Browse the corpus

Introduction Shared decision making is a process that helps patients and clinicians align therapies with patients’ values and preferences. This is particularly important for invasive therapies for life-threatening illness, such as left ventricular assist devices as destination therapy (DT LVADs). These devices are offered to patients with severe heart failure who are ineligible for cardiac transplantation. Placement of a DT LVAD can be life-prolonging, but also comes with considerable changes in lifestyle, need for caregiver support, and a high likelihood of complications. Unfortunately, for many major decisions involving newer medical technologies—including DT LVAD—education, consent, and shared decision-making processes are suboptimal. Current consent documents are too long and poorly written, industry materials tend to be biased, and clinicians often lack the skills to support high-quality decision making. Patient decision aids are a form of decision support that standardize the information received by patients to support a shared decision-making process. Decision aids have been shown to improve knowledge and reduce decisional conflict. However, few tools have been developed to engage seriously ill patients in shared decision making, and until recently, none were available for LVAD. In addition, the effectiveness of decision aids in clinical practice is largely unknown and wide-scale implementation remains a substantial challenge.

decisional conflict. However, few tools have been developed to engage seriously ill patients in shared decision making, and until recently, none were available for LVAD. In addition, the effectiveness of decision aids in clinical practice is largely unknown and wide-scale implementation remains a substantial challenge. In this context, we performed a series of studies exploring the decisional needs for patients and their caregivers considering DT LVAD. Based on this research and following the International Patient Decision Aid Standards (IPDAS), we developed pamphlet and video decision aids for patients and their caregivers considering DT LVAD. We aimed to study the effectiveness of these decision aids coupled with a clinician-directed support training through a multicenter, cluster-randomized, stepped-wedge design. Methods Design and Sites The Multicenter Trial of a Shared Decision Support Intervention for Patients and their Caregivers Offered Destination Therapy for End-Stage Heart Failure (DECIDE-LVAD) used a hospital-level, randomized phased roll out (stepped wedge) in 6 mechanical circulatory support (MCS) programs across the United States. This approach was chosen because the intervention engages clinicians and other program staff in addition to patients. The study was overseen by the institutional review board at the University of Colorado and approved by institutional review boards at all sites. The trial protocol is available in Supplement 1.

States. This approach was chosen because the intervention engages clinicians and other program staff in addition to patients. The study was overseen by the institutional review board at the University of Colorado and approved by institutional review boards at all sites. The trial protocol is available in Supplement 1. Study Participants Patient and caregiver dyads were enrolled from the 6 sites during a 20-month enrollment period. Patient eligibility criteria included age 18 years or older, end-stage heart failure, and active consideration for a DT LVAD. Eligible patients were identified by the study team at each site at the time a DT LVAD evaluation was initiated. This was triggered either by a preauthorization request to the patient’s health insurance for LVAD evaluation or a provider’s request for formal education about LVAD. Written informed consent was obtained from all study participants. Participants were compensated $25 for completing surveys at each time point. Intervention The content and development of the decision aids is described separately; the pamphlet is available in Supplement 2.

Study Participants Patient and caregiver dyads were enrolled from the 6 sites during a 20-month enrollment period. Patient eligibility criteria included age 18 years or older, end-stage heart failure, and active consideration for a DT LVAD. Eligible patients were identified by the study team at each site at the time a DT LVAD evaluation was initiated. This was triggered either by a preauthorization request to the patient’s health insurance for LVAD evaluation or a provider’s request for formal education about LVAD. Written informed consent was obtained from all study participants. Participants were compensated $25 for completing surveys at each time point. Intervention The content and development of the decision aids is described separately; the pamphlet is available in Supplement 2. All sites began in the control period using their existing materials during formal education. This process typically consisted of a patient teaching session with an MCS coordinator and use of industry pamphlets/videos and program-specific documents. At 4 stepped time intervals, programs were randomly assigned to cross over to the intervention. The decision support intervention included (1) delivery of a 2.5-hour clinician-directed decision support training and (2) use of the 26-minute video and 8-page pamphlet decision aids developed by our group. At the time of intervention implementation, staff directly involved in LVAD patient education and care were asked to attend a 60-minute grand rounds style presentation about shared decision making for DT LVAD, followed by a 90-minute coaching session that included demonstration and discussion of the decision aid materials. With facilitation by local physician champions, sites were then instructed to formally integrate the decision aids and tenets learned from the coaching session into existing education, decision making, and informed consent processes. The only requirements of sites around the use of the decision aids were that they be delivered by clinicians and not research staff. This design allowed for sites to implement the decision aids in a way that was most appropriate for that site; thus, local differences in how the intervention was delivered were possible.

The only requirements of sites around the use of the decision aids were that they be delivered by clinicians and not research staff. This design allowed for sites to implement the decision aids in a way that was most appropriate for that site; thus, local differences in how the intervention was delivered were possible. Data Collection Data collection was the same during both control and intervention periods. For all patients meeting initial eligibility criteria, basic demographic and health status data were captured in the screening form. Enrolled participants were administered surveys (Supplement 2) at 4 time points: prior to formal LVAD education (baseline 1), immediately after formal education (baseline 2), 1 month after enrollment, and 6 months after enrollment. Baseline surveys were administered in person (with verbal assessments completed), whereas follow-up surveys were administered in person or through telephone or mail. Medical records were reviewed at enrollment, 1 month, and 6 months for relevant clinical information, outcomes, and adverse events. Decision aid use by individual patients was reported by clinical staff. All data were entered into a REDCap database by the individual sites. Outcomes The primary outcome was decision quality: the extent to which medical decision making reflects the considered preferences of a well-informed patient. As such, coprimary endpoints were chosen comprising the 2 main IPDAS domains of decision quality—knowledge and values-choice concordance.

Data Collection Data collection was the same during both control and intervention periods. For all patients meeting initial eligibility criteria, basic demographic and health status data were captured in the screening form. Enrolled participants were administered surveys (Supplement 2) at 4 time points: prior to formal LVAD education (baseline 1), immediately after formal education (baseline 2), 1 month after enrollment, and 6 months after enrollment. Baseline surveys were administered in person (with verbal assessments completed), whereas follow-up surveys were administered in person or through telephone or mail. Medical records were reviewed at enrollment, 1 month, and 6 months for relevant clinical information, outcomes, and adverse events. Decision aid use by individual patients was reported by clinical staff. All data were entered into a REDCap database by the individual sites. Outcomes The primary outcome was decision quality: the extent to which medical decision making reflects the considered preferences of a well-informed patient. As such, coprimary endpoints were chosen comprising the 2 main IPDAS domains of decision quality—knowledge and values-choice concordance. A 10-item knowledge test was developed by the study team and validated by clinicians and patients. Consistent with the methods of Sepucha et al, the study team created a list of knowledge items based on clinical needs, local post-LVAD education standards, and needs assessment work with patients. We then surveyed patients, caregivers, MCS coordinators, and physicians to narrow the list and determine the key knowledge items and assure content validity. The acceptability of this measure was further assessed with patients and caregivers in a pilot of the trial protocol. Improvement in knowledge from baseline 1 to baseline 2 was a coprimary endpoint. A values scale was also developed, modeled after a well-accepted values evaluation tool. During previous needs assessment work, 1 value rose above all others in considering DT LVAD: maximizing chances of survival with aggressive medical care vs not. We developed a single-item, 10-tier Likert values measure using the dichotomy of “Do everything I can to live longer, even if that means having major surgery and being dependent on a machine” (score 1) vs “Live with whatever time I have left, without going through major surgery or being dependent on a machine” (score 10). Concordance between 1-month value score and patient-reported treatment choice (DT LVAD or medical treatment without LVAD) at 1 month was the other coprimary endpoint. Concordance between 1-month value score and actual treatment received by 6 months was also assessed.

surgery or being dependent on a machine” (score 10). Concordance between 1-month value score and patient-reported treatment choice (DT LVAD or medical treatment without LVAD) at 1 month was the other coprimary endpoint. Concordance between 1-month value score and actual treatment received by 6 months was also assessed. Secondary outcomes included validated measures of decision conflict, decision regret, control preferences, illness acceptance, perceived stress, depression (Patient Health Questionnaire-2), and quality of life (EuroQol Visual Analogue Scale). Acceptability of the decision aids was also measured at baseline 2. Analysis We determined that a sample size of 168 participants with standard deviation of 18% would yield a power of 0.95 to detect an improvement in knowledge by 10%. We anticipated a dropout rate of 25% by 6 months based on expected death rates and other loss to follow-up. We compared baseline characteristics between participants enrolled in the study to those screened but not enrolled using χ2 tests. We compared characteristics between those enrolled during the control period with those enrolled during the intervention period using χ2 tests and t tests.

