figure 3. Receiver Operating Characteristic (ROC) Analysis of Plasma Biomarkers in Subpopulations in BioFINDER ROC curves and corresponding areas under the curve (AUCs) from logistic regression models for plasma Aβ together with the additional predictors APOE, plasma tau, and neurofilament light chain (NFL), to assess accuracy when detecting Aβ positivity (cerebrospinal fluid Aβ42/Aβ40 ≤ 0.059) in cognitively unimpaired participants (A and B, n = 513), cognitively impaired participants (C and D, n = 329), the younger half of the cohort (E and F, n = 428; 60-72 y), and the older half of the cohort (G and H, n = 414; 73-88 y). Cognitively unimpaired comprised of cognitively healthy controls and participants with subjective cognitive decline. Cognitively impaired comprised of participants with mild cognitive impairment and Alzheimer disease dementia. AUC indicates area under the curve; and NFL, neurofilament light chain.

H, n = 414; 73-88 y). Cognitively unimpaired comprised of cognitively healthy controls and participants with subjective cognitive decline. Cognitively impaired comprised of participants with mild cognitive impairment and Alzheimer disease dementia. AUC indicates area under the curve; and NFL, neurofilament light chain. Aβ Detection With Additional Predictors The accuracy of predicting Aβ status was further examined by adding APOE genotype, and plasma levels of tau, NFL, and NFH to plasma Aβ42 and Aβ40 in logistic regression models (Figure 2A-B). When adding plasma tau, AUC increased nonsignificantly to 0.81 (95% CI, 0.78-0.84) and further improved the model fit (ΔAIC, –27). However, when instead adding APOE genotype to plasma Aβ42 and Aβ40, AUC increased significantly from 0.80 to 0.85 (95% CI, 0.82-0.88; P < .001; Figure 2A-B; eTable 4 in the Supplement). Adding plasma tau to plasma Aβ42, Aβ40, and APOE increased the AUC slightly to 0.86 (95% CI, 0.83-0.88; ΔAIC, –20). A further slight increase was seen when adding plasma NFL to plasma Aβ42, Aβ40, tau, and APOE (AUC, 0.87; 95% CI, 0.84-0.89; ΔAIC –16; Figure 2A-B). The results were similar in CU and cognitively impaired participants, respectively, except that plasma tau and NFL were not a significant predictor in the cognitively impaired group (eTable 4 in the Supplement). The results were also similar when the CSF P-tau/Aβ42 ratio was used to define Aβ positivity (eFigure 3 in the Supplement). Plasma NFH did not contribute to Aβ prediction in addition to plasma Aβ42 and Aβ40 (eTable 5 in the Supplement).

gnificant predictor in the cognitively impaired group (eTable 4 in the Supplement). The results were also similar when the CSF P-tau/Aβ42 ratio was used to define Aβ positivity (eFigure 3 in the Supplement). Plasma NFH did not contribute to Aβ prediction in addition to plasma Aβ42 and Aβ40 (eTable 5 in the Supplement). Independent Validation Cohort Among the 237 study participants in the independent validation cohort, mean (SD) age was 66 (10) years with a range of 23 to 85 years, and 120 (50.6%) were female. The demographic characteristics are shown in eTable 7 in the Supplement and the accuracy of the plasma assays in Figure 2C and D. The AUC for plasma Aβ42 and Aβ40 to predict Aβ positivity was 0.86 (95% CI, 0.81-0.91) when applying the estimates from the model established in BioFINDER (compared with an AUC of 0.80, 95% CI 0.77-0.83 in BioFINDER). When applying the BioFINDER model that included plasma Aβ42, Aβ40, and tau in the validation cohort, the AUC was slightly lower than when using only plasma Aβ42 and Aβ40 (AUC, 0.84; 95% CI, 79-89). With the alternative reference standard for Aβ status (CSF P-tau/Aβ42 ≥ 0.022; eFigure 4 in the Supplement), the accuracy was slightly lower for plasma Aβ42 and Aβ40 (AUC, 0.83; 95% CI, 0.78-0.89) but still better than the corresponding results in the BioFINDER cohort (AUC, 0.79; 95% CI, 0.76-0.82; eFigure 3 in the Supplement). Sensitivities and specificities using the cutoffs established in BioFINDER are shown in Table 2. Plasma NFH, NFL, and APOE genotype were not available in the validation cohort.

0.78-0.89) but still better than the corresponding results in the BioFINDER cohort (AUC, 0.79; 95% CI, 0.76-0.82; eFigure 3 in the Supplement). Sensitivities and specificities using the cutoffs established in BioFINDER are shown in Table 2. Plasma NFH, NFL, and APOE genotype were not available in the validation cohort. Cost-Benefit Analysis Finally, we performed a cost-benefit analysis (eFigure 5 in the Supplement) where we show a scenario in which 1000 Aβ-positive participants are included in a trial where the screening cost for Aβ PET is $4000 per participant.43 For example, using the highest Youden index cutoff (Table 2) for plasma Aβ42, Aβ40, and APOE reduces the number of PET scans by approximately 800 and lowers the PET costs by approximately $3.2 million (from a total cost of approximately $9.2 million).

rial where the screening cost for Aβ PET is $4000 per participant.43 For example, using the highest Youden index cutoff (Table 2) for plasma Aβ42, Aβ40, and APOE reduces the number of PET scans by approximately 800 and lowers the PET costs by approximately $3.2 million (from a total cost of approximately $9.2 million). Discussion In this study of 842 participants, we found that plasma Aβ42 and Aβ40 using the fully automated Elecsys platform detected abnormal levels of Aβ in the brain with an AUC of 0.80 (Figure 2A and B). The addition of APOE genotype increased the AUC significantly to 0.85 (Figure 2A and B). Plasma tau and NFL had a slight effect on the accuracy (AUC, +0.01 to 0.02; Figure 2A and B). The results were similar in cognitively impaired and unimpaired and older and younger participants (Figure 3), with the exception that plasma tau and NFL generally did not improve accuracy in addition to plasma Aβ and APOE genotype in cognitively impaired participants (eTable 4 in the Supplement). When applying the plasma Aβ42 and Aβ40 model from BioFINDER to the independent validation cohort (n = 237), the AUC was greater compared with BioFINDER (AUC, 0.86; 95% CI, 0.81-0.91), but no improvement was seen when adding plasma tau (Figure 2C and D).

in cognitively impaired participants (eTable 4 in the Supplement). When applying the plasma Aβ42 and Aβ40 model from BioFINDER to the independent validation cohort (n = 237), the AUC was greater compared with BioFINDER (AUC, 0.86; 95% CI, 0.81-0.91), but no improvement was seen when adding plasma tau (Figure 2C and D). Although previous studies have found associations between CSF and PET Aβ and plasma Aβ using different immunoassays,7,12,13,14,44,45 the present Elecsys assays produced among the best accuracies and they are the first fully automated assays to have these greater accuracies. In mass spectrometry–based techniques, 2 recent studies have provided overall better accuracies for plasma Aβ42/Aβ40 (AUC, 0.84-0.97 depending on population and reference standard).19,20 However, these are labor-intensive, time-consuming, low-throughput methods that currently are not feasible to implement in clinical practice on a large scale. Fully automated Elecsys assays, on the other hand, are already implemented in many clinical chemistry laboratories worldwide that provide analyses (eg, for primary care).

these are labor-intensive, time-consuming, low-throughput methods that currently are not feasible to implement in clinical practice on a large scale. Fully automated Elecsys assays, on the other hand, are already implemented in many clinical chemistry laboratories worldwide that provide analyses (eg, for primary care). Historically, the ratio of plasma Aβ42 to Aβ40 has been used to optimize the concordance with CSF or PET Aβ. Here, Aβ40 acts as a reference peptide that accounts for interindividual variability in the overall Aβ production and CSF turnover. We found that instead of using the fixed ratio of Aβ42/Aβ40, both the model fit (AIC) and accuracy (AUC) were slightly but significantly improved when the model was adjusted for Aβ40 concentrations independent of Aβ42 (ie, used as a separate predictor in the logistic regression models) (AUC 0.77 vs 0.80; P = .01; ΔAIC –66; eTable 4 in the Supplement). As a single additional biomarker to Aβ42 and Aβ40, APOE genotype increased the accuracy most markedly, from AUC 0.80 to 0.85 (P < .001) (Figure 2A and B; eTable 4 in the Supplement). Plasma tau increased the AUC slightly, and provided a better model fit (ΔAIC –27), but clinically this is not comparable with the contribution CSF tau has combined with CSF Aβ42,24 and improved plasma tau assays are probably needed in the future, such as measurement of specifically phosphorylated tau,46 to increase the added value of plasma tau to plasma Aβ42 and Aβ40.

a better model fit (ΔAIC –27), but clinically this is not comparable with the contribution CSF tau has combined with CSF Aβ42,24 and improved plasma tau assays are probably needed in the future, such as measurement of specifically phosphorylated tau,46 to increase the added value of plasma tau to plasma Aβ42 and Aβ40. Despite the present and previous results showing relatively high correlations between plasma and CSF NFL (eTable 3 in the Supplement and the study by Hansson et al47), we saw a modest increase in accuracy in addition to plasma Aβ42 and Aβ40 (Figure 2; eTable 4 in the Supplement). Because NFL generally is late biomarker in the disease process and a non–AD specific biomarker for axonal degeneration,15 the poor result could be because most of the participants (513 of 842) were cognitively unimpaired and only 64 had AD dementia. Compared with plasma tau and NFL, plasma NFH had a poorer performance and did not improve accuracy (eTable 5 in the Supplement). However, 101 plasma NFH measurements were below the detection limit of the assays, and development of more sensitive plasma NFH assays is thus warranted to establish whether this biomarker could further improve the diagnostic performance of plasma Aβ.

rformance and did not improve accuracy (eTable 5 in the Supplement). However, 101 plasma NFH measurements were below the detection limit of the assays, and development of more sensitive plasma NFH assays is thus warranted to establish whether this biomarker could further improve the diagnostic performance of plasma Aβ. Overall, the accuracies of the Aβ42 and Aβ40 assays are not sufficient to be used on their own as a clinical test of Aβ positivity; additional assay development is needed before this can be recommended, possibly together with other blood biomarkers and screening tools in diagnostic algorithms. In the present study, we showed that the Aβ assays perform similarly in CU populations with lower prevalence of Aβ positivity (Figure 3A and B; Aβ-positive prevalence, 29%). Nonetheless, further studies would be valuable in populations with lower prevalence of Aβ positivity, such as primary care settings, as well as more heterogeneous dementia cohorts with different neurodegenerative disorders. To some extent, the generalizability of the BioFINDER results has already been shown in the present study where the plasma Aβ42 and Aβ40 model established in BioFINDER could be applied in the independent validation cohort with better accuracy (AUC, 0.86 vs 0.80; Figure 2). This robust result is similar to what has been shown when using the Elecsys assays for CSF to establish a cutoff in one cohort and replicating it in a second cohort.24

a Aβ42 and Aβ40 model established in BioFINDER could be applied in the independent validation cohort with better accuracy (AUC, 0.86 vs 0.80; Figure 2). This robust result is similar to what has been shown when using the Elecsys assays for CSF to establish a cutoff in one cohort and replicating it in a second cohort.24 Limitations Limitations of the present validation analysis include the lack of APOE data, the lack of improvement when replicating the model that included plasma tau, and the smaller population size, resulting in a lack of analyses in subpopulations. The latter was, however, tested in BioFINDER and the accuracies were similar in different subsamples including CU (Figure 3A and B) and younger participants (Figure 3E and F) where Aβ positivity might be more difficult to identify using alternative methods such as cognitive testing and age stratification.40

opulations. The latter was, however, tested in BioFINDER and the accuracies were similar in different subsamples including CU (Figure 3A and B) and younger participants (Figure 3E and F) where Aβ positivity might be more difficult to identify using alternative methods such as cognitive testing and age stratification.40 Conclusions From a practical perspective, we believe that the most advantageous future use of optimized blood Aβ assays is as a screening tool for identifying subjects at a higher risk of being Aβ positive. They could, for example, be applied as an initial test together with other noninvasive, cost-efficient tools that aid the decision about whom a general practitioner should refer for further investigation at memory clinics where CSF or PET and more extensive clinical assessment could be used to support the AD diagnosis. Another useful setting for the blood biomarkers are clinical AD trials enrolling Aβ-positive participants, where they can be used for prescreening to minimize the number of unnecessary (Aβ-negative) lumbar punctures and Aβ PET scans, as well as lowering the costs for the examinations up to 30% to 50% depending on the cutoff (eFigure 5 in the Supplement).48 Although further validation studies are needed, this illustrates the potential usefulness blood assays might have, especially considering the ongoing great need to recruit large cohorts for AD drug trials in preclinical and prodromal stages. Supplement. eMethods. eResults. eTable 1. Performance Characteristics of the Plasma Aβ42, Aβ40 and Tau Elecsys Assays

Introduction Antibodies to myelin oligodendrocyte glycoprotein (MOG), detected by live cell-based assays, are found in a proportion of patients with acquired demyelinating syndromes (ADSs) of the central nervous system from early childhood through late adulthood. Since some of these MOG antibody–positive individuals will remain monophasic while others will relapse, identifying those destined for recurrent disease has important implications for both prognosis and treatment decisions. This is particularly relevant in the pediatric age group, where MOG antibodies can be found frequently (up to 40% of those with ADSs), with studies variably reporting subsequent relapsing disease in 36% to 61% of cases. While it has been suggested that persistent MOG antibody seropositivity following ADS is associated with relapsing disease in both children and adults, the studies to date assessing serial serologic MOG antibody status may have preferentially selected patients prone to relapse, resulting in an overestimation of relapse risk. Here, we report on the long-term clinical and imaging outcomes and their association with serial serum MOG antibody measurements in a large unselected cohort of children with ADS followed prospectively from clinical onset.

referentially selected patients prone to relapse, resulting in an overestimation of relapse risk. Here, we report on the long-term clinical and imaging outcomes and their association with serial serum MOG antibody measurements in a large unselected cohort of children with ADS followed prospectively from clinical onset. Methods Participants The Canadian Pediatric Demyelinating Disease Study recruited and longitudinally observed a total of 430 children presenting with a first clinical episode consistent with ADS between July 2004 and February 2017. Of these, all participants for whom a first serum sample was available within 45 days of presentation (n = 274) contributed to the current study (Figure 1A). All participants underwent comprehensive clinical assessments and were offered brain magnetic resonance imaging (MRI) scans and collection of serum samples at the time of clinical presentation; at 3, 6, and 12 months postenrollment; and annually thereafter. Clinical attacks or new MRI lesions were considered evidence of new disease activity if occurring more than 30 days from onset or 90 days in case of an acute disseminated encephalomyelitis (ADEM) presentation. Data were locked as of May 2018, and participant status was adjudicated based on information available as of their most recent examination to confer a diagnosis of either monophasic ADS (defined as the absence of new clinical attacks and of any MRI evidence of new disease activity on all serial examinations) or of multiple sclerosis (MS), neuromyelitis optica spectrum disorder, or relapsing non-MS disease, using established criteria. Guardians and participants provided written informed consent. Younger children provided verbal assent. The study was approved by The Hospital for Sick Children and the research ethics boards of all participating institutions.

uromyelitis optica spectrum disorder, or relapsing non-MS disease, using established criteria. Guardians and participants provided written informed consent. Younger children provided verbal assent. The study was approved by The Hospital for Sick Children and the research ethics boards of all participating institutions. Figure 1. Study Design A, A total of 2022 serial samples from 274 participants with acquired demyelinating syndrome (ADS) with complete clinical data were tested for anti–myelin oligodendrocyte glycoprotein (MOG) IgG1 and anti–aquaporin 4 (AQP4) antibodies. Seven participants were recruited after July 2014, at which time the protocol entered a new phase, requiring participants to meet McDonald 2010 diagnostic criteria. None of these 7 patients were positive for anti-MOG antibodies. B, Example images of typical positive, borderline, and negative results for MOG IgG1 antibody testing.

cipants were recruited after July 2014, at which time the protocol entered a new phase, requiring participants to meet McDonald 2010 diagnostic criteria. None of these 7 patients were positive for anti-MOG antibodies. B, Example images of typical positive, borderline, and negative results for MOG IgG1 antibody testing. Clinical Characteristics Clinical features at presentation were recorded, including the presence of signs and symptoms indicating optic neuritis (ON) or transverse myelitis (TM) and those meeting the criteria for ADEM. Clinical course was evaluated in terms of time from ADS onset to first relapse, annualized relapse rate over the first 4 years (computed including the presenting episode), and relapse phenotype. Disability outcome was recorded using extrapolated Expanded Disability Status Scale (EDSS) scores, as described previously. The functional system score for visual function was also evaluated. To ensure that EDSS and visual functional system scores could be compared at a consistent time from onset across all participants, we selected the scores obtained at 4 years following clinical presentation (the median postonset observation time of the cohort). Dates and durations of treatment with disease-modifying therapies were recorded.