Analysis We determined that a sample size of 168 participants with standard deviation of 18% would yield a power of 0.95 to detect an improvement in knowledge by 10%. We anticipated a dropout rate of 25% by 6 months based on expected death rates and other loss to follow-up. We compared baseline characteristics between participants enrolled in the study to those screened but not enrolled using χ2 tests. We compared characteristics between those enrolled during the control period with those enrolled during the intervention period using χ2 tests and t tests. To assess the change in patient DT LVAD knowledge over time, we fit a linear mixed model proposed for the analysis of stepped wedge designs. This model accounted for the repeated within-person measures, included a random effect for site and fixed effect indicators of intervention group and stepped wedge time period. This model adjusts for trends over time, assuming that changes occur similarly across all sites. Owing to differences observed at baseline, we included 2 covariate indicator variables: outpatient status and diagnosis of heart failure less than 4 years prior to enrollment. We evaluated whether the change in knowledge score (percent correct) between baseline 1 and baseline 2 was different between the control and intervention groups.

ces observed at baseline, we included 2 covariate indicator variables: outpatient status and diagnosis of heart failure less than 4 years prior to enrollment. We evaluated whether the change in knowledge score (percent correct) between baseline 1 and baseline 2 was different between the control and intervention groups. To assess values-choice concordance, we calculated the Kendall’s τ correlation coefficient between the stated values score at 1 month and each of the treatment outcomes (patient-reported treatment choice at 1 month, actual treatment received at 6 months), and looked at the difference in this correlation coefficient by intervention group. To generate a confidence interval for this difference, we performed 500 bootstrap samples and calculated the 2.5th and 97.5th percentiles. In analysis of secondary survey scores described previously and the LVAD implantation rates by intervention group, we applied the same mixed model methods described above. Owing to site differences in LVAD implantation rate and differences in location of patient enrollment over time, we performed separate sensitivity analyses accounting for each, as well as sensitivity analysis without including the site random effect. Missing data analyses can be viewed in Supplement 2. All analyses were performed using SAS statistical software (version 9.4, SAS Inc).

In analysis of secondary survey scores described previously and the LVAD implantation rates by intervention group, we applied the same mixed model methods described above. Owing to site differences in LVAD implantation rate and differences in location of patient enrollment over time, we performed separate sensitivity analyses accounting for each, as well as sensitivity analysis without including the site random effect. Missing data analyses can be viewed in Supplement 2. All analyses were performed using SAS statistical software (version 9.4, SAS Inc). Results Patients Between June 2015 and January 2017, of 385 patients who were actively considering a DT LVAD, 248 were enrolled (Figure 1). Compared with patients who were screened but not enrolled in the study, enrolled patients were more likely to be white non-Hispanic (75.8% vs 63.9%, P = .03); other demographics and clinical status were not different between the 2 groups. Figure 1. Screening and Enrollment Flow Diagram Patients in the intervention period were more likely to be enrolled as outpatient (31% vs 17%, P = .007) and to have been diagnosed with heart failure less than 4 years prior (29.8% vs 18.2%, P = .03) than those enrolled in the control period. See Table 1 for other demographic information.

g and Enrollment Flow Diagram Patients in the intervention period were more likely to be enrolled as outpatient (31% vs 17%, P = .007) and to have been diagnosed with heart failure less than 4 years prior (29.8% vs 18.2%, P = .03) than those enrolled in the control period. See Table 1 for other demographic information. Table 1. Participant Baseline Characteristics Characteristic No. (%) Control (n = 135)a Intervention (n = 113)a Age, mean (SD), yb 63.5 (9.7) 63.2 (10.2) Sex, male 111 (82.2) 98 (86.7) Race/ethnicityc White, non-Hispanic 102 (79.1) 86 (82.7) Black 19 (14.7) 12 (11.5) Other 8 (6.2) 6 (5.8) Some college or more 74 (56.4) 72 (69.2) Receiving disability 35 (27.6) 33 (32.0) Annual household income <$40 000 64 (51.6) 37 (40.2) Married 95 (72.5) 68 (65.4) First diagnosed with heart failure Within past 2 years 15 (11.9) 12 (12.4) 2-4 years 9 (7.1) 19 (19.6) 4 or more years 102 (81.0) 66 (68.0) INTERMACS score 1 5 (4.3) 8 (7.9) 2-3 89 (77.4) 48 (47.5) 4-7 21 (18.3) 45 (44.6) Comorbiditiesd Peripheral vascular disease 7 (5.2) 4 (3.5) Major stroke 2 (1.5) 0 (0) Severe diabetes 12 (8.9) 11 (9.7) Chronic renal disease 31 (23.0) 23 (20.4) Pulmonary disease 12 (8.9) 4 (3.5) Liver dysfunction 6 (4.4) 5 (4.4) History of solid organ or blood cancer 10 (7.4) 8 (7.1) History of alcohol or illicit drug use 13 (9.6) 12 (10.6) Enrollment location Outpatient 23 (17.0) 35 (31.0) Inpatient (non-ICU) 83 (61.5) 48 (42.5) ICU 29 (21.5) 30 (26.5) Cognitive Function (SPMSQ) Score, mean (SD)e 0.7 (1.5) 0.7 (1.2) Intact functioning 123 (93.9) 94 (93.1) Mild impairment 5 (3.8) 6 (5.9) Severe impairment 3 (2.3) 1 (1.0) Literacy (REALM-R) Score, mean (SD)f 6.93 (1.9) 6.95 (2.0) At risk for poor literacy 30 (23.4) 27 (26.7) Subjective Numeracy Score, mean (SD)g 4.0 (1.1) 4.2 (1.1) Abbreviations: ICU, intensive care unit; INTERMACS, Interagency Registry for Mechanically Assisted Circulatory Support; REALM-R, rapid estimate of adult literacy in medicine, revised; SPMSQ, short portable mental status questionnaire.

iteracy 30 (23.4) 27 (26.7) Subjective Numeracy Score, mean (SD)g 4.0 (1.1) 4.2 (1.1) Abbreviations: ICU, intensive care unit; INTERMACS, Interagency Registry for Mechanically Assisted Circulatory Support; REALM-R, rapid estimate of adult literacy in medicine, revised; SPMSQ, short portable mental status questionnaire. a Some participants refused to answer certain demographic questions; the following items had missing data: race/ethnicity (n = 15), education (n = 13), disability status (n = 18), income (n = 32), marital status (n = 13), heart failure diagnosis timing (n = 25), SPMSQ (n = 16), REALM-R (n = 19), Numeracy Score (n = 14). b Reported from patient medical record. c Patient-reported from survey. d Used from INTERMACS preimplant data collection form, section “Concerns and Contraindications.” e Number of incorrect answers out of 10 questions. f Number of correctly read words out of 8 listed. g Range of 1 to 6, higher numeracy toward 6. Exposure to and Impression of Decision Support Materials In the control period, patients most often received site-specific consent forms, locally made documents, and industry pamphlets/videos during formal education. In the intervention period, 94.7% of patients received the decision aid (99 patients received both pamphlet and video, 2 received pamphlet only, 6 received video only, and 6 received neither); 3 of the 6 sites stopped using industry pamphlets/videos labeled as decision making materials. Patient-reported acceptability of the educational materials was not significantly different between the control and intervention periods.

et and video, 2 received pamphlet only, 6 received video only, and 6 received neither); 3 of the 6 sites stopped using industry pamphlets/videos labeled as decision making materials. Patient-reported acceptability of the educational materials was not significantly different between the control and intervention periods. Decision Quality: Knowledge and Values-Choice Concordance Patient knowledge (mean test performance) during the decision-making period improved from 59.5% to 64.9% in the control group vs 59.1% to 70.0% in the intervention group (difference of difference, 5.5%; P = .03) (Figure 2). Figure 2. Primary Outcome, Decision Quality LVAD indiciates left ventricular assist device. Decision quality measured by improved knowledge following the education and consent process and concordance between patient values and treatment choice at 1 month after initiation of evaluation for LVAD. Patient values on the 10-tier Likert scale were generally in the direction of more aggressive care to maximize survival (ie, closer to 1): control period baseline 1 score mean of 2.19 (standard error [SE], 0.26); 1-month, 2.37 (SE, 0.28); 6-month, 3.12 (SE, 0.33); intervention period baseline, 2.98 (SE, 0.30); 1-month, 3.33 (SE, 0.32); 6-month, 3.65 (SE, 0.39); adjusted overall difference P = .06 (Table 2).