nd visual functional system scores could be compared at a consistent time from onset across all participants, we selected the scores obtained at 4 years following clinical presentation (the median postonset observation time of the cohort). Dates and durations of treatment with disease-modifying therapies were recorded. MOG Antibody Status and Cerebrospinal Fluid Studies Myelin oligodendrocyte glycoprotein antibody status was assessed in 2022 serum samples. Of these, 5 samples obtained within 30 days of plasma exchange were excluded. All samples were tested using a live MOG IgG1-specific cell-based assay and scored semiquantitatively by a single blinded operator (Figure 1B). Any positive sample was titrated using antihuman IgG (H+L) as detecting antibody. Results of the IgG1 assay in samples obtained within 45 days from clinical onset were used to determine the serologic status at presentation. For assessment of serial samples, subsequent samples with a titer of 1:200 or greater were considered seropositive. All samples were also screened using a live cell-based aquaporin 4 (AQP4) assay as a specificity control. Where indicated, presence of cerebrospinal fluid oligoclonal bands was assessed by isoelectric focusing as part of clinical evaluation.

sequent samples with a titer of 1:200 or greater were considered seropositive. All samples were also screened using a live cell-based aquaporin 4 (AQP4) assay as a specificity control. Where indicated, presence of cerebrospinal fluid oligoclonal bands was assessed by isoelectric focusing as part of clinical evaluation. MRI Analysis Research MRI scans included axial and sagittal T2-weighted images, fluid-attenuated inversion recovery images, T1-weighted images, and T1-weighted images after administration of gadolinium. Scans acquired within 45 days of clinical presentation were analyzed by applying a standardized scoring tool modified to include additional MRI variables and blinded to MOG status and to clinical outcome. Serial scans were analyzed for the occurrence of new T2 lesions and lesion resolution. T2 lesions were segmented using a bayesian classifier and then manually corrected by a trained expert. T1 lesion volumes were automatically segmented as a subset of T2 lesion voxels with T1 intensity less than 87% of the normal-appearing white matter. When available, clinical scans obtained apart from the research protocol were also evaluated for the presence of new lesions.

then manually corrected by a trained expert. T1 lesion volumes were automatically segmented as a subset of T2 lesion voxels with T1 intensity less than 87% of the normal-appearing white matter. When available, clinical scans obtained apart from the research protocol were also evaluated for the presence of new lesions. Statistical Analysis Figure 1A summarizes the study design. First, presenting clinical, MRI, and laboratory features were compared between anti-MOG antibody–positive and anti-MOG antibody–negative participants, as defined by the result of their baseline sample. The analysis was also repeated separately in participants older and younger than 11 years at onset and in participants with or without ADEM at presentation. We then compared the presenting clinical and MRI features of participants who were persistently positive for anti-MOG antibodies (defined as anti-MOG antibody–positive results in all serial samples up to and including at least 1 year) with those who became negative for anti-MOG antibodies. To minimize confounding of missing values and different lengths of follow-up, we conducted an additional sensitivity analysis focusing on participants for whom samples were available at all time points of baseline, 3 months, 1 year, and 2 years.

g at least 1 year) with those who became negative for anti-MOG antibodies. To minimize confounding of missing values and different lengths of follow-up, we conducted an additional sensitivity analysis focusing on participants for whom samples were available at all time points of baseline, 3 months, 1 year, and 2 years. Clinical and MRI courses were compared between participants positive and negative for anti-MOG antibodies with a minimum follow-up of 6 months (except for annualized relapse rate and EDSS, which were further restricted to participants observed for a minimum of 4 years). The analysis was repeated after excluding participants with a diagnosis of MS. The presenting features were then compared between anti-MOG antibody–positive participants with monophasic or relapsing course and used as inputs for a random forest classifier (1000 trees; Python SKLearn). A backward elimination process was applied to identify features that could work in combination to predict a relapsing outcome, with the feature with the lowest Gini impurity removed from the input at each repetition.

sic or relapsing course and used as inputs for a random forest classifier (1000 trees; Python SKLearn). A backward elimination process was applied to identify features that could work in combination to predict a relapsing outcome, with the feature with the lowest Gini impurity removed from the input at each repetition. Feature comparisons between groups were performed using χ2 or Fisher exact tests for categorical variables and Mann-Whitney U tests for continuous variables. The correlation between continuous variables was assessed by Spearman rank correlation. A P value less than .05 was considered statistically significant, and all P values were 2-tailed. Time to conversion to seronegative status was plotted using the Kaplan-Meier method. Analyses were performed using Python version 3.6.5 (Python Software Foundation) and R version 3.1.3 (The R Foundation). Results Presenting Features of the ADS Cohort The presence of MOG and AQP4 antibodies was assessed in 274 children with ADS, of whom 140 (51.1%) were female, and the median (interquartile range [IQR]) age of all participants was 10.8 (6.2-13.9) years (Table 1). Myelin oligodendrocyte glycoprotein antibodies were detected in 84 participants (30.7%), while 179 (65.3%) tested negative and 11 (4.0%) had borderline results. Aquaporin 4 antibodies were detected in 1 participant, who was negative for anti-MOG antibodies. Anti-MOG antibody–positive participants were younger than anti-MOG antibody–negative participants, and 81 of 84 (96%) presented with ON, TM, ADEM, or a combination of these.

ed negative and 11 (4.0%) had borderline results. Aquaporin 4 antibodies were detected in 1 participant, who was negative for anti-MOG antibodies. Anti-MOG antibody–positive participants were younger than anti-MOG antibody–negative participants, and 81 of 84 (96%) presented with ON, TM, ADEM, or a combination of these. Table 1. Demographic, Clinical, Imaging, and Cerebrospinal Fluid (CSF) Laboratory Features at Presentation by Initial Myelin Oligodendrocyte Glycoprotein Status in All Participants and in Participants Stratified by Age at Clinical Onset Characteristic No./Total No. (%) All Participants <11 y ≥11 y Total Borderline Seropositive Seronegative P Valuea Seropositive Seronegative P Valuea Seropositive Seronegative P Valuea Demographic and Clinical Features at Presentation Participants, No.

Participants and in Participants Stratified by Age at Clinical Onset Characteristic No./Total No. (%) All Participants <11 y ≥11 y Total Borderline Seropositive Seronegative P Valuea Seropositive Seronegative P Valuea Seropositive Seronegative P Valuea Demographic and Clinical Features at Presentation Participants, No. 274 11 84 179 NA 65 69 NA 19 110 NA Time from onset to first serum procurement, median (IQR), d 9 (5-19) 22 (12-28) 10 (5-20) 9 (5-17) .23 10 (6-21) 8 (4-15) .06 12 (4-18) 10 (5-18) .33 Female 140/274 (51.1) 7/11 (64) 46/84 (55) 87/179 (48.6) .42 38/65 (58) 30/69 (43) .12 8/19 (42) 57/110 (51.8) .59 Age at clinical onset, median (IQR), y 10.80 (6.18-13.87) 6.18 (5.44-11.32) 7.31 (4.93-10.57) 12.41 (8.09-14.43) <.001 6.27 (4.44-8.65) 6.31 (2.79-9.57) .46 13.31 (12.62-14.35) 14.04 (12.82-15.31) .06 Duration of clinical follow up, median (IQR), y 6.68 (4.24-8.86) 6.92 (4.36-8.05) 6.74 (4.77-8.75) 6.35 (4.08-8.99) .36 6.91 (4.79-8.45) 7.56 (4.23-9.01) .32 6.68 (4.18-9.08) 6.03 (4.07-8.69) .42 Presenting phenotypeb ADEM 67/274 (24.5) 5/11 (45) 32/84 (38) 30/179 (16.8) <.001 29/65 (45) 22/69 (32) .18 3/19 (16) 8/110 (7.3) .21 ADEM with ON 3/274 (1.1) 0 3/84 (4) 0 .03 2/65 (3) 0 .23 1/19 (5) 0 .15 ADEM with TM 10/274 (3.6) 0 9/84 (11) 1/179 (0.6) <.001 9/65 (14) 0 .001 0 1/110 (0.9) >.99 ADEM with ON and TM 2/274 (0.7) 1/11 (9) 1/84 (1) 0 .32 1/65 (2) 0 .49 0 0 >.99 ON Monofocal 68/274 (24.8) 3/11 (27) 32/84 (38) 33/179 (18.4) <.001 21/65 (32) 10/69 (14) .03 11/19 (58) 23/110 (20.9) .002 Polyfocal 12/274 (4.4) 1/11 (9) 2/84 (2) 9/179 (5.0) .51 0 2/69 (3) .50 2/19 (11) 7/110 (6.4) .62 TM Monofocal 53/274 (19.3) 0 7/84 (8) 46/179 (25.7) .002 6/65 (9) 14/69 (20) .12 1/19 (5) 32/110 (29.1) .04 Polyfocal 15/274 (5.5) 0 5/84 (6) 10/179 (5.6) .87 4/65 (6) 6/69 (9) .75 1/19 (5) 4/110 (3.6) .56 ON with TM 4/274 (1.5) 0 3/84 (4) 1/179 (0.6) .04 2/65 (3) 0 .49 1/19 (5) 1/110 (0.9) .27 Other 55/274 (20.1) 2/11 (18) 3/84 (4) 50/179 (27.9) <.001 3/65 (5) 15/69 (22) .005 0 35/110 (31.8) .002 MRI and CSF Laboratory Features at Presentationc Brain lesions Patients with lesions 172/253 (68.0) 9/11 (82) 51/76 (67) 112/166 (67.5) >.99 40/57 (70) 44/67 (66) .70 11/19 (58) 68/99 (69) .52 Count, median (IQR)d 3 (0->15) 8 (0.75->15) 5.5 (0->15) 2 (0->15) .13 14 (0->15) 2 (0-13) .02 1 (0-4.5) 4 (0->15) .07 Excluding patients without lesions 12 (3->15) >15 (6->15) >15 (5.5->15) 8 (2->15) .02 >15 (10.75->15) 7 (2->15) .001 4 (2-12) 9 (3->15) .07 Presence of ≥1 discrete lesions 117/172 (68.0) 2/9

99 (69) .52 Count, median (IQR)d 3 (0->15) 8 (0.75->15) 5.5 (0->15) 2 (0->15) .13 14 (0->15) 2 (0-13) .02 1 (0-4.5) 4 (0->15) .07 Excluding patients without lesions 12 (3->15) >15 (6->15) >15 (5.5->15) 8 (2->15) .02 >15 (10.75->15) 7 (2->15) .001 4 (2-12) 9 (3->15) .07 Presence of ≥1 discrete lesions 117/172 (68.0) 2/9 (22) 27/51 (53) 88/112 (78.6) .002 18/40 (45) 23/44 (52) .65 9/11 (82) 65/68 (96) .14 Only well-defined lesions 83/172 (48.3) 1/9 (11) 14/51 (27) 68/112 (60.7) <.001 7/40 (18) 16/44 (36) .09 7/11 (64) 52/68 (76) .46 Diffuse bilateral pattern 61/172 (35.5) 6/9 (67) 31/51 (61) 24/112 (21.4) <.001 28/40 (70) 18/44 (41) .01 3/11 (27) 6/68 (9) .11 ≥1 Cerebellar lesions 67/172 (39.0) 5/9 (56) 22/51 (43) 40/112 (35.7) .46 19/40 (48) 17/44 (39) .55 3/11 (27) 23/68 (34) >.99 ≥1 Cerebellar peduncle lesions 52/172 (30.2) 2/9 (22) 20/51 (39) 30/112 (26.8) .16 14/40 (35) 13/44 (30) .76 6/11 (55) 17/68 (25) .10 ≥1 Brainstem lesions 103/172 (59.9) 7/9 (78) 34/51 (67) 62/112 (55.4) .23 25/40 (62) 27/44 (61) .91 9/11 (82) 35/68 (51) .10 ≥1 Periventricular lesions 98/172 (57.0) 4/9 (44) 30/51 (59) 64/112 (57.1) .98 25/40 (62) 17/44 (39) .049 5/11 (45) 47/68 (69) .23 ≥3 Periventricular lesions 59/172 (34.3) 3/9 (33) 19/51 (37) 37/112 (33.0) .73 18/40 (45) 6/44 (14) .003 1/11 (9) 31/68 (46) .02 ≥1 Lesion perpendicular to major axis of corpus callosum 53/172 (30.8) 2/9 (22) 7/51 (14) 44/112 (39.3) .002 3/40 (8) 5/44 (11) .72 4/11 (36) 39/68 (57) .21 ≥1 Basal ganglia lesions 37/172 (21.5) 4/9 (44) 17/51 (33) 16/112 (14.3) .009 17/40 (42) 9/44 (20) .05 0 7/68 (10) .58 ≥1 Thalamic lesions 57/172 (33.1) 4/9 (44) 30/51 (59) 23/112 (20.5) <.001 27/40 (68) 13/44 (30) .001 3/11 (27) 10/68 (15) .38 ≥1 Juxtacortical lesions 116/172 (67.4) 8/9 (89) 42/51 (82) 66/112 (58.9) .006 37/40 (92) 24/44 (55) <.001 5/11 (45) 42/68 (62) .49 ≥1 T1 hypointense lesions 77/172 (44.8) 1/9 (11) 15/51 (29) 61/112 (54.5) .005 15/40 (38) 12/44 (27) .44 0 49/68 (72) <.001 ≥1 Lesion enhancement 47/149 (31.5) 2/7 (29) 5/42 (12) 40/100 (40.0) .002 5/34 (15) 7/38 (18) .92 0 33/62 (53) .006 ≥1 Gadolinium-negative T1 hypointense lesions 54/149 (36.2) 0 11/42 (26) 43/100 (43.0) .09 11/34 (32) 7/38 (18) .28 0 36/62 (58) .002 OCBs 47/169 (27.8) 0 8/49 (16) 39/113 (34.5) .02 6/38 (16) 8/41 (20) .89 2/11 (18) 31/72 (43) .19 Abbreviations: ADEM, acute disseminated encephalomyelitis; IQR, interquartile range; MRI, magnetic resonance imaging; NA, not applicable; OCB, oligoclonal band; ON, optic neuritis; TM, transverse myeliti

0 36/62 (58) .002 OCBs 47/169 (27.8) 0 8/49 (16) 39/113 (34.5) .02 6/38 (16) 8/41 (20) .89 2/11 (18) 31/72 (43) .19 Abbreviations: ADEM, acute disseminated encephalomyelitis; IQR, interquartile range; MRI, magnetic resonance imaging; NA, not applicable; OCB, oligoclonal band; ON, optic neuritis; TM, transverse myeliti s. a P values are computed based on comparisons between seropositive and seronegative participants. Participants with borderline results were not considered in the comparisons of participants stratified by age. b For each participant, the clinical presentation was classified as monofocal if all clinical deficits were localizable to a single central nervous system site, while findings implicating more than 1 central nervous system location were classified as polyfocal. Magnetic resonance imaging scans were not used to define monofocal or polyfocal designations. c Brain MRI scans acquired within 45 days from onset were available in 253 participants. The frequencies of all features pertaining to lesion aspect and location were computed only among patients with brain lesions at baseline. The denominator for each feature corresponds to the total number of participants in which that feature was evaluated. Analyses of lesion enhancement and gadolinium-negative T1 hypointense lesions were further restricted to participants who had gadolinium administered. d The total number of T2 lesions were estimated by manual count, with lesion numbers exceeding 15 listed as greater than 15.

c Brain MRI scans acquired within 45 days from onset were available in 253 participants. The frequencies of all features pertaining to lesion aspect and location were computed only among patients with brain lesions at baseline. The denominator for each feature corresponds to the total number of participants in which that feature was evaluated. Analyses of lesion enhancement and gadolinium-negative T1 hypointense lesions were further restricted to participants who had gadolinium administered. d The total number of T2 lesions were estimated by manual count, with lesion numbers exceeding 15 listed as greater than 15. Several clinical and MRI features differed by age at presentation. Consistent with the common presentation of ADEM, anti-MOG antibody–positive participants younger than 11 years had a high number of lesions that were often ill defined and/or distributed in a diffuse bilateral pattern, with frequent thalamic and juxtacortical involvement. Conversely, anti-MOG antibody–positive participants older than 11 years presented with fewer focal lesions, and 8 of 19 (42%) (all presenting with ON) had normal brain MRI findings at onset. When present in older anti-MOG antibody–positive participants, brain lesions more frequently showed well-defined borders. The full description of clinical, MRI, and laboratory features relative to MOG status at presentation is presented in Table 1, and a comparison between participants with and without ADEM is reported in eTable 1 in the Supplement.

antibody–positive participants, brain lesions more frequently showed well-defined borders. The full description of clinical, MRI, and laboratory features relative to MOG status at presentation is presented in Table 1, and a comparison between participants with and without ADEM is reported in eTable 1 in the Supplement. Temporal Evolution of Serologic Status The evolution of serological status was evaluated in 216 participants with samples available for follow-up more than 1 year after baseline (median [IQR] follow-up, 4.70 [3.07-6.11] years) (Figure 2A). A total of 137 of 139 participants (98.6%) who were negative for anti-MOG antibodies at presentation remained negative in all subsequent examinations.