ive care to maximize survival (ie, closer to 1): control period baseline 1 score mean of 2.19 (standard error [SE], 0.26); 1-month, 2.37 (SE, 0.28); 6-month, 3.12 (SE, 0.33); intervention period baseline, 2.98 (SE, 0.30); 1-month, 3.33 (SE, 0.32); 6-month, 3.65 (SE, 0.39); adjusted overall difference P = .06 (Table 2). Table 2. Outcomes Outcome Visit Control (n = 132) Intervention (n = 104) P Value Knowledge score, percent correct (10-item test), mean (SE), %a BL1 59.5 (1.9) 59.1 (2.2) .92 BL2 64.9 (1.8) 70.0 (2.1) .09 1 mo 67.8 (1.9) 66.4 (2.3) .64 6 mo 68.6 (1.8) 67.1 (2.2) .63 Values score (scale 1-10), mean (SE)b BL1 2.19 (0.26) 2.98 (0.30) .06 1 mo 2.37 (0.28) 3.33 (0.32) .03 6 mo 3.12 (0.33) 3.65 (0.39) .32 Treatment choice, “wanted LVAD,” No. (%)c 1 mod 95 (92.2) 47 (61.0) <.001 6 moe 88 (90.7) 46 (69.7) .01 Treatment received, LVAD, No. (%)f 6 mo 110 (83.3) 54 (52.4) <.001 HeartMate IIf 6 mo 68 (61.8) 22 (40.7) .02 HeartMate 3f 6 mo 29 (26.4) 22 (40.7) .02 HVADf 6 mo 10 (9.1) 10 (18.5) .02 Decision conflict part b score (0-100), mean (SE)g BL1 20.2 (1.99) 23.4 (2.24) .28 BL2 16.5 (1.95) 18.4 (2.23) .52 1 mo 15.5 (1.89) 17.9 (2.17) .42 6 mo 15.4 (1.89) 14.2 (2.21) .67 Decision regret score (0-100), mean (SE)h 1 mo 14.3 (2.15) 17.9 (2.84) .37 6 mo 12.1 (2.28) 19.1 (2.96) .09 Control preferences scale (preferred), active role, No. (%)i 1 mod 90 (86.6) 71 (89.8) .87 6 moe 85 (86.7) 61 (92.4) .70 Control preferences scale (actual), active role, No. (%)j 1 mod 91 (87.5) 66 (83.6) .81 6 moe 84 (85.8) 59 (89.4) .95 PEACE: acceptance of illness score (5-20), mean (SE)k BL1 17.5 (0.26) 17.1 (0.31) .44 1 mo 17.4 (0.27) 17.4 (0.32) .90 6 mo 17.5 (0.28) 18.2 (0.34) .18 PEACE: struggle with illness score (7-28), mean (SE)l BL1 14.0 (0.42) 13.1 (0.50) .25 1 mo 13.6 (0.47) 12.9 (0.57) .41 6 mo 12.9 (0.50) 12.0 (0.62) .29 Perceived stress score (0-40), mean (SE)m BL1 16.1 (0.68) 14.1 (0.81) .09 6 mo 12.6 (0.82) 11.9 (1.03) .61 Patient Health Q-2 Score (0-6), mean (SE)n BL1 1.80 (0.21) 1.56 (0.24) .47 1 mo 1.64 (0.23) 1.39 (0.26) .47 6 mo 1.06 (0.21) 0.97 (0.25) .80 EuroQol visual analogue scale (0-100), mean (SE)o BL1 44.6 (2.69) 48.6 (3.07) .36 1 mo 64.3 (2.67) 60.5 (3.13) .38 6 mo 69.6 (2.57) 68.8 (3.07) .86 Abbreviations: BL1, baseline 1 survey; BL2, baseline 2 survey; PEACE, peace, equanimity, and acceptance in the cancer experience; SE, standard error; 1 mo, 1-month follow-up; 6 mo, 6-month follow-up.

le (0-100), mean (SE)o BL1 44.6 (2.69) 48.6 (3.07) .36 1 mo 64.3 (2.67) 60.5 (3.13) .38 6 mo 69.6 (2.57) 68.8 (3.07) .86 Abbreviations: BL1, baseline 1 survey; BL2, baseline 2 survey; PEACE, peace, equanimity, and acceptance in the cancer experience; SE, standard error; 1 mo, 1-month follow-up; 6 mo, 6-month follow-up. a 10-item measure assessing knowledge of DT LVAD, number of correct answers. b Likert scale of 1 to 10, with 1 being “Do everything I can to live longer, even if that means having major surgery and being dependent on a machine” and 10 being “Live with whatever time I have left, without going through major surgery or being dependent on a machine.” c Patient-reported treatment decision of “I wanted the DT LVAD and decided to get it” and “I first decided not to get the DT LVAD but then decided I wanted it.” d Missing 1-month surveys: 31 (23%) in control group and 34 (30.1%) in intervention group. e Missing 6-month surveys: 37 (27.4%) in control group and 46 (40.7%) in intervention group. f Medical record report at 6 months on patients’ treatment received, LVAD or no LVAD. g 16 Items, scoring 0 to 100 with higher score indicating greater decisional conflict. h 5 Items, scoring 0 to 100 with higher score indicating greater decision regret.

e Missing 6-month surveys: 37 (27.4%) in control group and 46 (40.7%) in intervention group. f Medical record report at 6 months on patients’ treatment received, LVAD or no LVAD. g 16 Items, scoring 0 to 100 with higher score indicating greater decisional conflict. h 5 Items, scoring 0 to 100 with higher score indicating greater decision regret. i 1 Item assessing preferred control in decision making, “active role” includes answers of “I prefer to make the final selection about which treatment I will receive,” “I prefer to make the final selection of my treatment after seriously considering my doctor’s opinion,” or “I prefer that my doctor and I share responsibility for deciding which treatment is best.” j 1 Item assessing actual control in decision making, “active role” includes answers of “I made the final selection about which treatment I would receive,” “I made the final selection of my treatment after seriously considering my doctor’s opinion,” or “My doctor and I shared responsibility for deciding which was treatment best for me.” k Questions 1 through 5 of 12 items, scoring 5 to 20 with higher score indicating greater acceptance of illness. l Questions 6 through 12 of 12 items, scoring 7 to 28 with higher score indicating greater struggle with illness. m 10 Items, scoring 0 to 40 with higher score indicating greater stress. n 2 Items, score of 0 to 6 with higher score indicating greater depression. o 1-Item scale, score of 0 to 100 with 0 being “worst imaginable health state” and 100 being “best imaginable health state.”

l Questions 6 through 12 of 12 items, scoring 7 to 28 with higher score indicating greater struggle with illness. m 10 Items, scoring 0 to 40 with higher score indicating greater stress. n 2 Items, score of 0 to 6 with higher score indicating greater depression. o 1-Item scale, score of 0 to 100 with 0 being “worst imaginable health state” and 100 being “best imaginable health state.” At 1 month, patient-reported treatment choice favored LVAD more in the control group than the intervention group: “wanted LVAD and decided to get it” 78.8% control, 54.4% intervention; “first decided not to get the DT LVAD but then decided he/she wanted it” 12.5% control, 5.1% intervention; “decided not to get LVAD” 1.0% control, 7.6% intervention (overall P < .001). Concordance between stated values and patient-reported treatment choice (eg, values score closer to 1 combined with “wanted LVAD”) at 1-month was higher in the intervention than in the control group (Kendall’s τ correlation coefficient: control 0.17, intervention 0.48 (difference in correlation control to intervention, 0.28; 95% CI, 0.05-0.45; P = .01) (Figure 2). Patient-reported treatment choices were stable from 1 month to 6 months (Table 2).

) at 1-month was higher in the intervention than in the control group (Kendall’s τ correlation coefficient: control 0.17, intervention 0.48 (difference in correlation control to intervention, 0.28; 95% CI, 0.05-0.45; P = .01) (Figure 2). Patient-reported treatment choices were stable from 1 month to 6 months (Table 2). By 6 months, 110 (83.3%) (adjusted rate, 79.9%) patients in the control group and 54 (52.4%) (adjusted rate, 53.9%) patients in the intervention group had undergone LVAD implantation (P = .008), with significant differences among sites (Figure 3). The medical team decided the patient was not eligible for a DT LVAD in 11 (8.3%) control and 25 (24.3%) intervention patients (P < .001). Concordance between stated values at 1 month and actual treatment received by 6 months (in contrast to the 1-month patient-reported treatment choice above) was the same in both the control and intervention groups (Kendall’s τ correlation coefficient: intervention, 0.26; control, 0.27; difference in correlation control to intervention: 0.01; 95% CI, −0.24 to 0.25; P = .97). Figure 3. LVAD Implant by 6 Months Abbreviation: LVAD, left ventricular assist device. Secondary Outcomes At baseline and follow up, decision conflict, decision regret, control preferences, illness acceptance, stress, depression, and quality of life were not significantly different between the control and intervention groups (Table 2).

Figure 3. LVAD Implant by 6 Months Abbreviation: LVAD, left ventricular assist device. Secondary Outcomes At baseline and follow up, decision conflict, decision regret, control preferences, illness acceptance, stress, depression, and quality of life were not significantly different between the control and intervention groups (Table 2). Sensitivity Analyses When individual sites were removed from the analysis or when the analysis was restricted only to those enrolled in the inpatient setting, the primary knowledge and values-choice concordance findings remained significantly in favor of the intervention period. In sensitivity analysis of the knowledge score that did not include the site random effect, our findings remained the same. Missing data can be viewed in Supplement 2. Discussion The DECIDE-LVAD trial offers unique insights into one of the most challenging medical decisions created by modern medicine. Rather than test the efficacy of decision support tools administered by research personnel in a patient-randomized trial, this study used a pragmatic effectiveness design to assess how programmatic integration of decision aids and clinician training into standard processes of care may influence DT LVAD decision making. Through its conduct, DECIDE-LVAD created one of the largest prospective LVAD-eligible cohorts, enrolling most of the patients considered for DT LVAD during the study period, from 6 geographically diverse sites, and nearly a quarter from intensive care. In this context, the intervention was associated with better decision quality.

conduct, DECIDE-LVAD created one of the largest prospective LVAD-eligible cohorts, enrolling most of the patients considered for DT LVAD during the study period, from 6 geographically diverse sites, and nearly a quarter from intensive care. In this context, the intervention was associated with better decision quality. Decision support studies most commonly measure knowledge and decision conflict as the primary outcomes. However, central to a high-quality decision is whether the choice matches the patient’s values, goals, and preferences. Knowledge alone is insufficient to guarantee high-quality decision making, particularly for emotionally charged decisions around life and death. Similarly, nudging patients in fear and denial to address life-threatening situations—rather than providing them with false reassurances—may transiently increase feelings of conflict, anxiety, and even regret. Thus, it is not surprising to us that DECIDE-LVAD intervention did not reduce conflict at 1 month; we predicted this a priori. In contrast, we leveraged the dominant value that emerges when considering DT LVAD (ie, aggressive care to optimize survival chances vs not) and found that the intervention was associated with improved concordance between stated values and patient-reported treatment choice. This did not translate to improved concordance between stated values and actual treatment received, perhaps because DT LVAD implantation is influenced by a wide range of factors (eg, medical eligibility, presence of an adequate caregiver), many of which are not in the patient’s control.

tient-reported treatment choice. This did not translate to improved concordance between stated values and actual treatment received, perhaps because DT LVAD implantation is influenced by a wide range of factors (eg, medical eligibility, presence of an adequate caregiver), many of which are not in the patient’s control. The site-based intervention with a randomized stepped wedge roll out distinguishes the DECIDE-LVAD study from most other assessments of shared decision making. Despite strong efficacy data, uptake of decision aids in routine practice has been slow. To be successful, decision support tools must integrate easily into existing care and facilitate necessary discussions. Leveraging the formal consent and education process for LVAD, we were able to implement the DECIDE-LVAD intervention into this existing structure, while observing the programmatic transitions in all 6 sites over time. Given the importance of widespread adoption of shared decision making, pragmatic studies such as this one are needed to address real-world complexities and implementation challenges. The decrease in device implantation rates from control to intervention in 5 of the 6 sites supports the ability to influence institutional culture and decision-making processes related to major medical interventions.

c studies such as this one are needed to address real-world complexities and implementation challenges. The decrease in device implantation rates from control to intervention in 5 of the 6 sites supports the ability to influence institutional culture and decision-making processes related to major medical interventions. Unlike most prospective studies in MCS that follow patients from the time of implant, the DECIDE-LVAD study moved upstream to follow the entire population of patients formally considered for DT LVAD. By focusing on a choice rather than a specific therapy (ie, only those who have received an LVAD), DECIDE-LVAD expands insights into the patients, experiences, and processes that lead up to decisions about device implantation. We found that a significant number of patients facing high morbidity and mortality from heart failure decline DT LVAD in favor of avoiding aggressive therapies. The declination rates reported here are among patients who have agreed to undergo formal evaluation; thus, we suspect that DT LVAD declination may be more prevalent in the broader community. This is concordant with prior work showing diversity in the relative emphasis patients place on quality vs quantity of life, even when actively facing life-threatening illness. It also reinforces DT LVAD as a relatively preference-sensitive decision.

t DT LVAD declination may be more prevalent in the broader community. This is concordant with prior work showing diversity in the relative emphasis patients place on quality vs quantity of life, even when actively facing life-threatening illness. It also reinforces DT LVAD as a relatively preference-sensitive decision. Limitations A number of limitations should be recognized. First, missing data were somewhat frequent and concentrated among the group of patients who did not undergo implantation of DT LVAD. Death was the most common cause of missing data, followed by withdrawal from the study, both of which are common in studies targeting patients with life threatening illness. Our missing data rates are comparable to similar study types, and our models adjusted for missing data as much as possible. Second, time trends in rapidly evolving fields are particularly problematic for the stepped wedge design. Fortunately, device technology was relatively stable between 2015 and 2017 and durable LVAD implant rates across the US plateaued somewhat during this period. Third, the phased implementation randomized the site with the lowest LVAD implant rate to spend the most time in the intervention and the site with the highest LVAD implant rate to spend the most time in the control. Linear mixed models accounting for site effects and sensitivity analyses were used to explore and diminish these differences for the patient-based effectiveness outcomes, but do not necessarily fully account for such effects. Finally, the population was mostly white males. Although enrollees were 12% more likely to be white than those excluded, the final study cohort reflects contemporary use of LVADs in the United States. This bias makes it difficult to extrapolate the findings here to decision making for women and underrepresented races/ethnicities.

e population was mostly white males. Although enrollees were 12% more likely to be white than those excluded, the final study cohort reflects contemporary use of LVADs in the United States. This bias makes it difficult to extrapolate the findings here to decision making for women and underrepresented races/ethnicities. Conclusions A shared decision-making intervention for patients considering DT LVAD—implemented programmatically, integrating novel patient decision aids, and including clinician training—was associated with a significant improvement in knowledge and an increase in concordance between stated values and patient-reported treatment choice. Although these changes did not translate to improvements in concordance between stated values and actual treatment received, patients were less likely during intervention than control to proceed to LVAD implant. These results suggest that institutional culture and processes can influence medical decisions in life-threatening illness. Supplement 1. Trial Protocol Click here for additional data file. Supplement 2. eAppendix 1. Pamphlet. A decision aid for Left Ventricular Assist Device (LVAD) as Destination Therapy eAppendix 2. Sample Surveys. Search strategy for EMBASE (using Embase.com) eAppendix 3. Missing Data. Flow of information through the different phases of the review eTable. Participant Characteristics by Missing Primary Knowledge (at Baseline 1 and Baseline 2) and Values (at 1-Month) Outcomes Click here for additional data file.