ty was measured using a custom-designed digital pathology macro on hematoxylin-eosin–stained sections of the nbM. Hippocampal sparing (HpSp) AD has fewer remaining neurons compared with typical AD, which has fewer compared with limbic predominant AD. Pairwise comparisons were performed using Mann-Whitney rank sum test. To examine the overlap in NFT accumulation and neuronal density differences observed among AD subtypes in Figure 2, we performed multivariable regression analyses within each AD subtype for a total of 6 models. Table 2 summarizes the expected change in NFT accumulation and neuronal density in the nbM, adjusted for demographic and clinical variables of interest and their associations within the AD subtype. Younger age at onset of cognitive symptoms was significantly associated with higher NFT counts in the nbM of HpSp AD (P < .001) (Figure 3A). Thus, for every 10 years younger age at onset, the number of NFTs was expected to be higher by 1.5 (95% CI, −2.9 to −0.15; P = .03) in HpSp AD cases and by 3.2 (95% CI, −3.9 to −2.4; P < .001) in typical AD cases. In addition, within the typical AD cases, females were expected to have 2.5 (95% CI, 1.4-3.5) more NFTs than males (P < .001) and APOE ε4 carriers to have 1.3 (95% CI, 0.15-2.5) more NFTs than APOE ε4 noncarriers (P = .03). For every 10-point decrease in final MMSE of typical AD cases, the number of nbM NFTs was expected to increase by 1.8 (95% CI, −3.2 to −0.31; P = .02). Accumulation of NFTs in the nbM of limbic predominant AD cases was not associated with the observed demographic and clinical variables. Regression analyses of neuronal density largely reflected the same pattern of associations with demographic and clinical variables observed with NFT accumulation in HpSp AD and typical AD (Table 2, eAppendix 2 in the Supplement). However, new associations with neuronal density emerged in limbic predominant AD. For every 10 years’ younger age at onset, the number of neurons was expected to be lower by 4.6 (95% CI, 2.3-7.0) in limbic predominant AD cases (P < .001) (Figure 3B). In addition, limbic predominant cases were observed to have 4.3 fewer neurons (95% CI, 0.47-8.1) for every 10-point decrease in MMSE.

atus was evaluated in 216 participants with samples available for follow-up more than 1 year after baseline (median [IQR] follow-up, 4.70 [3.07-6.11] years) (Figure 2A). A total of 137 of 139 participants (98.6%) who were negative for anti-MOG antibodies at presentation remained negative in all subsequent examinations. Figure 2. Myelin Oligodendrocyte Glycoprotein (MOG) Results on Serial Samples A, Evolution of serologic status in participants with follow-up greater than 1 year. Only 2 of 139 participants who were negative for anti-MOG antibodies at presentation changed serological status in subsequent examinations: 1 participant became seropositive at 3 months and was persistently positive for the subsequent 8 years and 1 had an isolated finding of a borderline result at 2 years from onset. Of the 10 participants with borderline results at onset, 7 became seronegative, 2 fluctuated between negative and borderline status and 1 became seropositive in the last follow-up sample (7 years from onset; titer, 1:200). B, Kaplan-Meier curve for the time to conversion to seronegative status in all participants who were positive for anti-MOG antibodies at time of presentation. The shaded areas indicate 95% CIs. C, Kaplan-Meier curves in participants with acute disseminated encephalomyelitis (ADEM) vs non-ADEM presentations. Participants with ADEM had a considerably shorter time to seroconversion than participants without ADEM. The shaded areas indicate 95% CIs. D and E, Trajectory of serial MOG titers of anti-MOG antibody–positive participants with ADEM at onset (D) and without ADEM at onset (E). The data points indicate the titers measured at each serological evaluation.

derably shorter time to seroconversion than participants without ADEM. The shaded areas indicate 95% CIs. D and E, Trajectory of serial MOG titers of anti-MOG antibody–positive participants with ADEM at onset (D) and without ADEM at onset (E). The data points indicate the titers measured at each serological evaluation. Of the 67 participants who were positive for anti-MOG antibodies at onset and observed for at least 1 year from presentation, 24 (36%) remained persistently positive, 5 (7%) fluctuated between positive and negative status, and 38 (57%) became seronegative (median [IQR] time to conversion, 1.02 [0.32-1.75] years) (Figure 2B). Of 22 initially anti-MOG antibody–positive participants with ADEM with at least 1 year of follow-up, 17 (77%) became seronegative after a median (IQR) of 4.63 (3.45-12.71) months (Figure 2C). Participants who were persistently seropositive tended to be older at presentation compared with those who converted to seronegative status (median [IQR] age, 9.06 [6.60-13.36] years vs 6.95 [5.28-9.96] years; P = .03) and were more likely to have presented with monofocal ON (14 of 24 [58%] vs 9 of 38 [24%]; P = .01) (eTable 2 in the Supplement).

tently seropositive tended to be older at presentation compared with those who converted to seronegative status (median [IQR] age, 9.06 [6.60-13.36] years vs 6.95 [5.28-9.96] years; P = .03) and were more likely to have presented with monofocal ON (14 of 24 [58%] vs 9 of 38 [24%]; P = .01) (eTable 2 in the Supplement). Myelin oligodendrocyte glycoprotein titers showed a global decrease over follow-up that was steeper in the first year (Figure 2D). These changes in MOG antibody titers were unlikely to be influenced by treatment, as only 2 of 84 participants (2%) who were positive for anti-MOG antibodies at the time of initial presentation were ever exposed to disease-modifying therapies during their follow-up. In the sensitivity analysis of the 126 participants for whom all samples at presentation, 3 months, 1 year, and 2 years were available, the proportion of participants who converted to seronegative status was consistent with findings observed in the larger cohort (eFigure 1 in the Supplement).

es during their follow-up. In the sensitivity analysis of the 126 participants for whom all samples at presentation, 3 months, 1 year, and 2 years were available, the proportion of participants who converted to seronegative status was consistent with findings observed in the larger cohort (eFigure 1 in the Supplement). Clinical Disease Course Diagnostic Outcome as a Function of MOG Status at Presentation During the prospective follow-up, 66 participants met McDonald 2010 diagnostic criteria for MS, while 11 were diagnosed as having a relapsing non-MS disease, 1 as having AQP4-positive neuromyelitis optica spectrum disorder, and 196 as having a monophasic course. Among the 66 children who eventually met diagnostic criteria for MS, 11 (17%) were positive for anti-MOG antibodies at presentation. Despite meeting formal MS diagnostic criteria, all 11 of these participants had clinical and MRI features that were considered by the study clinicians to be atypical for MS (eg, clinical attacks largely restricted to ON, near-complete resolution of initial T2 lesions, no or few small lesions on follow-up scans). At the time of the most recent evaluation of these 11 patients (mean [range] follow-up from initial presentation, 8.96 [1.69-12.13] years), none of these participants were considered by their treating clinician to have a clinical disease course consistent with MS.

2 lesions, no or few small lesions on follow-up scans). At the time of the most recent evaluation of these 11 patients (mean [range] follow-up from initial presentation, 8.96 [1.69-12.13] years), none of these participants were considered by their treating clinician to have a clinical disease course consistent with MS. Myelin oligodendrocyte glycoprotein antibodies were found in almost all participants eventually diagnosed as having relapsing non-MS disease (10 of 11 [91%]) and in 63 of 196 participants with monophasic ADS (32.1%). Of the 10 anti-MOG antibody–positive children with relapsing non-MS disease, 6 had relapsing ON, 1 had relapsing TM, and 3 had multiple neurological relapses (including either ON or TM and other neurological deficits) following initial ADEM presentation.

[91%]) and in 63 of 196 participants with monophasic ADS (32.1%). Of the 10 anti-MOG antibody–positive children with relapsing non-MS disease, 6 had relapsing ON, 1 had relapsing TM, and 3 had multiple neurological relapses (including either ON or TM and other neurological deficits) following initial ADEM presentation. Clinical and MRI Disease Activity Outcomes as a Function of MOG Status at Time of Initial Presentation Clinical relapses occurred in similar proportions of patients positive and negative for anti-MOG antibodies (53 of the latter met criteria for MS) observed for a minimum of 6 months from presentation (Table 2). When excluding participants who met diagnostic criteria for MS, the occurrence of clinical relapses was more common among anti-MOG antibody–positive participants (9 of 71 [13%] vs 2 of 121 [2%]; P = .002), ie, 10 of 11 participants with relapsing non-MS disease (91%) were positive for anti-MOG antibody at presentation. Relapses in anti-MOG antibody–positive participants were more commonly restricted to the optic nerve or spinal cord. A total of 5 of 20 anti-MOG antibody–positive participants (25%) who presented with monofocal ON and had normal findings on brain MRI had subsequent clinical relapses compared with none of the 17 anti-MOG antibody–negative children with the same presentation (P = .049). Additionally, 32 children who presented with monofocal TM had normal findings on brain MRI at onset (3 anti-MOG antibody–positive and 29 anti-MOG antibody–negative children), only 1 of whom (who was negative for anti-MOG antibodies) experienced a relapse (TM). Among participants presenting with ADEM, clinical relapses occurred in 3 of 32 participants (9%) who were positive for anti-MOG antibodies at presentation vs 1 of 30 (3%) who were seronegative. While EDSS scores were comparably low across seropositive and seronegative participants at 4 years of follow-up (Table 2), anti-MOG antibody–positive participants with a monophasic disease course had lower EDSS scores at 4 years than monophasic anti-MOG antibody–negative participants (median [IQR] score, 0 [0-1] vs 0 [0-1.5]; P = .007). This difference did not persist when the comparison was restricted to participants with the same clinical presentation of monofocal ON, monofocal TM, or ADEM.

sic disease course had lower EDSS scores at 4 years than monophasic anti-MOG antibody–negative participants (median [IQR] score, 0 [0-1] vs 0 [0-1.5]; P = .007). This difference did not persist when the comparison was restricted to participants with the same clinical presentation of monofocal ON, monofocal TM, or ADEM. Table 2. Clinical and Magnetic Resonance Imaging (MRI) Disease Activity as a Function of Baseline Myelin Oligodendrocyte Glycoprotein Status in Participants With More Than 6 Months’ Follow-up Characteristic No./Total No. (%) Total With Non-MS Disease Borderline Seropositive Seronegative P Valuea Borderline Seropositive Seronegative P Valuea Participants, No.b 11 82 174 NA 10 71 121 NA Duration of clinical follow-up, median (IQR), y 6.92 (4.63-8.05) 6.86 (4.79-8.87) 6.78 (4.30-9.01) .38 6.41 (4.37-8.08) 6.68 (4.76-8.23) 7.00 (4.66-8.98) .28 Clinical diagnosis MS 1/11 (9) 11/82 (13)c 53/174 (30.5) .005 NA NA NA NA Relapsing non-MS disease 0 10/82 (12)c 1/174 (0.6) <.001 0 10/71 (14) 1/121 (0.8) <.001 NMOSD 0 0 1/174 (0.6) >.99 0 0 1/121 (0.8) >.99 Monophasic ADS 10/11 (91) 61/82 (74) 119/174 (68.4) .40 10/10 (100) 61/71 (86) 119/121 (98.3) .001 Exposure to DMT 1/11 (9) 2/82 (2) 42/174 (24.1) <.001 0 0 0 >.99 Clinical relapses 0 16/82 (20) 39/174 (22.4) .72 0 9/71 (13) 2/121 (1.7) .002 Time to second attack, median (IQR), y NA 0.92 (0.31-1.60) 0.83 (0.41-1.69) .47 NA 1.43 (0.52-2.12) 0.81 (0.36-1.25) .28 Relapse ON only NA 10/16 (63) 1/39 (3) <.001 NA 7/9 (78) 0 .11 TM only NA 1/16 (6) 2/39 (5) >.99 NA 0 1/2 (50) .18 ON or TM with other NA 5/16 (31) 14/39 (36) .40 NA 2/9 (22) 1/2 (50) .49 Other NA 0 22/39 (56) <.001 NA 0 0 >.99 ARR in the first 4 y, median (IQR)d NA 0.50 (0.50-0.75) 0.75 (0.75-1.00) .18 NA 0.50 (0.50-0.75) 0.75 (0.63-0.88) .40 EDSS score at 4 y, median (IQR)e 0 (0-1.50) 0 (0-1.00) 1.00 (0-1.50) .049 0 (0-0.38) 0 (0-1.00) 0 (0-1.50) .02 Visual functional system score at 4 y, median (IQR)f 0 (0-1.00) 0 (0-1.00) 0 (0-0) .04 0 (0-0.25) 0 (0-0) 0 (0-0) .18 MRI follow-up Duration of MRI follow-up, median (IQR), y 5.03 (1.50-5.54) 4.04 (1.11-5.98) 3.43 (1.04-5.98) .34 4.54 (1.32-5.13) 3.98 (1.10-5.93) 3.01 (1.03-5.97) .37 No.

0.38) 0 (0-1.00) 0 (0-1.50) .02 Visual functional system score at 4 y, median (IQR)f 0 (0-1.00) 0 (0-1.00) 0 (0-0) .04 0 (0-0.25) 0 (0-0) 0 (0-0) .18 MRI follow-up Duration of MRI follow-up, median (IQR), y 5.03 (1.50-5.54) 4.04 (1.11-5.98) 3.43 (1.04-5.98) .34 4.54 (1.32-5.13) 3.98 (1.10-5.93) 3.01 (1.03-5.97) .37 No. of scans per participant, median (IQR) 5 (3-6) 5 (3-8) 5 (3-8) .28 5 (3-5) 5 (3-8) 5 (3-7) .38 Complete brain lesion resolution 5/8 (62) 26/49 (53) 20/109 (18.3) <.001 5/7 (71) 25/42 (60) 20/62 (32) .009 New MRI lesionsg 1/10 (10) 15/78 (19) 56/162 (34.6) .03 0 5/67 (7) 8/112 (7.1) .85 Lesion volume at last scan, median (IQR), cm3 Total T2 0.19 (0-0.38) 0 (0-0.03) 0.22 (0-3.41) .006 0 (0-0.19) 0 (0-0.01) 0 (0-0.12) .18 Total T1 0 (0-0.01) 0 (0-0) 0 (0-1.01) .01 0 (0-0) 0 (0-0) 0 (0-0) .17 Abbreviations: ADS, acquired demyelinating syndrome; ARR, annualized relapse rate; DMT, disease-modifying treatment; EDSS, Expanded Disability Status Scale; IQR, interquartile range; MS, multiple sclerosis; NA, not applicable; NMOSD, neuromyelitis optica spectrum disorder; ON, optic neuritis; TM, transverse myelitis. a P values are computed based on comparisons between seropositive and seronegative participants. b Six participants with a diagnosis of monophasic ADS (2 seropositive and 4 seronegative patients) and 1 seronegative participant diagnosed as having MS had less than 6 months of clinical follow-up and are not included here.

a P values are computed based on comparisons between seropositive and seronegative participants. b Six participants with a diagnosis of monophasic ADS (2 seropositive and 4 seronegative patients) and 1 seronegative participant diagnosed as having MS had less than 6 months of clinical follow-up and are not included here. c Of the 11 seropositive patients with a diagnosis of MS, 7 experienced clinical relapses and 4 were diagnosed as having MS based on evidence of new lesions on MRI. Of the 10 seropositive participants with a diagnosis of relapsing non-MS disease, 9 experienced clinical relapses not meeting the current MS diagnostic criteria and 1 developed new, asymptomatic lesions 3 years after an ADEM presentation. d ARR is computed among participants with relapsing disease, including the presenting episode. e EDSS score at 4 years from presentation was evaluated in 65 seropositive and 123 seronegative participants. f Visual functional system score at 4 years from presentation was evaluated in 62 seropositive and 116 seronegative participants. g Serial brain MRI scans for more than 6 months from presentation were available in 78 seropositive and 162 seronegative participants.

e EDSS score at 4 years from presentation was evaluated in 65 seropositive and 123 seronegative participants. f Visual functional system score at 4 years from presentation was evaluated in 62 seropositive and 116 seronegative participants. g Serial brain MRI scans for more than 6 months from presentation were available in 78 seropositive and 162 seronegative participants. Serial brain MRI scans revealed that participants who were positive for anti-MOG antibodies at presentation were more likely to exhibit either complete resolution of all baseline lesions or smaller T2 and T1 total lesion volumes on their last MRI (Table 2). Among the presenting features of anti-MOG antibody–positive participants, older age was identified by the random forest analysis as the feature that was most strongly associated with a relapsing outcome, with a Gini impurity-derived importance 3-fold and 4-fold greater than the following 2 features in the ranking (T1 hypointense lesions and lesion count, respectively) (eTable 3 in the Supplement).