We have recently shown that hydrochlorothiazide use increases the risk of lip and nonmelanoma skin cancer, notably squamous cell carcinoma.1,2 It would have substantial implications if the carcinogenic effect of hydrochlorothiazide also extended to malignant melanoma. Methods Similarly to our recent studies of hydrochlorothiazide,1,2 we identified histologically verified melanoma cases (January 2004 to December 2015), each matched 1:10 (risk-set sampling; age, sex, and date) to cancer-free population controls. We required cases and controls to be between ages 18 and 90 years, without previous history of cancer (except nonmelanoma skin cancer), organ transplantation, HIV infection, or azathioprine use, and to have resided continuously in Denmark for 10 years. Using conditional logistic regression, we calculated odds ratios (ORs), with 95% CIs, for melanoma associated with cumulative hydrochlorothiazide use compared with never-use, adjusting for potential confounders (Table 1 and 2). We performed stratified analyses by localization, stage, histologic subtype, and subgroups of age, sex, and history of nonmelanoma skin cancer. To evaluate potential confounding by indication, we performed analyses for other antihypertensive drugs, including bendroflumethiazide, angiotensin-converting enzyme inhibitors, angiontensin-II receptor antagonists, and calcium-channel blockers. This study was approved by Statistics Denmark and the Danish Data Protection Agency.

e potential confounding by indication, we performed analyses for other antihypertensive drugs, including bendroflumethiazide, angiotensin-converting enzyme inhibitors, angiontensin-II receptor antagonists, and calcium-channel blockers. This study was approved by Statistics Denmark and the Danish Data Protection Agency. Table 1. Association of Exposure to Hydrochlorothiazide and Risk of Malignant Melanoma Use and Dose Cases, No. (n = 19 273) Controls, No. (n = 192 730) Adjusted OR (95% CI) Model 1a Model 2b Hydrochlorothiazide Never used 17 315 175 486 1 [Reference] 1 [Reference] Ever used 1958 17 244 1.16 (1.11-1.22) 1.17 (1.11-1.23) High use (≥50 000 mg) 413 3406 1.24 (1.12-1.38) 1.22 (1.09-1.36) Cumulative dose 1-24 999 mg 1160 10 483 1.14 (1.07-1.21) 1.14 (1.07-1.22) 25 000-49 999 mg 385 3355 1.18 (1.06-1.32) 1.18 (1.05-1.32) 50 000-99 999 mg 219 1852 1.22 (1.06-1.41) 1.21 (1.05-1.40) ≥100 000 mg 194 1554 1.26 (1.08-1.46) 1.21 (1.04-1.42) Test for trend 1958 17 244 P = .24 P = .42 a Adjusted for age, sex, and calendar time (by use of risk-set matching and conditional analysis).

22) 25 000-49 999 mg 385 3355 1.18 (1.06-1.32) 1.18 (1.05-1.32) 50 000-99 999 mg 219 1852 1.22 (1.06-1.41) 1.21 (1.05-1.40) ≥100 000 mg 194 1554 1.26 (1.08-1.46) 1.21 (1.04-1.42) Test for trend 1958 17 244 P = .24 P = .42 a Adjusted for age, sex, and calendar time (by use of risk-set matching and conditional analysis). b Fully adjusted model additionally adjusted for history of nonmelanoma skin cancer, other comorbidity (diabetes, chronic obstructive pulmonary disease, alcohol abuse–associated disorders, chronic renal failure), Charlson Comorbidity Index score (0, low; 1-2, medium; or ≥3, high), use of certain drugs (topical retinoids, oral retinoids, tetracycline, macrolides, aminoquinolines, amiodarone, methoxypsoralen, low-dose aspirin, nonaspirin nonsteroidal anti-inflammatory drugs, statins, or oral steroids), and highest achieved education (short, medium, long, or unknown).

ow; 1-2, medium; or ≥3, high), use of certain drugs (topical retinoids, oral retinoids, tetracycline, macrolides, aminoquinolines, amiodarone, methoxypsoralen, low-dose aspirin, nonaspirin nonsteroidal anti-inflammatory drugs, statins, or oral steroids), and highest achieved education (short, medium, long, or unknown). Table 2. Association of Exposure to Hydrochlorothiazide and Risk of Malignant Melanoma According to Amount of Hydrochlorothiazide Use and Specified by Melanoma Subtype Melanoma Cases. No. Controls, No. Adjusted OR (95% CI) Model 1a Model 2b Superficial Spreading Melanoma Never used 12 494 126 216 1 [Reference] 1 [Reference] Ever used 1287 11 594 1.13 (1.06-1.20) 1.13 (1.06-1.20) High use (≥50 000 mg) 254 2268 1.14 (0.99-1.30) 1.11 (0.97-1.27) Cumulative dose 1-24 999 mg 783 7023 1.14 (1.05-1.23) 1.13 (1.05-1.23) 25 000-49 999 mg 250 2303 1.11 (0.97-1.27) 1.10 (0.96-1.27) 50 000-99 999 mg 140 1252 1.15 (0.96-1.38) 1.14 (0.95-1.37) ≥100 000 mg 114 1016 1.12 (0.91-1.36) 1.06 (0.87-1.30) Test for trend 1287 11 594 P = .94 P = .73 Nodular Melanoma Never used 1465 15 108 1 [Reference] 1 [Reference] Ever used 230 1842 1.31 (1.12-1.53) 1.28 (1.09-1.49) High use (≥50 000 mg) 68 351 2.13 (1.61-2.80) 2.05 (1.54-2.72) Cumulative dose 1-24 999 mg 119 1142 1.08 (0.88-1.32) 1.05 (0.86-1.29) 25 000-49 999 mg 43 349 1.24 (0.90-1.72) 1.17 (0.84-1.64) 50 000-99 999 mg 34 195 1.90 (1.30-2.78) 1.81 (1.23-2.67) ≥100 000 mg 34 156 2.34 (1.59-3.45) 2.26 (1.52-3.36) Test for trend 230 1842 P = .01 P = .01 Lentigo Melanoma Never used 386 4198 1 [Reference] 1 [Reference] Ever used 114 802 1.57 (1.25-1.97) 1.58 (1.25-2.00) High use (≥50 000 mg) 28 177 1.72 (1.13-2.62) 1.61 (1.03-2.50) Cumulative dose 1-24 999 mg 58 476 1.32 (0.98-1.78) 1.35 (0.99-1.83) 25 000-49 999 mg 28 149 2.22 (1.44-3.43) 2.30 (1.46-3.60) 50 000-99 999 mg 11 99 1.25 (0.66-2.38) 1.09 (0.56-2.11) ≥100 000 mg 17 78 2.26 (1.30-3.91) 2.24 (1.25-3.99) Test for trend 114 802 P = .11 P = .16 a Adjusted for age, sex, and calendar time (by use of risk-set matching and conditional analysis).

.35 (0.99-1.83) 25 000-49 999 mg 28 149 2.22 (1.44-3.43) 2.30 (1.46-3.60) 50 000-99 999 mg 11 99 1.25 (0.66-2.38) 1.09 (0.56-2.11) ≥100 000 mg 17 78 2.26 (1.30-3.91) 2.24 (1.25-3.99) Test for trend 114 802 P = .11 P = .16 a Adjusted for age, sex, and calendar time (by use of risk-set matching and conditional analysis). b Fully adjusted model additionally adjusted for history of nonmelanoma skin cancer, other comorbidity (diabetes, chronic obstructive pulmonary disease, alcohol abuse associated disorders, chronic renal failure), Charlson Comorbidity Index score (0, low; 1-2, medium; or ≥3, high), use of certain drugs (topical retinoids, oral retinoids, tetracycline, macrolides, aminoquinolines, amiodarone, methoxypsoralen, low-dose aspirin, nonaspirin nonsteroidal anti-inflammatory drugs, statins, or oral steroids), and highest achieved education (short, medium, long, or unknown). Results We identified 22 010 cases of melanoma. After exclusions, the final study population comprised 19 273 cases and 192 730 population controls. Cases had slightly lower comorbidity, higher educational level, and higher prevalence of previous nonmelanoma skin cancer than controls. Remaining characteristics were similar between cases and controls.

es of melanoma. After exclusions, the final study population comprised 19 273 cases and 192 730 population controls. Cases had slightly lower comorbidity, higher educational level, and higher prevalence of previous nonmelanoma skin cancer than controls. Remaining characteristics were similar between cases and controls. Overall, 413 cases (2.1%) and 3406 controls (1.8%) were classified as high-users (≥50 000 mg) of hydrochlorothiazide, yielding an adjusted OR of 1.22 (95% CI, 1.09-1.36) for melanoma. No clear dose-response pattern emerged between hydrochlorothiazide use and melanoma risk (Table 1). Analyses by melanoma localization, stage, age, sex, and history of nonmelanoma skin cancer yielded results comparable to the main analysis (data not shown). When stratifying by histological subtype (Table 2), higher ORs occurred for nodular melanoma (n = 1695 cases [8.8%]; OR, 2.05; 95% CI, 1.54-2.72; P for trend = .01) and lentigo melanoma (n = 500 cases [2.6%]; OR, 1.61; 95% CI, 1.03-2.50; P for trend = .16) than for superficial spreading melanoma (n = 13 781 cases [72%]; OR, 1.11; 95% CI, 0.97-1.27; P for trend = .73).

, higher ORs occurred for nodular melanoma (n = 1695 cases [8.8%]; OR, 2.05; 95% CI, 1.54-2.72; P for trend = .01) and lentigo melanoma (n = 500 cases [2.6%]; OR, 1.61; 95% CI, 1.03-2.50; P for trend = .16) than for superficial spreading melanoma (n = 13 781 cases [72%]; OR, 1.11; 95% CI, 0.97-1.27; P for trend = .73). In secondary analyses, we observed associations close to the null for overall melanoma risk with long-term use of bendroflumethiazide (OR, 1.10; 95% CI, 1.02-1.19; P for trend = 0.47), angiotensin-converting enzyme inhibitors (OR, 1.07; 95% CI, 0.99-1.16; P for trend = .53), angiotensin-II receptor antagonists (OR, 1.18; 95% CI, 1.07-1.29; P for trend = .07), and calcium-channel blockers (OR, 1.06; 95% CI, 0.97-1.14; P for trend  = .94). These associations remained neutral in subanalyses stratified by melanoma subtype (data not shown). Discussion The main strength of our study is the use of high-quality nationwide registry data.3 The main limitations are a lack of information on risk factors such as sun exposure, skin pigmentation, and family history of melanoma. However, these characteristics are unlikely to be substantially associated with hydrochlorothiazide use, and thus unlikely to confound out estimates.

quality nationwide registry data.3 The main limitations are a lack of information on risk factors such as sun exposure, skin pigmentation, and family history of melanoma. However, these characteristics are unlikely to be substantially associated with hydrochlorothiazide use, and thus unlikely to confound out estimates. Thiazide use and melanoma risk has been investigated in a few previous studies; however, only 2 studies,4,5 both from northern Denmark, have specifically examined hydrochlorothiazide. The first study reported an OR of 1.32 (95% CI, 1.03-1.70) for melanoma risk overall associated with 10 000 mg increments of hydrochlorothiazide.4 The corresponding OR for hydrochlorothiazide in combination with amiloride was 1.43 (95% CI, 1.09-1.88).4 The other study found no association between hydrochlorothiazide use combined with amiloride and melanoma risk (OR, 1.02; 95% CI, 0.78-1.33).5 Neither of these studies included dose-response or histology-specific analyses. The findings for melanoma subtype are somewhat surprising, as lentigo and superficial spreading melanoma are known to be associated with high sun exposure, whereas the etiology of nodular melanomas is less elucidated.6 It is worrying that hydrochlorothiazide use appears to be associated with an increased risk of melanoma, and the particular associations observed for lentigo melanoma and nodular melanoma warrant further research.

Introduction Indwelling urethral catheters are a mainstay in the care of hospitalized patients. An estimated 30 million indwelling urethral (ie, Foley) catheters are sold annually in the United States,1 and approximately 20% of hospitalized patients have a urethral catheter at any given time.2,3 While indwelling urethral catheters are important for the care of certain hospitalized patients (eg, monitoring hourly urine output in critically ill patients),4 they are invasive medical devices and, as such, are a potential threat to patient safety. The most commonly recognized complication is catheter-associated urinary tract infection.3,5 Although infectious complications associated with indwelling urethral catheters are well described, few studies have examined the noninfectious complications associated with their use, and Foley catheters lead to several patient safety problems.6 For example, a study by Leuck et al7 identified 100 instances of Foley catheter–associated trauma in 6513 catheter-days during a period of 16 months. Using administrative data, Aaronson and colleagues8 found that urethral catheter–associated complications were linked to an increased length of stay and an increased risk of urinary tract infection among surgical patients.8 Finally, Hollingsworth and colleagues9 reported in their systematic review of 37 studies that noninfectious urethral catheter–associated complications were as common as clinically significant infections.

ere linked to an increased length of stay and an increased risk of urinary tract infection among surgical patients.8 Finally, Hollingsworth and colleagues9 reported in their systematic review of 37 studies that noninfectious urethral catheter–associated complications were as common as clinically significant infections. Several limitations exist with the previous studies. For example, most were single-site investigations, did not collect data after catheter removal, and often focused on specific patient populations (eg, men, surgical patients, or intensive care unit patients),7,10,11 limiting their generalizability. The data sources used are also problematic, with studies using urology consultations likely to overestimate complication rates,10,11 whereas those using diagnosis codes8 are likely to underestimate complication frequency. Finally, few studies to date have included the perspective of the patient, even though patients can be a reliable source of information about adverse events.12 We thus conducted a prospective study to measure the incidence of all complications associated with the indwelling urethral catheter and to capture patient insights about how the use of this device affects their well-being and safety.

even though patients can be a reliable source of information about adverse events.12 We thus conducted a prospective study to measure the incidence of all complications associated with the indwelling urethral catheter and to capture patient insights about how the use of this device affects their well-being and safety. Methods Study Overview We conducted a prospective cohort study at 4 US hospitals in which hospitalized patients with an indwelling urethral catheter were identified and followed up for 30 days after its insertion, even if the catheter had been removed during that time. Data about both infectious and noninfectious catheter–associated complications during this 30-day period were collected through patient interviews. This process included an in-person interview with the patient within 3 days after placement of the catheter and follow-up assessments after 14 days and 30 days, either in-person or through brief telephone interviews. The 4 study sites were the University of Michigan Medical Center and Veterans Affairs Ann Arbor Healthcare System, both in Ann Arbor, Michigan, and Ben Taub Hospital and the Michael E. DeBakey Veterans Affairs Medical Center, both in Houston, Texas. The primary objectives were to determine rates and types of patient-reported infectious and noninfectious complications associated with urethral catheters. Institutional review board approval was received at each participating facility. All participants provided written informed consent.

Center, both in Houston, Texas. The primary objectives were to determine rates and types of patient-reported infectious and noninfectious complications associated with urethral catheters. Institutional review board approval was received at each participating facility. All participants provided written informed consent. Data Collection and Inclusion Criteria Each weekday morning (Monday through Friday) from August 26, 2015, to August 18, 2017, research staff identified patients on selected inpatient units who, during the previous 3 days, had placement of a urethral catheter. Patient eligibility criteria were (1) hospitalization in an acute care unit, including medical/surgical wards, and intensive care and progressive care units; (2) placement of a urethral catheter for the first time during this hospital stay; (3) the urethral catheter had been in place for no longer than 3 days; (4) patients were at least 18 years of age; and (5) patients were able to speak either English or Spanish. Exclusion criteria were (1) inability to provide self-consent or participate in the interview/assessment process (eg, obtundation, severe dementia, or delirium); (2) refusal to provide written informed consent to participate; or (3) previous participation in this study. After identifying potentially eligible patients by using data from the electronic medical record system, study staff conducted bedside visits to begin the recruitment process. After confirming eligibility, patients were invited to participate regardless of whether the urethral catheter was still in place at the time of recruitment. Once they were recruited, patient follow-up continued for 30 days from the initial date of catheter insertion. Participants received a souvenir magnet as a thank-you gift for their participation.

ients were invited to participate regardless of whether the urethral catheter was still in place at the time of recruitment. Once they were recruited, patient follow-up continued for 30 days from the initial date of catheter insertion. Participants received a souvenir magnet as a thank-you gift for their participation. Information about patient characteristics as well as infectious and noninfectious complications associated with the urethral catheter were collected directly from patients. After providing written consent, an initial in-person patient interview was conducted that included questions about why the urethral catheter was placed. The American Urological Association (AUA) Symptom Index was used to identify any urologic symptom in the previous month (ie, precatheterization).13 Follow-up assessments were conducted approximately 14 days and 30 days after catheter insertion. The follow-up assessments, which asked patients about their symptoms and experiences during the previous 2 weeks, were conducted in person if the patient was still hospitalized. After discharge, a study team member contacted the patient by telephone. The patient assessment was designed to elicit the patient’s perspectives about the catheter and to identify complications, such as pain, that may not be well documented in the medical record. Indeed, patient assessments may be the only way to identify complications that occur after hospital discharge, particularly complications that do not lead to a medical visit or require intervention. Questions were primarily closed ended, except for a concluding question that allowed patients to discuss other possible complications.

nt assessments may be the only way to identify complications that occur after hospital discharge, particularly complications that do not lead to a medical visit or require intervention. Questions were primarily closed ended, except for a concluding question that allowed patients to discuss other possible complications. Study Measures Primary outcomes were infectious and noninfectious complication events associated with the urethral catheter. Infectious complications for participants included being told that they had a urinary tract infection or a positive endorsement of any of the following symptoms in the previous 2 weeks: fever, chills, burning with urination, urinary frequency, urinary urgency, or other symptoms suggestive of an infection that required the patient to see a physician. Noninfectious complications for participants whose catheters had been removed included a sense of urgency or bladder spasms, blood in the urine, leaking urine, and difficulty with starting or stopping the urine stream. For those with a catheter still in place, noninfectious complications included pain or discomfort, a sense of urgency or bladder spasms, blood in the urine, and trauma to the skin associated with catheter placement or securement. A full list of complications by category is provided in eTable 1 in the Supplement. Secondary outcomes of interest focused on patient perspectives about their catheters, such as their effect on activities of daily living, social activities, and the general level of comfort.