age was identified by the random forest analysis as the feature that was most strongly associated with a relapsing outcome, with a Gini impurity-derived importance 3-fold and 4-fold greater than the following 2 features in the ranking (T1 hypointense lesions and lesion count, respectively) (eTable 3 in the Supplement). Clinical Relapses as a Function of Serial MOG Status Of the 16 participants who were positive for anti-MOG antibodies at presentation who experienced clinical relapses, 9 (56%) were persistently seropositive (accounting for 28% of all persistently seropositive participants), 2 (13%) had fluctuating serologic status, and 5 (31%) converted to seronegative status (Figure 3A) (eFigure 2 in the Supplement). Of the latter, 4 participants experienced at least 1 relapse after detection of the first seronegative sample without being exposed to disease-modifying therapies prior to or at the time of the relapse.

fluctuating serologic status, and 5 (31%) converted to seronegative status (Figure 3A) (eFigure 2 in the Supplement). Of the latter, 4 participants experienced at least 1 relapse after detection of the first seronegative sample without being exposed to disease-modifying therapies prior to or at the time of the relapse. Figure 3. Evolution of Serologic Status and Clinical Relapses in Participants Who Were Positive for Anti–Myelin Oligodendrocyte Glycoprotein (MOG) Antibodies at Time of Initial Presentation A, Serologic status over time in 77 anti-MOG antibody–positive participants with serial samples. Each bar represents an individual participant, and the square at the end of each bar indicates the serological status in the last follow-up sample. Dark blue bars indicate seropositive status, and light blue bars indicate seronegative status. Black dots represent the time of sampling. Colored circles indicate clinical relapses. B, Representative examples of antibody titers in participants during follow-up. Graphs 1 and 2 show 2 cases in which transient antibody titer increases were associated with clinical relapses (arrowheads). Graphs 3 and 4 show patients with reappearance of MOG antibodies after initial seroconversion from seropositive to seronegative status, in one case associated with a clinical attack. Graphs 5 and 6 show examples of patients who experienced clinical relapses after conversion to seronegative status.

elapses (arrowheads). Graphs 3 and 4 show patients with reappearance of MOG antibodies after initial seroconversion from seropositive to seronegative status, in one case associated with a clinical attack. Graphs 5 and 6 show examples of patients who experienced clinical relapses after conversion to seronegative status. A monophasic course was observed in most participants with an ADEM presentation, including 3 of 5 (60%) who were persistently seropositive, 27 of 28 (96%) who were seronegative, and 17 of 18 (94%) who converted to seronegative status. The 1 patient who relapsed after seroconversion had multiple relapses. While our protocol did not include sampling at the time of relapse, all of the samples acquired within 3 months of a clinical relapse in participants who were positive for anti-MOG antibodies at presentation were also seropositive at collection. One participant who was seropositive at presentation who then became seronegative experienced a subsequent relapse at the time of a planned study visit; findings from serum analysis at that time demonstrated re-emergence of MOG antibodies (Figure 3).

-MOG antibodies at presentation were also seropositive at collection. One participant who was seropositive at presentation who then became seronegative experienced a subsequent relapse at the time of a planned study visit; findings from serum analysis at that time demonstrated re-emergence of MOG antibodies (Figure 3). Discussion In this prospective cohort of pediatric patients with incident ADS, we serially evaluated MOG serostatus from clinical onset, with rigorous longitudinal clinical and MRI assessments extending for up to 13 years. Several of our findings reinforce and extend prior reports, showing that MOG antibodies are common at onset in pediatric ADS, particularly among younger children and those who present with ADEM or ON. We also confirmed a bimodal distribution of clinical presentations and MRI features in anti-MOG antibody–positive children, with younger patients being more likely to present with ADEM and/or a large number of ill-defined MRI lesions and older children more often presenting with ON and no or few well-defined brain lesions.

lso confirmed a bimodal distribution of clinical presentations and MRI features in anti-MOG antibody–positive children, with younger patients being more likely to present with ADEM and/or a large number of ill-defined MRI lesions and older children more often presenting with ON and no or few well-defined brain lesions. A strength of our study was the procurement of serial samples at predefined intervals, enabling us to demonstrate that a negative result for MOG antibodies in proximity to the ADS presentation almost entirely excludes subsequent positive results over time. The absence of MOG antibodies at onset has key clinical implications. First, in anti-MOG antibody–negative children with isolated ON or TM (ie, with normal findings on brain MRI and no clinical features suggesting lesions in the brain) and in anti-MOG antibody–negative patients with ADEM, a monophasic outcome is highly likely. Second, anti-MOG antibody–negative patients with clinical, MRI, and cerebrospinal fluid findings meeting McDonald MS diagnostic criteria follow a typical MS disease course. The circumstance of MOG antibodies identified in children meeting MS diagnostic criteria requires specific comment. Careful review of all such patients in our cohort demonstrated a clinical and imaging disease evolution atypical of MS. As such, we agree with previous authors advocating that the presence of MOG antibodies “plead against MS diagnosis.”

es identified in children meeting MS diagnostic criteria requires specific comment. Careful review of all such patients in our cohort demonstrated a clinical and imaging disease evolution atypical of MS. As such, we agree with previous authors advocating that the presence of MOG antibodies “plead against MS diagnosis.” Most patients who were seropositive at time of initial presentation subsequently became seronegative (after a median of 1 year from onset) and experienced a monophasic clinical course. Although most relapses in patients who were initially seropositive occurred while they were seropositive, nearly 40% of these patients experienced 1 or more clinical relapses after their first seronegative result. Of these, 4 of 16 participants (25%) were persistently seronegative, while 2 of 16 (13%) had anti-MOG antibodies again detected in subsequent samples. It is notable that we did not collect samples at the time of relapse and, as such, we may have missed more dynamic changes in serostatus; however, when serum was obtained within 3 months before or following a clinical attack in participants with prior seropositive status, MOG antibodies were reliably detected. Defining the optimal frequency of sampling required to more formally assess the relationship between an acute relapse and continued presence or re-emergence of MOG antibodies will require frequent regular sampling as well as sampling at time of relapse. It should also be noted that 72% of persistently seropositive patients in our cohort have not experienced a second attack despite prolonged observation. Among all presenting features of the anti-MOG antibody–positive patients evaluated in our study, an older age at presentation was most strongly associated with a relapsing outcome.

o be noted that 72% of persistently seropositive patients in our cohort have not experienced a second attack despite prolonged observation. Among all presenting features of the anti-MOG antibody–positive patients evaluated in our study, an older age at presentation was most strongly associated with a relapsing outcome. In line with previous reports, most anti-MOG antibody–positive children in our cohort had a favorable outcome, with monophasic courses for more than 80% of patients, low EDSS scores at last follow-up, and either complete resolution or very low residual lesion volumes at last brain MRI scan. However, the extended long-term outcome for such patients remains to be determined. Based on our findings, we would advocate that MOG testing has relevance for all children with ADS. Negative testing at onset provides reassurance for families of a likely monophasic illness if their child has presented with isolated ON, TM, or ADEM. Furthermore, a positive MOG result in a pediatric patient is associated with a non-MS disease course.

advocate that MOG testing has relevance for all children with ADS. Negative testing at onset provides reassurance for families of a likely monophasic illness if their child has presented with isolated ON, TM, or ADEM. Furthermore, a positive MOG result in a pediatric patient is associated with a non-MS disease course. Serial serological surveillance is valuable in patients found to be seropositive at presentation, whereas it is likely to be of little clinical utility in patients identified as seronegative at presentation. Conversion to seronegative status was associated with lesser risk of subsequent relapse, although it does not entirely preclude it. In turn, persistent seropositivity is not necessarily associated with disease relapse. If serial testing following presentation is limited by resources, then a 12-month time point following initial presentation will identify more than half of the patients destined to become seronegative, including most patients with ADEM likely to convert.

eropositivity is not necessarily associated with disease relapse. If serial testing following presentation is limited by resources, then a 12-month time point following initial presentation will identify more than half of the patients destined to become seronegative, including most patients with ADEM likely to convert. Given that many of the children found to be positive for anti-MOG antibodies at presentation will remain monophasic, we would not advocate broad institution of long-term immunomodulatory therapy following their initial presentation. In contrast, anti-MOG antibody–positive patients who do relapse are typically offered treatment with immunoglobulins, anti-CD20, or more general immunosuppressive therapies. Further research is required to confirm whether relapsing MOG demyelination is a lifelong illness and to define the morbidity of seropositive-related relapses more fully. In particular, future studies should aim to quantify the effect of seropositive-related demyelination on age-expected brain growth and white matter maturation, both of which are negatively affected in MS, and on cognitive performance. All of this information is essential to better inform the risk-to-benefit ratio of long-term immunosuppression.

ies should aim to quantify the effect of seropositive-related demyelination on age-expected brain growth and white matter maturation, both of which are negatively affected in MS, and on cognitive performance. All of this information is essential to better inform the risk-to-benefit ratio of long-term immunosuppression. Limitations This study had limitations. The study was not designed to acquire serological samples at time of relapses and might have therefore not captured transient changes in serological status associated with reoccurrence of either clinical or subclinical disease activity. Future studies, with frequent acquisition of serological samples following a first clinical episode and at the time of subsequent attacks, are needed to fully clarify the dynamic changes in serostatus in patients initially seropositive for anti-MOG antibodies. Additionally, since this study was originally designed for the study of patients with possible ADSs, patients with encephalitis-like presentations, also described in association with anti-MOG antibodies, would not have been included in our cohort.

n serostatus in patients initially seropositive for anti-MOG antibodies. Additionally, since this study was originally designed for the study of patients with possible ADSs, patients with encephalitis-like presentations, also described in association with anti-MOG antibodies, would not have been included in our cohort. Conclusions Anti MOG-antibodies are common in children with ADSs and are transient in approximatively half of cases. Most anti-MOG antibody–positive children experience a monophasic disease, with clinical relapses occurring more commonly, although not exclusively, in children with persistent seropositivity. Overall, most anti-MOG antibody–positive children have a favorable outcome, and the presence of anti-MOG antibodies at the time of incident demyelination should not immediately prompt the initiation of long-term immunomodulatory therapy. Supplement. eTable 1. Baseline clinical, MRI, and laboratory features in patients with and without ADEM. eTable 2. Baseline clinical and MRI features of persistently seropositive patients, patients converting to seronegative status, or patients with fluctuating serological status. eTable 3. Ranking of baseline clinical, MRI, and laboratory features most strongly associated with relapses in seropositive children. eFigure 1. Sensitivity analysis for evolution of serological status. eFigure 2. Kaplan-Meier curve for risk of a second clinical attack in participants persistently seropositive vs those who converted to seronegative status. Click here for additional data file.

Introduction Alzheimer disease (AD) is a devastating neurodegenerative disorder neuropathologically characterized by abnormal tau accumulation in neurofibrillary tangles (NFTs) and the presence of extracellular amyloid-β plaque deposits. Postmortem studies of AD and more recent neuroimaging studies provide evidence that involvement of the nucleus basalis of Meynert (nbM) may be critical and early in the molecular cascade of events. The accumulation of NFTs in the nbM may precede entorhinal cortex and locus coeruleus involvement, making the nbM potentially one of the earliest sites where NFT accumulation occurs. We have previously shown 3 distinct regional patterns of corticolimbic NFT accumulation in subtypes of AD. Hippocampal sparing (HpSp) AD has a relatively spared hippocampus compared with the greater NFT accumulation in the cortex. Limbic predominant AD has a severely involved hippocampus compared with a relatively spared cortex. Lying between these 2 extreme AD subtypes, typical AD has the expected hippocampal and cortical NFT accumulation based on the widely accepted staging scheme proposed by Braak and Braak. In addition to the neuropathologic distinctions, these AD subtypes have striking differences in demographics and clinical progression.

Lying between these 2 extreme AD subtypes, typical AD has the expected hippocampal and cortical NFT accumulation based on the widely accepted staging scheme proposed by Braak and Braak. In addition to the neuropathologic distinctions, these AD subtypes have striking differences in demographics and clinical progression. Given the significance of the nbM for targeted cholinergic therapy, we sought to test the hypothesis that clinicopathologic heterogeneity of AD subtypes underlies the variability of NFT accumulation and neuronal loss in the nbM. Our primary goal was to investigate selective vulnerability of the cholinergic system in AD by examining the severity of NFT accumulation and neuronal loss in the nbM among AD subtypes. Our secondary goal was to evaluate whether any associations exist between NFT accumulation in the nbM and demographic or clinicopathologic changes among AD subtypes.

gate selective vulnerability of the cholinergic system in AD by examining the severity of NFT accumulation and neuronal loss in the nbM among AD subtypes. Our secondary goal was to evaluate whether any associations exist between NFT accumulation in the nbM and demographic or clinicopathologic changes among AD subtypes. Methods Study Samples The Florida Autopsied Multi-Ethnic (FLAME) cohort, which had been accessioned from 1991 to 2015, is derived from a consecutive series of patients who elected to participate in a deeded autopsy program via memory disorder clinic referral services. These services included community-based educational seminars for caregivers of patients with dementia and Alzheimer Association educational support groups. The FLAME cohort, which comprised 2809 individuals (1436 [51%] males and 1373 [49%] females), with an age at death ranging between 36 and 104 years, was queried for neuropathologically diagnosed AD cases and normal controls who were nondemented. We excluded 1084 study brains that were not neuropathologically diagnosed as having AD; 124 AD cases with hippocampal sclerosis because it interfered with subtype classification; 101 AD cases lacking NFT data because these could not be subtyped; and 18 AD cases with known genetic mutations. A total of 1361 AD cases remained and were termed the FLAME-AD cohort. Of the remaining 121 controls who were nondemented (Braak tangle stage<IV that lacked significant neurodegenerative pathology) that we identified for the demographic and clinicopathologic comparisons given in Table 1, we excluded 18 that were younger than the youngest FLAME-AD case (<54 years). Thus, our final sample size for controls was 103. Of the 1464 AD cases and controls, 113 were excluded from NFT density analyses and 390 were excluded from neuronal density analyses. All brains were acquired with appropriate ethical approval, and the research performed on postmortem samples was approved by the Mayo Clinic Research Executive Committee.

ontrols was 103. Of the 1464 AD cases and controls, 113 were excluded from NFT density analyses and 390 were excluded from neuronal density analyses. All brains were acquired with appropriate ethical approval, and the research performed on postmortem samples was approved by the Mayo Clinic Research Executive Committee. Table 1. Demographic and Clinicopathologic Characteristics by Control and AD Subtype Characteristic Median (IQR) AD-Specific P Valuea Normal Controls (n = 103) AD Neuropathologic Subtype (n = 1361) HpSp (n = 175) Typical (n = 1014) Limbic Predominant (n = 172) Female, % total of AD type, No./total No. (%) 47/103 (46) 62/175 (35) 545/1014 (54) 121/172 (70) <.001 Educational level, y 16 (14 to 16) 16 (12 to 16) 14 (12 to 16) 14 (12 to 16) .007 APOE ε4, No./total No. (%) 8/21 (38) 64/140 (46) 488/767 (64) 93/129 (72) <.001 Clinical findings Age at onset, y NA 65 (56 to 72) 71 (65 to 77) 78 (72 to 81) <.001 Disease duration, y NA 9 (7 to 10) 9 (6 to 12) 9 (7 to 12) .16 Atypical presentation, No./total No. (%) NA 57/150 (38) 89/819 (11) 3/139 (2) <.001 MMSE Final score, points 27 (27 to 28) 7 (5 to 15) 13 (7 to 19) 18 (8 to 21) .01 Change in MMSE, points, yb 0 (0 to 0) –4 (–4 to –3) –2 (–2 to –1) –1 (–2 to –1) <.001 Postmortem findings Age at death, y 73 (60 to 80) 72 (66 to 80) 81 (76 to 86) 86 (82 to 90) <.001 Brain weight, g 1240 (1123 to 1338) 1042 (960 to 1145) 1040 (940 to 1140) 1040 (950 to 1120) .40 Braak tangle stage I (0 to III) VI (V to VI) VI (V to VI) VI (V to VI) <.001 Thal amyloid phase 0 (0 to 2) 5 (5 to 5) 5 (5 to 5) 5 (5 to 5) .67 Lewy body disease, No./total No. (%) 0/103 (0) 25/175 (14) 265/1014 (26) 44/172 (26) .003 nbM NFT density, per 0.125 mm2 1 (0 to 1) 14 (9 to 20) 10 (5 to 16) 8 (5 to 11) <.001 Neuronal density, per mm2 34 (30 to 39) 22 (17 to 28) 25 (19 to 30) 26 (19 to 32) .002 Abbreviations: AD, Alzheimer disease; APOE ε4, the ε4 allele of the apolipoprotein E gene; HpSp, hippocampal sparing; IQR, interquartile range; MMSE, Mini-Mental State Examination; nbM, nucleus basalis of Meynert; NA, not applicable; NFT, neurofibrillary tangle.

(30 to 39) 22 (17 to 28) 25 (19 to 30) 26 (19 to 32) .002 Abbreviations: AD, Alzheimer disease; APOE ε4, the ε4 allele of the apolipoprotein E gene; HpSp, hippocampal sparing; IQR, interquartile range; MMSE, Mini-Mental State Examination; nbM, nucleus basalis of Meynert; NA, not applicable; NFT, neurofibrillary tangle. a Normal controls were not included in Kruskal-Wallis rank sum test; thus, P values specifically reflect groupwise comparisons. b Estimated using a mixed linear regression model accounting for interaction of time from test to death and AD subtype.

(30 to 39) 22 (17 to 28) 25 (19 to 30) 26 (19 to 32) .002 Abbreviations: AD, Alzheimer disease; APOE ε4, the ε4 allele of the apolipoprotein E gene; HpSp, hippocampal sparing; IQR, interquartile range; MMSE, Mini-Mental State Examination; nbM, nucleus basalis of Meynert; NA, not applicable; NFT, neurofibrillary tangle. a Normal controls were not included in Kruskal-Wallis rank sum test; thus, P values specifically reflect groupwise comparisons. b Estimated using a mixed linear regression model accounting for interaction of time from test to death and AD subtype. Neuropathologic Procedures Standardized neuropathologic examination was performed by a single board-certified neuropathologist (D.W.D.) using the Dickson sampling scheme for neurodegenerative-centric brain dissection. To optimize sampling of the nbM at the time of brain cutting, the fixed hemibrain was cut into coronal slabs using 3 points to define the plane of section: the anterior commissure, infundibulum, and uncus. Formalin-fixed, paraffin-embedded tissue sections were cut to be 5 μm thick and mounted on to glass slides. An nbM tissue section was stained with thioflavin S and another with hematoxylin-eosin. The topographic distributions of both NFTs and amyloid-β plaques were assessed using thioflavin S immunofluorescence with an Olympus BH2 fluorescence microscope to assign Braak tangle stage and Thal amyloid phase. The Braak tangle stage ranged from 0 to III for controls and from IV to VI for AD cases. The NFT density is reported as counts per 0.125-mm2 microscopic field (×40 objective). Corticolimbic patterns of NFTs were examined using an AD subtype algorithm (eFigure 1 in the Supplement), which assigns an AD subtype of HpSp (175 [13%]), typical (1014 [74%]), or limbic predominant (172 [13%]). The algorithm specifically assesses the association between the hippocampus (Cornu Ammonis 1, more commonly known as CA1, and the subiculum) and the association cortices (frontal, parietal, and temporal). The α-synuclein antibody, nonamyloid-β protein component of AD amyloid, was used to assess the distribution of Lewy body pathology and classify as Lewy body disease (1:3000 dilution, rabbit, amino acids 98-115, with a cysteine residue at its C-terminus).

) and the association cortices (frontal, parietal, and temporal). The α-synuclein antibody, nonamyloid-β protein component of AD amyloid, was used to assess the distribution of Lewy body pathology and classify as Lewy body disease (1:3000 dilution, rabbit, amino acids 98-115, with a cysteine residue at its C-terminus). Neuropathologic Assessment of the nbM: NFT and Neuronal Loss Quantification The term nucleus basalis includes all neuronal components of the nbM, of which more than 90% are magnocellular neurons that are cholinergic. At the time of neuropathologic examination, thioflavin S microscopy was used to quantify NFT counts in the nbM (Figure 1). The prospective assessment of NFT density (NFT count per 0.125 mm2) was performed by D.W.D., who was blinded to AD subtype algorithm classification. An Olympus BH2 fluorescence microscope was used to evaluate greatest lesion density at low magnification. Subsequently, a ×40 objective was used for 2 or more microscopic fields to count the area of greatest density. A detailed overview of sample size by neuroanatomic level, as well as information regarding cases excluded, is given in eTables 1 and 2 in the Supplement.

evaluate greatest lesion density at low magnification. Subsequently, a ×40 objective was used for 2 or more microscopic fields to count the area of greatest density. A detailed overview of sample size by neuroanatomic level, as well as information regarding cases excluded, is given in eTables 1 and 2 in the Supplement. Figure 1. Selective Vulnerability of the Nucleus Basalis of Meynert (nbM) and Corticolimbic Structures to Neurofibrillary Tangles (NFTs) Among Neuropathologic Subtypes of Alzheimer Disease (AD) A. Thioflavin S microscopy (a, c, and e) shows greater NFT accumulation in the nbM of hippocampal sparing (HpSp) AD (a) compared with typical AD (c) and limbic predominant (limbic) AD (e). Hematoxylin-eosin–stained sections of the nbM (b, d, and f) were digitally quantified (b′, d′, and f′, respectively). Fewer neurons are observed in HpSp AD (b) compared with typical AD (d) and limbic predominant AD (f). Scale bar represents 50 μm. B. Heatmap of differences calculated between brain region of interest and the nbM, as exampled by the more severe involvement of the entorhinal cortex compared with the nbM in limbic predominant AD, shown in warmer colors, and the less severe involvement of the hippocampus (Hipp) in HpSp AD compared with the nbM, shown in cooler colors. C. We hypothesize that, although both the nbM and entorhinal cortex (ctx) are involved early among AD subtypes and across aging, the cortex may be more vulnerable in HpSp AD. By contrast, the pattern of greater vulnerability of limbic structures is manifested in both limbic predominant AD and perhaps as a function of older age. LOAD indicates late-onset AD; YOAD, young-onset AD.

cortex (ctx) are involved early among AD subtypes and across aging, the cortex may be more vulnerable in HpSp AD. By contrast, the pattern of greater vulnerability of limbic structures is manifested in both limbic predominant AD and perhaps as a function of older age. LOAD indicates late-onset AD; YOAD, young-onset AD. To quantify the neuronal density of the nbM among a large series of AD cases and controls, we implemented high-throughput digitization of hematoxylin-eosin–stained slides (Figure 1). Detailed digital pathology methods and neuroanatomic assessment can be found in eAppendix 1 in the Supplement. The level of the nbM was neuroanatomically classified based on the anterior-to-posterior extent of the nucleus (eFigure 2 in the Supplement). Given the lack of discrete boundaries of the nbM with neighboring cell groups, we implemented specific neuroanatomic boundaries that enabled us to systematically capture the neuronal density of nbM neurons. Our application of neuroanatomic boundaries and assessment of the anterior-to-posterior extent of the nbM was informed by Mesulam and Geula and the recently revisited anatomic assessment of the nbM by Liu et al. We used the anterior commissure, globus pallidus, fornix, and mammillary body to facilitate identification of nbM level. The slides with hematoxylin-eosin–stained tissue were annotated, being blinded to both disease status (control vs AD) and AD subtype, using ImageScope software (Leica Biosystems). The annotated nbM was then batch analyzed in Aperio eSlide Manager (Leica Biosystems) using a custom-designed digital pathology macro to identify surviving neurons in the nbM on hematoxylin-eosin–stained sections. The macro was built to recognize the well-circumscribed, basophilic properties of the nbM neuron, as shown in Figure 1. The data are exported as counts, which were divided by the area annotated. Neuronal density is reported as neurons per millimeters squared, not total number of neurons. The mean size of objects counted was additionally exported to examine neuronal shrinkage (eAppendix 2 in the Supplement). Data provided in the present study were derived from the anterior nbM (eFigure 1 in the Supplement) because this level contains the most widespread and noticeable portion of the nbM and was the most robustly sampled (eTables 1 and 2 in the Supplement). As shown in eFigure 2 in the Supplement, the cholinergic neurons at the anterior level correspond to the Ch4am, Ch4al, and Ch4ai subsectors of the nbM.

the Supplement) because this level contains the most widespread and noticeable portion of the nbM and was the most robustly sampled (eTables 1 and 2 in the Supplement). As shown in eFigure 2 in the Supplement, the cholinergic neurons at the anterior level correspond to the Ch4am, Ch4al, and Ch4ai subsectors of the nbM. Clinical History In this cross-sectional study, clinical history was abstracted from existing clinical records made available by brain bank participants or family members, as previously described. The neurologic summary or brain bank questionnaire was reviewed for details relating to the age when the first cognitive symptoms began. The date of birth was subtracted from the approximate date at onset to identify the age at onset in years. The date at onset was subtracted from date of death to identify the disease duration in years. Atypical clinical presentations were recorded that differed from the expected amnestic presentation more commonly observed among patients having a clinical diagnosis of AD, such as primary progressive aphasia, frontotemporal dementia, posterior cortical atrophy, and corticobasal syndrome. Any available Mini-Mental State Examination (MMSE) score (range, 0-30) and date of test were recorded. The MMSE scores obtained within 3 years of death were recorded as the final MMSE score, serving as a measure of cognitive impairment. Three or more MMSE scores were required to estimate cognitive decline measured as a change in MMSE (eAppendix 1 in the Supplement). The first and last MMSE test dates were required to be more than 1 year apart.

within 3 years of death were recorded as the final MMSE score, serving as a measure of cognitive impairment. Three or more MMSE scores were required to estimate cognitive decline measured as a change in MMSE (eAppendix 1 in the Supplement). The first and last MMSE test dates were required to be more than 1 year apart. Statistical Analysis Continuous variables (eg, age at onset, nbM NFT density) are represented using medians and interquartile ranges. Categorical variables (eg, sex) were summarized using frequencies and percentages. The Kruskal-Wallis rank sum test was used to test for differences in continuous measures, while the Pearson χ2 test was used to compare proportions among AD subtype groups. The associations of demographic (sex, educational level, and the apolipoprotein E [APOE] ε4 carrier status) and clinical variables (age at onset, disease duration, and final MMSE score) with NFT accumulation in the nbM among AD subtypes was examined using 3 multivariable linear regression models that were created to investigate within-subtype differences. Three additional models were used to investigate neuronal density. All tests were 2-sided, and P < .05 was considered statistically significant. All statistical analyses were performed using R statistical software, version 3.4.2 (R Foundation for Statistical Computing) and completed January 2019.

n-subtype differences. Three additional models were used to investigate neuronal density. All tests were 2-sided, and P < .05 was considered statistically significant. All statistical analyses were performed using R statistical software, version 3.4.2 (R Foundation for Statistical Computing) and completed January 2019. Results Demographic, Clinicopathologic, and Neuropathologic Differences Among AD Subtypes Table 1 provides the demographic, clinical, and neuropathologic findings among 1361 AD subtypes and 103 nondemented controls included for comparison. The HpSp AD cases were more commonly observed in men (113 [65%]) compared with typical AD (469 [46%]) and limbic predominant AD (51 [30%]) (P < .001). Individuals with HpSp AD had a higher level of education (median, 16 years; interquartile range [IQR], 12-16 years; P = .007) and lowest frequency of the APOE ε4 risk variant (64 of 140 [46%]; P < .001). Among clinical findings, HpSp AD cases were the youngest to present with cognitive symptoms (median, 65 years; IQR, 56-72 years) compared with typical AD cases (median, 71 years; IQR, 65-77 years) and limbic predominant AD cases (median, 78 years; IQR, 72-81 years) (P < .001). The proportion of cases with an atypical clinical presentation was highest in HpSp AD cases (57 of 150 [38%]), lower in typical AD (89 of 819 [11%]), and lowest in limbic predominant AD (3 of 139 [2%]) (P < .001). The final median (IQR) MMSE score was lowest in HpSp AD (7; 5-15 points), higher in typical AD (13; 7-19 points), and highest in limbic predominant AD (18; 8-21 points) (P = .01). Moreover, the change in median MMSE over time was faster in HpSp AD cases (4 points lost per year; IQR, −4 to −3 points) compared with both typical AD (2 points lost per year; IQR, −2 to −1 points) and limbic predominant AD (1 point lost per year; IQR, −2 to −1 points) (P < .001). Using linear regression modeling to adjust for age at death and sex, AD subtype differences remained significant, with HpSp AD cases estimated to decline the fastest at 5 points lost per year (IQR, −6 to −4 points; P < .001).

−1 points) and limbic predominant AD (1 point lost per year; IQR, −2 to −1 points) (P < .001). Using linear regression modeling to adjust for age at death and sex, AD subtype differences remained significant, with HpSp AD cases estimated to decline the fastest at 5 points lost per year (IQR, −6 to −4 points; P < .001). In addition to observed demographic and clinical differences, there were major differences in neuropathologic characteristics among AD subtypes (Table 1). The HpSp AD cases were the youngest at death (median, 72 years; IQR, 66-80 years) compared with typical AD (median, 81 years; IQR, 76-86 years), who were younger than limbic predominant AD (median, 86 years; IQR, 82-90 years) (P < .001). Braak tangle stage differed at the group level (P < .001); however, between-group differences were less evident because the median (IQR) stage was the same for each AD subtype (VI; V-VI). The presence of coexisting Lewy body disease was lower in HpSp AD (25 [14%]) compared with typical AD (265 [26%]) and limbic predominant AD (44 [26%]) (P = .003).

the group level (P < .001); however, between-group differences were less evident because the median (IQR) stage was the same for each AD subtype (VI; V-VI). The presence of coexisting Lewy body disease was lower in HpSp AD (25 [14%]) compared with typical AD (265 [26%]) and limbic predominant AD (44 [26%]) (P = .003). Demographic and Clinical Associations With NFT Accumulation and Neuronal Density in the nbM Stratified by AD Subtype The NFT accumulation (count per 0.125 mm2) in the anterior level of the nbM was highest in 163 HpSp AD cases (median, 14; IQR, 9-20), lower in 937 typical AD cases (median, 10; IQR, 5-16), and lowest in 163 limbic predominant AD (median 8; IQR, 5-11) (P < .001) (Figure 2A; Table 1). Neuronal density (per millimeter squared) in the nbM was lowest in 148 HpSp AD cases (median, 22; IQR, 17-28) compared with 727 typical AD (median, 25; IQR, 19-30) and 127 limbic predominant AD (median, 26; IQR, 19-32) (P = .002) (Figure 2B). To test that the difference in nbM NFT density and neuronal density did not simply reflect differences observed in cognitive impairment among individuals with the various AD subtypes, we used multivariable linear regression models to adjust for final MMSE score. Group differences remained among AD subtypes for nbM NFT density (P < .001) and nbM neuronal density (P = .001). To put in perspective the relative difference in NFT accumulation in areas routinely assessed in Braak tangle staging compared with the nbM, Figure 1B shows a heatmap of calculated differences between brain regions of interest and the nbM. This is exampled by the more severe involvement of the hippocampus compared with the nbM in limbic predominant AD, as shown by warmer colors, and the less severe involvement of the hippocampus compared with the nbM in HpSp AD, as shown by cooler colors. The calculated differences are given in eFigure 3 in the Supplement.

is is exampled by the more severe involvement of the hippocampus compared with the nbM in limbic predominant AD, as shown by warmer colors, and the less severe involvement of the hippocampus compared with the nbM in HpSp AD, as shown by cooler colors. The calculated differences are given in eFigure 3 in the Supplement. Figure 2. Neurofibrillary Tangle (NFT) and Neuronal Density in the Nucleus Basalis of Meynert (nbM) Differs Among Alzheimer Disease (AD) Subtypes Data are displayed as jitter plots overlaying box plots of the 25th to 75th percentile, with the middle horizontal line representing the median. Nondemented normal controls are displayed for reference. Within each AD subtype, individuals presenting with cognitive problems younger than 65 years of age are displayed on the left and individuals 65 years or older are displayed on the right (lighter color). A, The NFT density in the nbM was measured using thioflavin S fluorescence microscopy. HpSp AD cases have the greatest accumulation of NFTs compared with typical AD, which is greater than limbic predominant (limbic) AD. B, Neuronal density was measured using a custom-designed digital pathology macro on hematoxylin-eosin–stained sections of the nbM. Hippocampal sparing (HpSp) AD has fewer remaining neurons compared with typical AD, which has fewer compared with limbic predominant AD. Pairwise comparisons were performed using Mann-Whitney rank sum test.