ciated with catheter placement or securement. A full list of complications by category is provided in eTable 1 in the Supplement. Secondary outcomes of interest focused on patient perspectives about their catheters, such as their effect on activities of daily living, social activities, and the general level of comfort. Statistical Analysis We calculated general descriptive statistics for all variables of interest, including baseline patient characteristics (number with percentage or mean [SD]) and outcomes of interest (number with percentage and 95% CI). The primary outcome was the percentage of patients experiencing a complication from a urethral catheter at any time, which was calculated for select individual complications as well as by group (infectious vs noninfectious). The frequency of complications at each site was also assessed. The χ2 test was used to compare complication percentages by certain demographic and clinical characteristics. These characteristics included age, sex, indication for catheter placement (perioperative, urinary retention or bladder obstruction, or other), AUA symptom index score prior to catheter use (range, 1-7 [mild]; 8-19 [moderate]; or 20-35 [severe]),13 and catheter duration (≤3 days or >3 days). To test whether differences between the sexes persisted after accounting for age, AUA symptom index score, reason for placement, and catheter duration, we used multivariable Poisson regression with robust SEs.14

r use (range, 1-7 [mild]; 8-19 [moderate]; or 20-35 [severe]),13 and catheter duration (≤3 days or >3 days). To test whether differences between the sexes persisted after accounting for age, AUA symptom index score, reason for placement, and catheter duration, we used multivariable Poisson regression with robust SEs.14 All analyses were performed using SAS, version 9.4 (SAS Institute Inc) or Stata, version 15 (StataCorp). All statistical tests were 2-sided with α set to .05. Results Of 2967 eligible patients at 4 study sites, 2227 (75.1%) agreed to participate and 2076 patients were evaluated (Figure 1). The majority of patients (1482 of 2076 [71.4%]) were male; mean (SD) age was 60.8 (13.4) years. Participants who were missing a baseline assessment or both follow-up assessments were excluded (151 of 2227 patients [6.8%]), resulting in 2076 patients included in this analysis. Patients were recorded in both the group with a catheter in place and the group with the catheter removed if they had an indwelling catheter at 1 follow-up evaluation, but not at the other follow-up evaluation. Figure 1. Patient Flowchart of Study Enrollment and Follow-up Of the 2967 patients who met eligibility criteria, 2227 (75.1%) agreed to participate. The 4 participating study sites were the University of Michigan Medical Center and Veterans Affairs Ann Arbor Healthcare System, both in Ann Arbor, Michigan, and Ben Taub Hospital and the Michael E. DeBakey Veterans Affairs Medical Center, both in Houston, Texas.

ligibility criteria, 2227 (75.1%) agreed to participate. The 4 participating study sites were the University of Michigan Medical Center and Veterans Affairs Ann Arbor Healthcare System, both in Ann Arbor, Michigan, and Ben Taub Hospital and the Michael E. DeBakey Veterans Affairs Medical Center, both in Houston, Texas. Table 1 describes the patient demographics. Only 594 patients (28.6%) were women because the 2 Veterans Affairs medical centers included in the study provide care for a predominately male population. Urethral catheters were initially placed before surgical procedures in 1653 of 2076 participants (79.6%). Most catheters were placed for short durations, with 1578 patients (76.0%) having them removed within 3 days of insertion. Only 164 patients (7.9%) reported experiencing pain, discomfort, bleeding, or trauma during catheter placement. A significantly lower percentage of patients (33 of 1643 [2.0%]) who had the urethral catheter inserted for a surgical procedure reported complications with insertion, compared with 83 of 144 (57.6%) who had a catheter inserted for urinary retention or bladder obstruction (P < .001; eTable 2 in the Supplement). In addition, 414 of 1340 patients (30.9%) whose catheter had been removed prior to enrollment reported experiencing those complications during removal (Table 1).

th insertion, compared with 83 of 144 (57.6%) who had a catheter inserted for urinary retention or bladder obstruction (P < .001; eTable 2 in the Supplement). In addition, 414 of 1340 patients (30.9%) whose catheter had been removed prior to enrollment reported experiencing those complications during removal (Table 1). Table 1. Demographic and Baseline Characteristics of 2076 Study Participants Characteristic No. (%) Age, mean (SD), y 60.8 (13.4) Sex Male 1482 (71.4) Female 594 (28.6) Race/ethnicity White 1631 (78.6) African American 290 (14.0) American Indian or Alaskan Native 27 (1.3) Other or unknowna 128 (6.2) Hispanic 147 (7.1) Reason for urethral catheter placement Perioperative use for surgical procedure 1653 (79.6) Urinary retention or bladder obstruction 144 (6.9) Accurate measurement of urine output 30 (1.4) Patient convenience or incontinence 31 (1.5) Required prolonged immobilization 25 (1.2) Other or unknown 193 (9.3) Experienced pain, discomfort, bleeding, or other trauma during urinary catheter insertion 164 (7.9) Experienced pain, discomfort, bleeding, or other trauma during urinary catheter removal for participants whose catheter was already removed (n = 1340) 414 (30.9) Duration of urinary catheter use ≤3 d 1578 (76.0) a Asian (11 participants [0.5%]), Native Hawaiian or other Pacific Islander (10 [0.5%]), more than 1 race/ethnicity (64 [3.1%]), and unknown (43 [2.1%]).

her trauma during urinary catheter removal for participants whose catheter was already removed (n = 1340) 414 (30.9) Duration of urinary catheter use ≤3 d 1578 (76.0) a Asian (11 participants [0.5%]), Native Hawaiian or other Pacific Islander (10 [0.5%]), more than 1 race/ethnicity (64 [3.1%]), and unknown (43 [2.1%]). During the 30 days after a urethral catheter was inserted, 1184 of 2076 patients (57.0%; 95% CI, 54.9%-59.2%) reported at least 1 complication because of the indwelling urethral catheter. As shown in Table 2, noninfectious complications were 5 times as prevalent as infectious complications (noninfectious, 55.4%; 95% CI, 53.2%-57.6% vs infectious, 10.5%; 95% CI, 9.3%-12.0%; P < .001). Women were more likely than men to report an infectious complication (92 of 594 women [15.5%] vs 127 of 1482 men [8.6%]; P < .001), whereas men were more likely than women to report a noninfectious complication (869 of 1482 men [58.6%] vs 281 of 594 women [47.3%]; P < .001) as depicted in Figure 2.

3%-12.0%; P < .001). Women were more likely than men to report an infectious complication (92 of 594 women [15.5%] vs 127 of 1482 men [8.6%]; P < .001), whereas men were more likely than women to report a noninfectious complication (869 of 1482 men [58.6%] vs 281 of 594 women [47.3%]; P < .001) as depicted in Figure 2. Table 2. Specific Patient-Reported Complications Associated With Urethral Catheter Use During the Month After Insertiona Specific Complication No. (%) Catheter in Place (n = 124) Catheter Removed (n = 2034) Total (N = 2076)b Infectious complication 19 (15.3) 205 (10.1) 219 (10.5) Fevers, chills, burning with urination, urinary frequency, urinary urgency, or other symptoms suggestive of an infection that required you to see a physician 12 (9.7) 162 (8.0) 173 (8.3) Told you have a urinary tract infection 16 (13.0) 106 (5.2) 118 (5.7) Noninfectious complication 87 (70.2) 1106 (54.4) 1150 (55.4) Pain or discomfort 67 (54.5) NA NA A sense of urgency or bladder spasms 43 (34.7) 487 (24.0) 523 (25.2) Blood in the urine 34 (27.4) 179 (8.8) 207 (10.0) Trauma to your skin related to catheter securement or catheter placement 24 (19.4) NA NA Leaking urine NA 413 (20.3) NA Difficulty with starting or stopping your urine stream NA 395 (19.5) NA Pain or burning when you urinate NA 353 (17.4) NA Split stream of urine NA 245 (12.1) NA Spraying of urine stream NA 187 (9.2) NA Skin problems in the genital area NA 134 (6.6) NA Bleeding from where the urinary catheter entered or was attached to your body, or other type of discharge NA 94 (4.6) NA New urinary tract symptom NA 69 (3.4) NA Bladder or kidney stones NA 59 (2.9) NA Newly diagnosed urethral stricture disease NA 4 (0.2) NA Other complications 66 (53.2) 99 (4.9) 160 (7.7) Restrictions in activities of daily living associated with having the catheter 49 (39.5) NA NA Restrictions in social activities associated with having the catheter 54 (43.9) NA NA Sexual problems NA 99 (4.9) NA Mechanical or equipment issues with the catheter or securement device, eg, leaking, issues with leg band 16 (12.9) NA NA Abbreviation: NA, not applicable.