c predominant AD. For every 10 years’ younger age at onset, the number of neurons was expected to be lower by 4.6 (95% CI, 2.3-7.0) in limbic predominant AD cases (P < .001) (Figure 3B). In addition, limbic predominant cases were observed to have 4.3 fewer neurons (95% CI, 0.47-8.1) for every 10-point decrease in MMSE. Table 2. Regression Analyses of Demographic and Clinical Variables Estimating NFT Accumulation and Neuronal Density in the nbM Stratified by AD Subtype Variablea HpSp AD Typical AD Limbic Predominant AD β (95% CI) P Value β (95% CI) P Value β (95% CI) P Value Dependent variable: nbM NFT densityb Female sexc 0.91 (–1.7 to 3.6) .50 2.5 (1.4 to 3.5) <.001 1.1 (–1.0 to 3.2) .32 Educational level, 1 y –0.24 (–0.75 to 0.27) .35 –0.01 (–0.25 to 0.22) .91 –0.30 (–0.73 to 0.12) .16 APOE ε4c 1.3 (–1.3 to 3.9) .32 1.3 (0.15 to 2.5) .03 –0.33 (–2.7 to 2.1) .79 Age at onset, 10 y –1.5 (–2.9 to –0.15) .03 –3.2 (–3.9 to –2.4) <.001 0.61 (–1.1 to 2.3) .48 Disease duration, 1 y –0.14 (–0.59 to 0.32) .56 0.040 (–0.11 to 0.19) .57 0.090 (–0.21 to 0.39) .57 Last MMSE score, 10 points –0.89 (–4.1 to 2.3) .59 –1.8 (–3.2 to –0.31) .02 –0.050 (–2.9 to 2.8) .97 Dependent variable: nbM neuronal density Female sexc 1.1 (–1.6 to 3.8) .42 –1.1 (–2.3 to –0.010) .048 –1.2 (–4.2 to 1.8) .45 Educational level, 1 y 0.25 (–0.25 to 0.75) .33 0.10 (–0.15 to 0.34) .44 0.00 (–0.56 to 0.56) .99

t MMSE score, 10 points –0.89 (–4.1 to 2.3) .59 –1.8 (–3.2 to –0.31) .02 –0.050 (–2.9 to 2.8) .97 Dependent variable: nbM neuronal density Female sexc 1.1 (–1.6 to 3.8) .42 –1.1 (–2.3 to –0.010) .048 –1.2 (–4.2 to 1.8) .45 Educational level, 1 y 0.25 (–0.25 to 0.75) .33 0.10 (–0.15 to 0.34) .44 0.00 (–0.56 to 0.56) .99 APOE ε4c –0.26 (–2.9 to 2.3) .84 –0.13 (–1.4 to 1.1) .84 –2.0 (–5.3 to 1.2) .22 Age at onset, 10 y 1.5 (0.13 to 2.8) .03 4.0 (3.2 to 4.7) <.001 4.6 (2.3 to 7.0) <.001 Disease duration, 1 y –0.44 (–0.89 to 0.02) .06 0.11 (–0.040 to 0.26) .17 0.23 (–0.18 to 0.64) .27 Last MMSE score, 10 points –0.49 (–3.7 to 2.8) .77 1.4 (–0.030 to 2.9) .06 4.3 (0.47 to 8.1) .03 Abbreviations: AD, Alzheimer disease; APOE e4, the ε4 allele of the apolipoprotein E gene; HpSp, hippocampal sparing; MMSE, Mini-Mental State Examination; nbM, nucleus basalis of Meynert; NFT, neurofibrillary tangle. a The unit for each continuous variable is indicated alongside the variable. b Data are presented as the estimated change in the number of neurofibrillary tangles per unit change of the variable of interest in the 3 models. c Female sex and the presence of the ε4 risk allele are discrete variables set at 1 and can be interpreted as associating positively if the outcome is positive.

a The unit for each continuous variable is indicated alongside the variable. b Data are presented as the estimated change in the number of neurofibrillary tangles per unit change of the variable of interest in the 3 models. c Female sex and the presence of the ε4 risk allele are discrete variables set at 1 and can be interpreted as associating positively if the outcome is positive. Figure 3. Association Between Age at Onset of Cognitive Symptoms and Neurofibrillary Tangle (NFT) Density and Neuronal Density in the Nucleus Basalis of Meynert (nmB) Among Alzheimer Disease (AD) Subtypes Best-fit lines represent the association of age at onset per AD subtype while adjusting for other covariates in the regression models found in Table 2. A, Younger age at onset of cognitive symptoms is significantly associated with greater NFT accumulation measured in the nbM of hippocampal sparing (HpSp) AD and typical AD but not limbic predominant (limbic) AD. B, The association between age at onset of cognitive symptoms and neuronal density measured in hematoxylin-eosin–stained sections of the nbM among AD subtypes is significant among all of the AD subtypes.

mulation measured in the nbM of hippocampal sparing (HpSp) AD and typical AD but not limbic predominant (limbic) AD. B, The association between age at onset of cognitive symptoms and neuronal density measured in hematoxylin-eosin–stained sections of the nbM among AD subtypes is significant among all of the AD subtypes. Discussion In this retrospective study of patients who were derived from memory disorder clinics for diagnosis and treatment of neurocognitive disorders and who ultimately came to autopsy, our data support the novel concept that there is an association between the severity of NFT pathology in the nbM and corticolimbic patterns of NFT pathology in the brain. In the FLAME-AD cohort, the nbM was more vulnerable to neuropathologic insult in HpSp AD cases compared with typical AD, which were more vulnerable than limbic predominant AD and nondemented normal controls. Younger age at onset was associated with greater NFT accumulation in the nbM of HpSp AD and typical AD but not limbic predominant AD. However, we did observe fewer nbM neurons remaining in limbic predominant cases, suggesting perhaps a non-NFT mediated association.

limbic predominant AD and nondemented normal controls. Younger age at onset was associated with greater NFT accumulation in the nbM of HpSp AD and typical AD but not limbic predominant AD. However, we did observe fewer nbM neurons remaining in limbic predominant cases, suggesting perhaps a non-NFT mediated association. The nbM is one of the most vulnerable brain regions to NFT pathology in AD. On the basis of the importance of the nbM for targeted treatment by acetylcholinesterase inhibitor therapies, we sought to test the hypothesis that the cholinergic system is differentially involved among AD subtypes. Inherent to autopsy studies, investigations are often performed at the end stage of the disease. Evidence suggests that 80% to 88% of the cholinergic neurons in the posterior nbM are depleted in AD compared with 29% to 54% in the anterior nbM. Data from the current study were derived from the anterior nbM, which facilitates investigation of clinicopathologic contributors to variability in NFT accumulation and in nbM neuronal loss prior to the burnout observed in the posterior nbM of AD brains. We found twice the number of NFTs per microscopic field in the nbM of HpSp AD cases compared with limbic predominant AD cases. This finding emphasizes the association of the NFT pathology in the anterior nbM to cortical, rather than limbic, location. Evidence supports that cholinergic projections to the hippocampus are derived from cholinergic neurons in the medial septal nucleus (Ch1) and vertical nucleus of the diagonal band (Ch2) of the basal forebrain. However, cell loss is not particularly evident in these hippocampal projecting nuclei in comparison with cell shrinkage of cholinergic neurons. Retrograde transport of neurotrophic factors from target regions is hypothesized to account for the cell shrinkage along the rostrocaudal extent of basal forebrain structures. This process elicits an intriguing hypothesis in HpSp AD in which cortical vulnerability to NFTs is a result of diminished cholinergic innervation from the nbM, which, via a feedback loop, contributes to exacerbation of nbM vulnerability to NFTs through a reduction of neurotrophic support from the cortex.

n structures. This process elicits an intriguing hypothesis in HpSp AD in which cortical vulnerability to NFTs is a result of diminished cholinergic innervation from the nbM, which, via a feedback loop, contributes to exacerbation of nbM vulnerability to NFTs through a reduction of neurotrophic support from the cortex. We found the fewest neurons in the nbM of HpSp AD cases compared with typical AD and limbic predominant AD, which were all lower than the nbM neuronal counts of controls. As expected, an inverse association was observed between NFT accumulation and neuronal density in the nbM. There is evidence to suggest that neuronal loss in the nbM precedes that in the locus coeruleus or entorhinal cortex, 2 areas that are hypothesized to be initiation sites for NFT accumulation. Although we could not compare the rate of NFT accumulation in postmortem tissue, we provided data on differential patterns in the entorhinal cortex, hippocampus, and association cortices relative to the nbM among AD subtypes. These observations are especially interesting when taken together with previous findings of more severe NFT accumulation in the cortex of younger-onset AD cases compared with potentially greater vulnerability of the hippocampus in late-onset AD cases. We hypothesize that upstream factors, likely genetic, contribute both to the corticolimbic pattern of NFT vulnerability among AD subtypes and between young-onset AD and late-onset AD. We observed a wave of vulnerability in which the exacerbation of nbM NFTs in HpSp AD may leave the cortex more vulnerable to NFT accumulation, perhaps via a biologically accelerated process or through a mechanism of disinhibition. By contrast, the limbic predominant AD cases had an exacerbation of areas vulnerable early in the Braak-like pattern of NFT accumulation, perhaps via a biologically restrictive process that relatively confines pathology to limbic areas.

aps via a biologically accelerated process or through a mechanism of disinhibition. By contrast, the limbic predominant AD cases had an exacerbation of areas vulnerable early in the Braak-like pattern of NFT accumulation, perhaps via a biologically restrictive process that relatively confines pathology to limbic areas. Given the significant demographic and clinicopathologic differences among AD subtypes, we elected to perform an analysis with separate covariate effects by AD subtype to enhance our ability to detect meaningful associations not diluted by the contribution of AD as a whole. We observed greater NFT accumulation associated with younger age at onset in HpSp AD and typical AD but not limbic predominant AD. Our data support a neurochemical study that has shown more severe cholinergic deficits in the brains of younger decedents (died <79 years of age) compared with older decedents (≥80 years of age), which suggests NFT accumulation in the nbM may underlie more widespread pathology and likely more severe cholinergic deficits. This outcome is of particular interest given that age is the strongest risk factor for AD dementia, yet we and others have shown that individuals with young-onset AD may paradoxically represent a more aggressive form of the disease. Given the focused association of age at onset in HpSp AD cases to the exclusion of other demographic and clinical variables and lack of a similar finding in the limbic predominant AD cases, future studies may seek to identify as-of-yet unknown contributors that may be unique in each of these extreme AD subtypes. It will be of particular interest to examine non–APOE ε4-associated genetic mechanisms that affect gene transcription, which have been shown to influence the temporal course of nbM involvement in AD.

ure studies may seek to identify as-of-yet unknown contributors that may be unique in each of these extreme AD subtypes. It will be of particular interest to examine non–APOE ε4-associated genetic mechanisms that affect gene transcription, which have been shown to influence the temporal course of nbM involvement in AD. We observed significant differences in NFT accumulation in the nbM among APOE ε4 carriers and noncarriers in typical AD, which is consistent with previous findings showing subtle differences in response to cholinesterase inhibitors based on APOE ε4 carrier status. However, this finding is inconsistent with a previous immunohistochemical study that did not detect differences between APOE ε4 carriers and noncarriers using the Alz-50 antibody. This inconsistency could be the result of differences in detection methods of NFT pathology, or the finding may be specific to typical AD cases and not readily detected in a cohort also including the extreme subtypes, that is, HpSp AD and limbic predominant AD. Our study provides additional evidence that female sex was associated with selective vulnerability of the nbM to NFTs in typical AD. These findings are consistent with those reported in previous studies showing that women have greater amounts of global AD pathology, especially greater numbers of NFTs than men. Moreover, our results support the immunohistochemical study of the nbM that has shown greater tau immunoreactivity in women than in men.

AD. These findings are consistent with those reported in previous studies showing that women have greater amounts of global AD pathology, especially greater numbers of NFTs than men. Moreover, our results support the immunohistochemical study of the nbM that has shown greater tau immunoreactivity in women than in men. Limitations The results of the current study should be informative for clinicians diagnosing memory disorders and treating patients with such disorders. However, it is important to acknowledge various biases associated with consenting to autopsy, which can include self-selection, higher educational level, marital status, and race/ethnicity. Moreover, longitudinal data relating to cognitive decline measured using MMSE was limited to a subset of cases. Given the size of our autopsied AD cohort, we did not use stereologic methods to assess the volume of the nbM. To offset this limitation in our study design, we used digital pathology measures to objectively quantify neuronal density of the anterior nbM with neuroanatomic boundaries defined by neighboring structures. Conclusions The findings reported in the present study, as well as those reported by others, underscore the importance of considering age at onset, sex, and APOE genotype when interpreting outcomes in AD. Our future studies will expand on these findings in a longitudinally followed cohort to extend our understanding of the differential treatments administered to individuals with HpSp AD, typical AD, and limbic predominant AD. Supplement. eAppendix 1. Methods eAppendix 2. Results

Conclusions The findings reported in the present study, as well as those reported by others, underscore the importance of considering age at onset, sex, and APOE genotype when interpreting outcomes in AD. Our future studies will expand on these findings in a longitudinally followed cohort to extend our understanding of the differential treatments administered to individuals with HpSp AD, typical AD, and limbic predominant AD. Supplement. eAppendix 1. Methods eAppendix 2. Results eFigure 1. Alzheimer Disease’s (AD) Subtyping Algorithm Classification eFigure 2. The Anterior-to-Posterior Extent of the Nucleus Basalis of Meynert eFigure 3. Relative Difference in NFT Accumulation in the nbM Compared to Areas Involved Early (eg, Entorhinal) and Later (eg, Association Cortex) in the Disease Progression According to Braak Tangle Staging eTable 1. Sample Size Inclusion and Exclusion From Analyses Investigating Neurofibrillary Tangle Differences in the Nucleus Basalis of Meynert Within the Total Cohort of Normal Controls and Alzheimer’s Disease Cases or by AD Subtype eTable 2. Sample Size Inclusion and Exclusion From Analyses Investigating Neuronal Differences in the Nucleus Basalis of Meynert Within the Total Cohort of Normal Controls and Alzheimer’s Disease Cases or by AD Subtype eReferences Click here for additional data file.

Introduction Peripheral nerve hyperexcitability syndromes include neuromyotonia or Isaacs syndrome and can be genetic or acquired. Acquired neuromyotonia (NMT) was first associated with antibodies that immunoprecipitated voltage-gated potassium channels (VGKCs).1,2,3 Voltage-gated potassium channel antibodies were then identified in patients with the rare Morvan syndrome4 and in a form of limbic encephalitis.5,6,7 Subsequently, it was shown that these antibodies were not directed against the extracellular domains of VGKCs but to 3 proteins tightly complexed with the VGKCs in detergent extracts of mammalian brain tissue: leucine-rich glioma inactivated protein 1 (LGI1), contactin-associated protein 2 (CASPR2), and contactin 2.8 Antibodies to LGI1 were strongly associated with limbic encephalitis, whereas antibodies to CASPR2 were found more often in patients with Morvan syndrome or NMT,8,9,10,11 sometimes with LGI1.8,11 Contactin-associated protein 2 antibodies were associated with underlying thymomas,12 but contactin 2 antibodies were uncommon.8 The reported frequency of VGKC-complex antibodies in patients with NMT, usually at relatively low titers (100-400pM), or in those with the less severe variant, cramp fasciculation syndrome, has varied between 2% and 40%.13,14,15,16 Other studies have included only a few patients with NMT.8,10,17 The full spectrum of clinical features (including pain, which has not been well studied previously), the specific antigenic targets, and whether the clinical features relate to antibody specificity have not been determined in NMT.

n 2% and 40%.13,14,15,16 Other studies have included only a few patients with NMT.8,10,17 The full spectrum of clinical features (including pain, which has not been well studied previously), the specific antigenic targets, and whether the clinical features relate to antibody specificity have not been determined in NMT. We describe detailed clinical and serologic characteristics of patients with NMT from Japan and Australia combined with the results of a novel, independent, patient-led pain questionnaire sent to individuals registered on an online forum for persons with Isaacs syndrome. The study is particularly timely given recent evidence that injection of patient-derived CASPR2 antibodies or genetic deletion of CASPR2 causes afferent nerve hyperexcitability and mechanical allodynia in mice.18

-led pain questionnaire sent to individuals registered on an online forum for persons with Isaacs syndrome. The study is particularly timely given recent evidence that injection of patient-derived CASPR2 antibodies or genetic deletion of CASPR2 causes afferent nerve hyperexcitability and mechanical allodynia in mice.18 Methods Thirty-eight serum samples from patients with clinical and neurophysiologic features consistent with a diagnosis of NMT were collected in Sydney, Australia,19 after VGKC-complex antibody screening, and Kagoshima, Japan, between February 2007 and August 2009 for routine testing at the time of patient review and were studied in detail in 2012. Final data analysis was performed in 2016. The diagnosis was reached according to established criteria, with symptoms or signs of muscle twitching or muscle cramps affecting at least 2 regions of skeletal muscles.14 All patients demonstrated the characteristic electromyographic (EMG) discharges consisting of doublet, triplet, or multiplet single–motor unit discharges with a high intraburst frequency of between 40 and 400 per second.3,20,21 Eleven patients with similar symptoms and serologic results, but without confirmation or testing by EMG, were excluded. Ethical approval was granted by the South Eastern Sydney Local Health District (Sydney, Australia), the University of Sydney (Sydney), and the University of Kagoshima (Kagoshima, Japan) human research ethics committees, and patients gave written informed consent for their data to be used. The serologic study was performed under the authority of the Oxford Regional Ethics Committee.

ealth District (Sydney, Australia), the University of Sydney (Sydney), and the University of Kagoshima (Kagoshima, Japan) human research ethics committees, and patients gave written informed consent for their data to be used. The serologic study was performed under the authority of the Oxford Regional Ethics Committee. Antibody Tests Serum samples were originally tested for binding to VGKC-complex antibodies by radioimmunoprecipitation, as previously described.3,8 Antibodies to LGI1, CASPR2, and contactin 2 were detected by live cell–based assays. For these assays, DNA encoding the different proteins was transfected into human embryonic kidney 293 cells cultured on glass coverslips and left overnight at 37°C. The cells were gently washed and left in medium for 1 additional day before each coverslip was placed in a 20-well microtiter plate and incubated in 200 μL of medium containing 10 μL (1:20 for LGI1 or contactin 2) or 2 μL (1:100 for CASPR2) for 2 hours. Binding of IgG to the cells was detected with Alexa fluor anti-human IgG (Thermo Fisher Scientific) after fixation.8 The binding was scored visually on a scale from 0 (negative), 1 (low positive), and 2 to 4 (increasing strength of binding).

g 10 μL (1:20 for LGI1 or contactin 2) or 2 μL (1:100 for CASPR2) for 2 hours. Binding of IgG to the cells was detected with Alexa fluor anti-human IgG (Thermo Fisher Scientific) after fixation.8 The binding was scored visually on a scale from 0 (negative), 1 (low positive), and 2 to 4 (increasing strength of binding). Clinical Information and Pain Questionnaire The clinical data recorded (shown in Table 1; data requested are shown in eTable 1 in the Supplement) covered symptom history and symptoms at first clinic visit and at neurologic review, with specific reference to pain, autonomic symptoms, central nervous system features, tumors, investigations (all patients underwent EMG; magnetic resonance imaging, electroencephalogram, or cerebrospinal fluid abnormalities were documented if available), and modified Rankin Scale (mRS) scores at first review and follow-up, which was 2 to 4 years after review. The mRS measures disability on a range of 0 to 6, with 0 indicating normal and 6 indicating death.

netic resonance imaging, electroencephalogram, or cerebrospinal fluid abnormalities were documented if available), and modified Rankin Scale (mRS) scores at first review and follow-up, which was 2 to 4 years after review. The mRS measures disability on a range of 0 to 6, with 0 indicating normal and 6 indicating death. Table 1. Presenting Features and Neurophysiologic Findings in 38 Patients With Neuromyotonia Feature No. (%) Total (N = 38) Male (n = 25) Female (n = 13) Age, median (range), y 55 (12-85) 55 (12-85) 44 (25-79) EMG features defining NMT Multiplets 19 (50) 14 (56) 5 (38) Fasciculations 10 (26) 6 (24) 4 (31) Doublets 8 (21) 6 (24) 2 (15) Myokymic discharges 7 (18) 4 (16) 3 (23) Neuromyotonic discharges 7 (18) 3 (12) 4 (31) Bursts 6 (16) 5 (20) 1 (8) Repetitive discharges 3 (8) 0 3 (23) Symptoms reported on referral Cramps 16 (42) 13 (52) 3 (23) Fasciculations 7 (18) 5 (20) 2 (15) Myokymia 4 (11) 1 (4) 3 (23) Spasms 3 (8) 3 (12) 0 Stiffness 4 (11) 2 (8) 2 (15) Fatigue 1 (3) 0 1 (8) Paresthesia or pain 3 (8) 2 (8) 1 (8) Altered sensation 1 (3) 0 1 (8) Features in response to specific questions Cramps 32 (84) 22 (88) 10 (77) Twitching 30 (79) 20 (80) 10 (77) Sweating 12 (32) 9 (36) 3 (23) Weakness 13 (34) 10 (40) 3 (23) Stiffness 12 (32) 9 (36) 3 (23) Pseudomyotonia 7 (18) 4 (16) 3 (23) Autonomic disturbance Any 18 (47) 12 (48) 6 (46) Constipation 4 (11) 3 (12) 1 (8) Diarrhea 3 (8) 3 (12) 0 Excessive secretions, including sweating 8 (21) 4 (16) 4 (31) Tachycardia or tachypnea 4 (11) 2 (8) 3 (23) Other (1 each of dry mouth, hypothermia, hyperthermia, gastrointestinal, and erectile dysfunction) 5 (13) 4 (16) 1 (8) Sensory features Pain or paresthesia or both 20 (53) 14 (56) 6 (46) CNS features Any 11 (29) 9 (36) 2 (15) Neuropsychiatric features of agitation or anxiety 10 (26) 8 (32) 2 (15) Insomnia 10 (26) 8 (32) 2 (15) Other sleep disturbance 3 (8) 2 (8) 1 (8) Depression 5 (13) 4 (16) 1 (8) Cognitive problems 1 (3) 0 1 (8) Seizures 2 (5) 1 (Previous) 1 (8) Tumor types 8 Concurrent, 2 previous (26) 3 Thymomas, 3 prostates, 1 previous non–small cell lung cancer, 1 previous leukemia (32) 2 Thymomas (15) Abbreviations: CNS, central nervous system; EMG, electromyography; NMT, neuromyotonia.

5 (13) 4 (16) 1 (8) Cognitive problems 1 (3) 0 1 (8) Seizures 2 (5) 1 (Previous) 1 (8) Tumor types 8 Concurrent, 2 previous (26) 3 Thymomas, 3 prostates, 1 previous non–small cell lung cancer, 1 previous leukemia (32) 2 Thymomas (15) Abbreviations: CNS, central nervous system; EMG, electromyography; NMT, neuromyotonia. A patient-led, online pain questionnaire was initiated through a patient support network from April 2012 to May 2012, and individual patients were contacted by a patient representative (R.B.) through http://isaacsyndrome.proboards.com/ (now https://www.facebook.com/groups/isaacs.pnh/) and asked whether they had pain. One hundred seventy-six patients responded, of whom 165 reported pain and were sent the questionnaire (eTable 2 in the Supplement); full responses were obtained from 56 patients, deidentified by one of us (R.B.), and analyzed by another of us, a neuroimmunologist (A.V.). Statistical Analysis Prism software, version 7 (GraphPad Software) was used to create graphs and perform statistical analysis. Two-tailed paired t tests were used to compare scores before and after treatments, and a 2-sided P < .05 was considered significant. Results Of the 38 patients, 25 (66%) were male and the median (range) age was 55 (12-85) years. Twenty-three (60.5%) were Japanese and 15 (39.5%) were of white race/ethnicity. The cohort was defined by typical history and EMG findings of peripheral nerve hyperexcitability. Nerve conduction was normal in 10 of 12 patients examined; 2 patients had evidence of neuropathy (1 polyneuropathy; 1 sensory more than motor).

wenty-three (60.5%) were Japanese and 15 (39.5%) were of white race/ethnicity. The cohort was defined by typical history and EMG findings of peripheral nerve hyperexcitability. Nerve conduction was normal in 10 of 12 patients examined; 2 patients had evidence of neuropathy (1 polyneuropathy; 1 sensory more than motor). Clinical Symptoms of Patients With Diagnosed NMT Initial presenting symptoms on referral comprised cramps (16 patients [42%]), fasciculations (7 [18%]), and, less commonly, symptoms of myokymia, stiffness, or spasms (Table 1). Sensory features were reported by 8 patients (21%). Possible precipitating events were recorded by 2 patients (1 “seafood poisoning and vomiting,” 1 “infection and exhaustion”). The symptoms recorded in response to specific questions at the first clinic visit were more widespread, including cramps (32 [84%]) and muscle twitching (30 [79%]), as well as sweating (12 [32%]), weakness (13 [34%]), and stiffness (12 [32%]) (Table 1). In addition, autonomic disturbance involving excessive secretions, sweating, diarrhea, tachycardia, and tachypnea was evident in 18 patients (47%) (Table 1). One patient had evidence of orthostatic hypotension. The most striking additional complaint reported was sensory disturbance. Specifically, 20 patients (53%) complained of paresthesia (5 [13%]), pain (8 [21%]), or both (7 [18%]). Pain and paresthesia were typically reported in the legs or arms and sometimes in all limbs. Pain was described as burning or throbbing in 7 patients (18%). One patient specifically reported pain from muscle cramps.

Specifically, 20 patients (53%) complained of paresthesia (5 [13%]), pain (8 [21%]), or both (7 [18%]). Pain and paresthesia were typically reported in the legs or arms and sometimes in all limbs. Pain was described as burning or throbbing in 7 patients (18%). One patient specifically reported pain from muscle cramps. With respect to central nervous system symptoms, sleep disturbance, particularly insomnia, was present in 13 patients (34%), anxiety and agitation in 10 (26%), and depression in 5 (13%). Two male patients (5%) attempted suicide. One patient had new-onset seizures. Additional investigations included cerebrospinal fluid analysis in 8 patients, which had normal results with the exception of 2 patients (25%) who had a small increase in lymphocyte counts. Electroencephalogram results were normal in 8 of 9 patients (89%) (1 woman with a locally invasive thymoma had focal spikes and epilepsy with normal findings on magnetic resonance imaging), and normal findings on magnetic resonance imaging in 18 of 20 others (90%) (1 had degenerative spinal changes and 1 had temporal lobe atrophy of unknown cause). Other Autoimmune Disorders or Tumors Eight of the 38 patients (21%) had recent tumors, 5 thymomas (3 males, 2 females), and 3 prostate tumors; 1 had a previously treated non–small cell cancer; and 1 had acute lymphatic leukemia after bone marrow transplantation. Acetylcholine receptor antibodies were positive in 3 patients with thymoma and Hu antibodies in another patient with thymoma.

had recent tumors, 5 thymomas (3 males, 2 females), and 3 prostate tumors; 1 had a previously treated non–small cell cancer; and 1 had acute lymphatic leukemia after bone marrow transplantation. Acetylcholine receptor antibodies were positive in 3 patients with thymoma and Hu antibodies in another patient with thymoma. Overall, autoimmunity or other comorbidities (eTable 3 in the Supplement), mainly diabetes or hypertension, were reported in 20 patients (53%). Antinuclear antibodies were present in 4 patients and anti–thyroid peroxidase antibodies in 1. Treatments Modified Rankin Scale scores and information on treatments used were available for 28 patients. The scores were variable but similar between male and female patients and generally improved after treatment (Figure 1A), but patients with mild disease who were untreated (n = 3) or given treatment for symptoms only (2 received phenytoin; 4, carbamazepine; and 1, clonazepam plus diazepam) had insignificant benefits (Figure 1B). The remaining 18 patients had more severe disease (mean [SD] mRS score, 3.39 [1.04] vs 2.0 [0.81]; P = .001) and had received a range of immunotherapies only or in combination with drugs for symptoms, with clear benefits in most (Figure 1B). The 7 patients with tumors had more severe disease (mRS, 3.75 [1.04]) than the 21 patients without tumors mRS 2.5 [0.98], P = .004) but also responded well to treatments (Figure 1C). Despite the improvements, in a number of patients specific drugs were judged unhelpful (a summary of treatment responses is provided in eTable 4 in the Supplement).

had more severe disease (mRS, 3.75 [1.04]) than the 21 patients without tumors mRS 2.5 [0.98], P = .004) but also responded well to treatments (Figure 1C). Despite the improvements, in a number of patients specific drugs were judged unhelpful (a summary of treatment responses is provided in eTable 4 in the Supplement). Figure 1. Modified Rankin Scale (mRS) Scores in 28 Patients With Neuromyotonia (NMT) Before and After Treatmenta A, Scores in male and female patients before treatment and at follow-up (FU). B, Scores before treatment and at FU according to treatment type. Further information on treatment responses is given in eTable 4 in the Supplement. C, Scores in patients with tumors (3 thymomas, 3 prostate tumors, and 1 acute lymphocytic leukemia after bone marrow transplant) and without tumors. In most cases, mRS scores were lower at FU (2-tailed paired t tests). The mRS measures disability on a range of 0 to 6, with 0 indicating normal and 6 indicating death. Solid horizontal lines indicate means. P values were determined by 2-tailed t test. AED indicates antiepileptic drugs and other symptomatic therapies; IT, immunotherapy. aP < .001 for the change in mRS score at FU. bP = .02 for the change in mRS score at FU. cP = .001 for the change in mRS score at FU.

Figure 1. Modified Rankin Scale (mRS) Scores in 28 Patients With Neuromyotonia (NMT) Before and After Treatmenta A, Scores in male and female patients before treatment and at follow-up (FU). B, Scores before treatment and at FU according to treatment type. Further information on treatment responses is given in eTable 4 in the Supplement. C, Scores in patients with tumors (3 thymomas, 3 prostate tumors, and 1 acute lymphocytic leukemia after bone marrow transplant) and without tumors. In most cases, mRS scores were lower at FU (2-tailed paired t tests). The mRS measures disability on a range of 0 to 6, with 0 indicating normal and 6 indicating death. Solid horizontal lines indicate means. P values were determined by 2-tailed t test. AED indicates antiepileptic drugs and other symptomatic therapies; IT, immunotherapy. aP < .001 for the change in mRS score at FU. bP = .02 for the change in mRS score at FU. cP = .001 for the change in mRS score at FU. Autoantibodies and Clinical Associations All patient serum samples were studied together at the University of Oxford, Oxford, United Kingdom, for VGKC-complex antibodies by radioimmunoprecipitation and for LGI1, CASPR2, and contactin 2 antibodies by live cell–based assays (as used routinely in the University of Oxford clinical service). Eleven of the serum samples (29%) (9 from men and 2 from women)were VGKC-complex antibody positive. However, results of the live cell–based assays were positive in 17 of the 38 patients (45%), including CASPR2 antibodies in 11 (29%), LGI1 antibodies in 8 (21%), and contactin 2 antibodies in 5 (13%) (Figure 2A). Results from 8 of the 17 assays were positive for VGKC-complex antibodies by radioimmunoprecipitation, but high levels were only found in patients with both CASPR2 and LGI1 antibodies (Figure 2B). The antibodies occurred either singly or in combinations, and the live cell–based assay scores are shown in Figure 2C. Overall, antibody specificity did not influence mRS scores before or after treatments (Figure 2D and Table 2).

high levels were only found in patients with both CASPR2 and LGI1 antibodies (Figure 2B). The antibodies occurred either singly or in combinations, and the live cell–based assay scores are shown in Figure 2C. Overall, antibody specificity did not influence mRS scores before or after treatments (Figure 2D and Table 2). Figure 2. Antibodies in Patients With Neuromyotonia (NMT) A, A serum sample positive for both contactin-associated protein 2 (CASPR2) and leucine-rich glioma inactivated protein 1 (LGI1) antibodies. The binding of the patient’s IgG antibodies (Abs) to human embryonic kidney cells (enhanced green fluorescent protein [EGFP] [green] label) was detected with Alexa fluor anti-human IgG (red). Binding was scored on a scale of 0 to 4, with 0 indicating negative; 1, positive; and 2 to 4, increasing positivity. This serum sample scored 2.5 and 2.0 as shown. Results of tests for contactin 2 Abs were negative (data not shown). This patient had mild disease (modified Rankin Scale [mRS] score, 2), unlike the other 5 patients with LGI1 and CASPR2 Abs (mRS scores, 3-5), and his disease responded to carbamazepine with immunotherapies (mRS score, 1). B, Voltage-gated potassium channel (VGKC)–complex antibody titers associated with the presence of specific Abs or none. The horizontal line indicates the cutoff for positivity, 1. Eleven serum samples were positive (>100pM) by radioimmunoprecipitation, but high titers were only found in those with LGI1 and CASPR2 Abs. C, The cell-based assay (CBA) scores in patients with either a single or 2 different Ab specificities. The horizontal line indicates the cutoff for positivity, 100pM. CASPR2 and LGI1 antibodies together gave the highest binding scores, but few reached the maximum score of 4. D, Modified Rankin Scale scores were not different between patients with or without detectable CBA Abs, but in both cases mRS scores decreased at follow-up (FU) (2-tailed paired t tests).

for positivity, 100pM. CASPR2 and LGI1 antibodies together gave the highest binding scores, but few reached the maximum score of 4. D, Modified Rankin Scale scores were not different between patients with or without detectable CBA Abs, but in both cases mRS scores decreased at follow-up (FU) (2-tailed paired t tests). aP = .003 for the change in mRS score at FU. bP = .002 for the change in mRS score at FU.

for positivity, 100pM. CASPR2 and LGI1 antibodies together gave the highest binding scores, but few reached the maximum score of 4. D, Modified Rankin Scale scores were not different between patients with or without detectable CBA Abs, but in both cases mRS scores decreased at follow-up (FU) (2-tailed paired t tests). aP = .003 for the change in mRS score at FU. bP = .002 for the change in mRS score at FU. Table 2. Antibodies and Clinical Features in 38 Patients With Electromyography-Confirmed Neuromyotoniaa Antibody Finding No. of Patients (No. M:F) Clinical Feature, No. of Patients Mean (SD) mRS Score Before/After Treatments P Value Pain Autonomic Features Insomnia/Other Sleep Disturbance Anxiety/Agitation/Depression Additional Features or Abs Ab negative (n = 18) or VGKC-complex only (n = 3) 21 (14:7) 11 11 4/0 4/2/2 1 Thymoma, 1 prostate, 1 CRPS-like 2.92 (1.17)/1.25 (0.87) (n = 12) .002 Specific Ab positive 17 (11:6) 9 7 4/4 3/3/2 4 Thymoma, 2 prostate, 2 AChR-Ab 2.9 (1.2)/1.8 (0.68) (n = 16) .002 CASPR2 and LGI1 6 (5:1) 5 6 3/2 2/2/0 3 Thymoma (2 invasive, 1 AChR-Ab), 1 prostate, 1 limbic encephalitis 3.83 (1.17)/1.67 (0.52) (n = 6) .002 CASPR2 (n = 4) or CASPR2 and contactin 2 (n = 1) 5 (1:4) 1 0 0/2 0/0/1 1 Thymoma, 1 prostate, 1 AChR-Ab MG 2.50 (1.29)/1.75 (0.96) (n = 4) .39 Contactin 2 only 4 (3:1) 3 1 1/0 1/1/1 None 2.25 (0.50)/1.75 (0.96) (n = 4) .39 LGI1 only 2 (2:0) 0 0 0/0 0/0/0 None 2.0/2.0b NA Abbreviations: Ab, antibody; AChR, acetylcholine receptor; CASPR2, contactin-associated protein 2; CRPS, complex regional pain syndrome; LGI1, leucine-rich glioma inactivated protein 1; MG, myasthenia gravis; mRS, modified Rankin Scale; NA, not applicable; VGKC, voltage-gated calcium channel.

0/0 0/0/0 None 2.0/2.0b NA Abbreviations: Ab, antibody; AChR, acetylcholine receptor; CASPR2, contactin-associated protein 2; CRPS, complex regional pain syndrome; LGI1, leucine-rich glioma inactivated protein 1; MG, myasthenia gravis; mRS, modified Rankin Scale; NA, not applicable; VGKC, voltage-gated calcium channel. a Two-tailed t tests used that were not corrected for multiple comparisons. Numbers in the final column are limited to those with pretreatment and follow-up treatment mRS scores. b The SDs were not determined. However, of the 6 patients with both CASPR2 and LGI1 antibodies, 3 (50%) had thymoma and 1 had developed prostate cancer; 1 had mild disease that responded to carbamazepine alone. Each of these patients had autonomic symptoms as well as typical NMT, 5 complained of pain, and 4 had neuropsychiatric features. Five had sleep disturbance and 3 fulfilled the criteria for Morvan syndrome (NMT, autonomic disturbance, and insomnia),11 although 1 had an additional central nervous system feature (seizures). As a group, they had higher mean pretreatment mRS scores (mean [SD], 3.83 [1.17]) compared with all other patients (2.71 [1.08]), but both groups had similar posttreatment scores (1.67 [0.52] vs 1.52 [0.88]; Table 2). Serum samples were also tested simultaneously using a widely available commercial assay. Positivity for LGI1 and CASPR2 antibodies was confirmed except for 1 CASPR2 antibody that was not detected with the commercial assay (data not shown).

However, of the 6 patients with both CASPR2 and LGI1 antibodies, 3 (50%) had thymoma and 1 had developed prostate cancer; 1 had mild disease that responded to carbamazepine alone. Each of these patients had autonomic symptoms as well as typical NMT, 5 complained of pain, and 4 had neuropsychiatric features. Five had sleep disturbance and 3 fulfilled the criteria for Morvan syndrome (NMT, autonomic disturbance, and insomnia),11 although 1 had an additional central nervous system feature (seizures). As a group, they had higher mean pretreatment mRS scores (mean [SD], 3.83 [1.17]) compared with all other patients (2.71 [1.08]), but both groups had similar posttreatment scores (1.67 [0.52] vs 1.52 [0.88]; Table 2). Serum samples were also tested simultaneously using a widely available commercial assay. Positivity for LGI1 and CASPR2 antibodies was confirmed except for 1 CASPR2 antibody that was not detected with the commercial assay (data not shown). Results of the Independent Patient-Led Pain Questionnaire Because of the reports of pain in 20 patients (53%) that were often unrelated to muscle cramps, an independent survey was conducted via an Isaacs syndrome website, and questionnaires were sent to 165 individuals who reported pain. The patients were asked to describe the level of pain at best and at worst (0 representing no pain to 10 representing very bad or incapacitating pain), describe the nature of the pain, and detail where the pain was, what factors made it better or worse, and the extent to which it had responded to treatments. To assess the effects on quality of life, patients were asked the extent to which pain affected sleep, relationships, work, exercise, mood, and general activities (summarized in the Box and detailed in eTable 2 in the Supplement). Of the 56 of 165 individuals (34%) who returned detailed responses (32 males and 24 females; median [range] age, 50 [12-85] years) from different countries (mainly the United States, United Kingdom, and Australia), 8 had been given diagnoses of NMT, 28 of Isaacs syndrome, 4 of peripheral nerve hyperexcitability, 9 of cramp fasciculation syndrome, and 7, something else (1, Morvan; 1, cramps; 2, fibromyalgia; and 3, awaiting a diagnosis). The questions and scoring requested are summarized in the Box and detailed in eTable 2 in the Supplement. Pain was reported in the legs in 44 participants (79%) plus feet (16 [29%]) or arms (28 [50%]) but also in the trunk or neck (19 [34%]) (Figure 3A). Cramps and aching pain were the most common, but shooting or “electric” pain were noted by 17 (30%) and stabbing or pulsating pain by 13 (23%) (Figure 3B); burning pain was reported by 8 participants (14%). Most patients experienced intermittent pain, with scores differing substantially between “at best” and “at worst,” and treatments appeared helpful (Figure 3C). In many patients, pain had, or previously had, a major influence on aspects of daily living (Figure 3D).

3B); burning pain was reported by 8 participants (14%). Most patients experienced intermittent pain, with scores differing substantially between “at best” and “at worst,” and treatments appeared helpful (Figure 3C). In many patients, pain had, or previously had, a major influence on aspects of daily living (Figure 3D). Box. Summary of Pain Questionnaire Sent to Individuals Responding to a Patient-Led Online Surveya Country, sex, age, medical diagnosis (eg, NMT, Isaacs syndrome, peripheral nerve hyperexcitability, CFS, or other). How bad is the pain at best (1 = little pain, 10 = very painful-incapacitating)? How bad is the pain at worst (1 = little pain, 10 = very painful-incapacitating)? How would you describe the types of pain(s) you feel, and where on the body they occur? Please also add how bad is/are the pain(s) in each area. What if anything make the pain(s) worse (eg, temperature changes, any form of exercise or exertion, stress, or any foods)? When do your pain symptoms occur (eg, in the morning/evening, any time of day, intermittent, or constant)? How effective have your prescriptions been at helping with pain(s) (1 = totally ineffective, 10 = fully effective)? For the following questions, please select between 1 and 10 as appropriate (1 = has had little effect, 10 = affects sleep very much). How much would you say pain affects your enjoyment of life? How much would you say the pain(s) affects sleep? How much would you say pain affects your relations at /home/with loved ones/friends? How much would you say pain affects your ability to work (includes work around the home)?

For the following questions, please select between 1 and 10 as appropriate (1 = has had little effect, 10 = affects sleep very much). How much would you say pain affects your enjoyment of life? How much would you say the pain(s) affects sleep? How much would you say pain affects your relations at /home/with loved ones/friends? How much would you say pain affects your ability to work (includes work around the home)? How much would you say pain affects your walking ability? How much would you say pain affects your mood? How much would you say pain affects overall day-to-day living and general activities? Finally, if you wish, comment further about the pain(s) you suffer and how they affect you. Abbreviations: CFS, cramp fasciculation syndrome; NMT, neuromyotonia. a This is a simplified version of the form sent to patients who responded to the internet survey as having pain. The full form, which includes example answers for each question, is available in eTable 1 in the Supplement. The results were entered into a spreadsheet file after deidentification by one of us (R.B.) and analyzed by another of us (A.V.).

of the form sent to patients who responded to the internet survey as having pain. The full form, which includes example answers for each question, is available in eTable 1 in the Supplement. The results were entered into a spreadsheet file after deidentification by one of us (R.B.) and analyzed by another of us (A.V.). Figure 3. Characteristics of Pain in 56 Patients With Neuromyotonia (NMT) or Related Diagnoses A, Percentage of patients reporting pain in different anatomic regions. B, Percentage of patients reporting different types of pain. C, Pain scores at best, at worst, and after treatment. D, Influence of pain on quality of life. Patients responding to a patient-led online survey were asked to score their experiences of pain and its consequences on a scale of 0 to 10, with 0 indicating none or no effect and 10 indicating incapacitating or substantial effect. A summary of the questionnaire is in the Box, and the full questionnaire is provided in eTable 2 in the Supplement. Results in C and D are shown as median scores. Whiskers indicate interquartile range.

ces on a scale of 0 to 10, with 0 indicating none or no effect and 10 indicating incapacitating or substantial effect. A summary of the questionnaire is in the Box, and the full questionnaire is provided in eTable 2 in the Supplement. Results in C and D are shown as median scores. Whiskers indicate interquartile range. Discussion There have been few detailed descriptions of NMT,14,15,16,20 and none have studied in detail the associated clinical features, serologic characterization, and treatment responses. The results of this cohort study of 38 patients with EMG-defined NMT found 8 (21%) with concurrent tumors (5 thymoma, 3 prostate); when asked, a high proportion reported additional features of autonomic disturbance, pain, and central nervous system dysfunction. Antibodies to the now well-recognized VGKC-complex antigens were present in 17 of the 38 patients (45%), but these included 6 (16%) patients with both LGI1 and CASPR2 antibodies, and these patients had more severe disease, thymoma (in 50%), and features of Morvan syndrome. Contactin 2 antibodies were relatively common (5 patients [13%]) compared with their incidence in other VGKC-complex antibody disorders.8 After receiving treatment, all but 4 patients with follow-up data improved by at least 1 mRS score. Therapies to treat symptoms were moderately effective in the few patients with relatively mild disease; however, in others, improvement was related to immunotherapy with or without symptom treatments.

ntibody disorders.8 After receiving treatment, all but 4 patients with follow-up data improved by at least 1 mRS score. Therapies to treat symptoms were moderately effective in the few patients with relatively mild disease; however, in others, improvement was related to immunotherapy with or without symptom treatments. The high frequency of pain among the patients with EMG-defined NMT (20 patients [53%]) exceeded that reported in previous studies, and the nature of the pain in NMT has only occasionally been described13,22,23 and may previously have been ascribed mainly to muscle cramps. However, the study patients with NMT reported a variety of painful symptoms as reflected in the detailed responses reported via the independent, patient-led online questionnaire that included fuller descriptions of the pain, its distribution and effects on aspects of daily living, such as relationships and work. A systematic, prospective study of pain in NMT needs to be undertaken.

ul symptoms as reflected in the detailed responses reported via the independent, patient-led online questionnaire that included fuller descriptions of the pain, its distribution and effects on aspects of daily living, such as relationships and work. A systematic, prospective study of pain in NMT needs to be undertaken. Another feature of this study was the number of patients who reported autonomic features. These have always been recognized in NMT,3,13 but direct questioning resulted in almost 50% of both male and female patients reporting diverse symptoms. In addition, sleep disturbance, including insomnia, were frequent. Morvan syndrome, which is defined by NMT, autonomic disturbance, and insomnia with encephalopathy, is thought to be a rare disease, with fewer than 100 cases reported in the literature.10,11,24 One typical patient with Morvan syndrome and high-titer CASPR2 antibodies was not included here because the diagnosis had already been given and serum was no longer available. Nevertheless, 3 of the 6 patients (50%) with both CASPR2 and LGI1 antibodies had many features of Morvan syndrome, including thymoma; however, none had developed encephalopathy, and the diagnosis remained NMT. Thus, NMT can be seen as a forme fruste of Morvan syndrome.

eady been given and serum was no longer available. Nevertheless, 3 of the 6 patients (50%) with both CASPR2 and LGI1 antibodies had many features of Morvan syndrome, including thymoma; however, none had developed encephalopathy, and the diagnosis remained NMT. Thus, NMT can be seen as a forme fruste of Morvan syndrome. In this study, 3 of the patients had VGKC-complex antibodies without evidence of positivity for CASPR2, LGI1, or contactin 2. Although VGKC-complex antibodies can be at high titer (>400pM)8 in patients with NMT, the titers originally reported3,13 were mostly lower (<400pM), and in the present study, 47% of patients were negative for all antibody tests (<100pM, the cutoff used at the University of Oxford; other centers with lower cutoffs may calculate titers differently). In fact, the clinical relevance of many lower titers, often without evidence of antibodies to LGI1, CASPR2, or contactin 2, is unclear.25,26 By contrast, LGI1 antibodies modulate VGKC function on brain slices and α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid currents in hippocampal neurons27,28 and internalize the LGI1–disintegrin and metalloproteinase 22 receptor complex in transfected human embryonic kidney cells.29 CASPR2 antibodies did not internalize CASPR2 on hippocampal neurons in one study,30 but in vivo transfer of CASPR2-antibodies to mice produced afferent nerve hyperexcitability and mechanical allodynia with loss of surface CASPR2 and VGKC Kv1 subunits in the dorsal root ganglia; these and additional dorsal horn changes were found in CASPR2−/− mice.18 These findings are typical of neuropathic pain models and consistent with a contribution of CASPR2 antibodies to producing pain via an effect at the levels of the dorsal root in NMT. Autoimmunity in patients with pain is an emerging field of research.31

and additional dorsal horn changes were found in CASPR2−/− mice.18 These findings are typical of neuropathic pain models and consistent with a contribution of CASPR2 antibodies to producing pain via an effect at the levels of the dorsal root in NMT. Autoimmunity in patients with pain is an emerging field of research.31 The data here and from other studies17,22,23 suggest that CASPR2, LGI1, and contactin 2 antibodies are more relevant than VGKC-complex antibodies in patients with suspected NMT. Nevertheless, fewer than half of the patients with suspected acquired autoimmune NMT were positive for the antibodies tested here, and other targets may need to be identified. At present, neurophysiologic assessment remains the criterion standard for diagnosis combined with clinical phenotyping. Treatment protocols for NMT have not been formalized. Most patients will be given antiepileptic therapies or muscle relaxants as a first-line treatment because these can be sufficient; in the present series, these therapies were most often prescribed for patients with less severe disability. The remaining patients were given these treatments plus corticosteroids, immunotherapies, or both. Such therapies were effective and most patients reported benefit, but some patients noted little benefit. In patients with symptoms of NMT refractory to immunotherapies, the possibility of nonimmunologic disorders should be considered.

patients were given these treatments plus corticosteroids, immunotherapies, or both. Such therapies were effective and most patients reported benefit, but some patients noted little benefit. In patients with symptoms of NMT refractory to immunotherapies, the possibility of nonimmunologic disorders should be considered. Limitations Neuromyotonia is a rare disorder and consequently study numbers tend not to be large. Furthermore, as a study driven by patients and their clinical needs rather than through an established protocol, data were typically collected at follow-up reviews. As such, it was not always possible to obtain mRS scores from all patients before and after treatment. Serologic studies of LGI1, CASPR2, and contactin-2 antibodies were performed on archived samples that had been sent to the Oxford reference laboratory as part of routine diagnostic screening from Sydney and Kagoshima and were then retested for this study. Separately, although there are limitations with relying on a patient-initiated questionnaire—including a lack of response in those patients who did not report pain—it is increasingly recognized that patient-reported outcome measures may better reflect real-world experiences, which in this case have provided novel insight into the variable frequency and characteristics of pain in patients diagnosed with NMT and on the effects on their daily activities.

ents who did not report pain—it is increasingly recognized that patient-reported outcome measures may better reflect real-world experiences, which in this case have provided novel insight into the variable frequency and characteristics of pain in patients diagnosed with NMT and on the effects on their daily activities. Conclusions Although cramps, muscle fasciculations, and pain are common symptoms, neuromyotonia remains a relatively rare diagnosis, and it is likely that less severe cases are often unrecognized. Nevertheless, comorbidities such as thymoma or myasthenia and sometimes associated symptoms of autonomic and central nervous system disturbance, and pain, a potentially disabling feature that is not always appreciated in clinical practice, mean that the disease deserves to be investigated for underlying triggers while also treated appropriately. Serologic studies are not positive in all patients, and neurophysiology remains a key diagnostic tool that also serves as a means of phenotyping patients. Moreover, although heterogeneity of clinical features and related antibodies may limit interpretation of significant correlations, the coexistence of LGI1 and CASPR2 antibodies may suggest the presence of thymoma, often accompanied by autonomic and central nervous system involvement. Supplement. eTable 1. Clinical Datasheet eTable 2. Patient-led On-line Questionnaire as Sent to 165 Individual Patients Who Reported Pain eTable 3. Other Autoimmunity or Co-morbidities as Reported by Clinicians eTable 4. Responses to Different Treatments Click here for additional data file.