ociated with having the catheter 49 (39.5) NA NA Restrictions in social activities associated with having the catheter 54 (43.9) NA NA Sexual problems NA 99 (4.9) NA Mechanical or equipment issues with the catheter or securement device, eg, leaking, issues with leg band 16 (12.9) NA NA Abbreviation: NA, not applicable. a Patients could report more than 1 complication. b Patients who had a catheter in place at one follow-up evaluation but not the other follow-up evaluation contributed data to both the catheter in place and the catheter removed columns of this table. The column totals do not equal the total sample size. Figure 2. Percentage of 2076 Patients Reporting Infectious or Noninfectious Complications During the Month After Urethral Catheter Insertion Statistically significant differences by sex were found in both infectious and noninfectious patient-reported complications (P < .001).

b Patients who had a catheter in place at one follow-up evaluation but not the other follow-up evaluation contributed data to both the catheter in place and the catheter removed columns of this table. The column totals do not equal the total sample size. Figure 2. Percentage of 2076 Patients Reporting Infectious or Noninfectious Complications During the Month After Urethral Catheter Insertion Statistically significant differences by sex were found in both infectious and noninfectious patient-reported complications (P < .001). Table 2 shows the percentage of specific patient-reported complications during the month after urethral catheter insertion. Both infectious and noninfectious complications were reported more frequently by patients who still had their catheter. For 2034 patients who had had their catheter removed, the most frequently cited noninfectious complications were leaking urine (413 of 2034 patients [20.3%]; 95% CI, 18.6%-22.2%), feeling a sense of urgency or bladder spasms (487 [24.0%]; 95% CI, 22.2%-25.9%), and difficulty with starting or stopping the urine stream (395 [19.5%]; 95% CI, 17.8%-21.3%). The most common noninfectious complication cited by the 124 patients who still had a catheter in place was pain or discomfort (67 of 124 patients [54.5%]; 95% CI, 45.3%-63.5%), a sense of urgency or bladder spasms (43 [34.7%]; 95% CI, 26.4%-43.8%), and blood in the urine (34 [27.4%]; 95% CI, 19.8%-36.2%). In addition to infectious and noninfectious complications, many of the 124 patients who still had a catheter reported activities of daily living restrictions (49 of 124 [39.5%]; 95% CI, 30.9%-48.7%) and social activity restrictions (54 of 124 [43.9%]; 95% CI, 35.0%-53.1%). Of the 124 patients, 16 (12.9%) with a catheter in place reported mechanical or equipment issues with the catheter or securement device. Sexual problems were reported by 99 of 2034 patients (4.9%) who had a catheter that had been removed (95% CI, 4.0%-5.9%).

tivity restrictions (54 of 124 [43.9%]; 95% CI, 35.0%-53.1%). Of the 124 patients, 16 (12.9%) with a catheter in place reported mechanical or equipment issues with the catheter or securement device. Sexual problems were reported by 99 of 2034 patients (4.9%) who had a catheter that had been removed (95% CI, 4.0%-5.9%). Select clinical characteristics associated with patient-reported infectious and noninfectious complications are shown in Table 3. For example, a higher percentage of patients with a baseline AUA symptom index score in the severe range reported both infectious complications (37 of 213 patients [17.4%]; P < .001) and noninfectious complications (175 of 213 [82.2%]; P < .001) than those with mild or moderate scores (mild score: infectious, 98 of 1176 [8.3%] and noninfectious, 519 [44.1%]; moderate score: infectious, 79 of 670 [11.8%] and noninfectious, 442 [66.0%]). Longer duration of urethral catheter use was also associated with more reported infectious and noninfectious complications. Medical or surgical indication for catheter placement was not significantly associated with infectious complications. Noninfectious complications, however, were reported by a higher percentage of patients with a catheter inserted due to urinary retention or bladder obstruction. In a multivariable analysis adjusting for demographic and clinical characteristics (eTable 3 in the Supplement), the sex-related difference remained significant only for infectious complications (incident rate ratio [IRR] for female patients, 2.11; 95% CI, 1.59-2.79; P < .001). Moderate and severe AUA symptom index scores and urinary catheter duration of more than 3 days remained significantly associated with both infectious and noninfectious complications (eTable 3 in the Supplement).

infectious complications (incident rate ratio [IRR] for female patients, 2.11; 95% CI, 1.59-2.79; P < .001). Moderate and severe AUA symptom index scores and urinary catheter duration of more than 3 days remained significantly associated with both infectious and noninfectious complications (eTable 3 in the Supplement). Table 3. Clinical Characteristics Associated With Infectious and Noninfectious Indwelling Urethral Catheter Complications Clinical Characteristic Infectious Complication, No. (%) P Value Noninfectious Complication, No. (%) P Value Age, mean (SD), y 60.7 (13.6)a .87 62.4 (13.0)b <.001 Reason for urethral catheter placement, No. (%) Perioperative use for surgical procedure (n = 1653) 174 (10.5) .29 880 (53.2) <.001 Urinary retention or bladder obstruction (n = 144) 20 (13.9) 104 (72.2) Other or unknown (n = 279) 25 (9.0) 166 (59.5) AUA Symptom Index score, No. (%) Mild (n = 1176) 98 (8.3) <.001 519 (44.1) <.001 Moderate (n = 670) 79 (11.8) 442 (66.0) Severe (n = 213) 37 (17.4) 175 (82.2) Duration of urinary catheter, No. (%) ≤3 d (n = 1578) 154 (9.8) .04 796 (50.4) <.001 >3 d (n = 498) 65 (13.1) 354 (71.1) Abbreviation: AUA, American Urological Association. a Comparison group: no infectious complications; mean (SD) age, 60.8 (13.4) years. b Comparison group: no noninfectious complications; mean (SD) age, 58.8 (13.8) years.

Table 3. Clinical Characteristics Associated With Infectious and Noninfectious Indwelling Urethral Catheter Complications Clinical Characteristic Infectious Complication, No. (%) P Value Noninfectious Complication, No. (%) P Value Age, mean (SD), y 60.7 (13.6)a .87 62.4 (13.0)b <.001 Reason for urethral catheter placement, No. (%) Perioperative use for surgical procedure (n = 1653) 174 (10.5) .29 880 (53.2) <.001 Urinary retention or bladder obstruction (n = 144) 20 (13.9) 104 (72.2) Other or unknown (n = 279) 25 (9.0) 166 (59.5) AUA Symptom Index score, No. (%) Mild (n = 1176) 98 (8.3) <.001 519 (44.1) <.001 Moderate (n = 670) 79 (11.8) 442 (66.0) Severe (n = 213) 37 (17.4) 175 (82.2) Duration of urinary catheter, No. (%) ≤3 d (n = 1578) 154 (9.8) .04 796 (50.4) <.001 >3 d (n = 498) 65 (13.1) 354 (71.1) Abbreviation: AUA, American Urological Association. a Comparison group: no infectious complications; mean (SD) age, 60.8 (13.4) years. b Comparison group: no noninfectious complications; mean (SD) age, 58.8 (13.8) years. Discussion This prospective cohort study of urethral catheter–associated complications conducted at 4 US medical centers in 2 states has 3 key findings. First, the overall patient-reported complication rate of indwelling urethral catheter use was found to be 57.0%, higher than what is commonly reported in the literature.7,9 Second, noninfectious complications were reported 5 times more often than infectious complications. However, women were almost twice as likely to report an infectious complication (15.5% vs 8.6%), while men were more likely to report a noninfectious complication (58.6% vs 47.3%). Finally, more than one-third of patients with catheters in place reported restrictions in activities of daily living (39.5%) and social activity (43.9%). These findings are novel, relevant to patient safety, and could not have been uncovered without direct follow-up in our study of patients who had been catheterized.

. Finally, more than one-third of patients with catheters in place reported restrictions in activities of daily living (39.5%) and social activity (43.9%). These findings are novel, relevant to patient safety, and could not have been uncovered without direct follow-up in our study of patients who had been catheterized. Other investigators have also evaluated the noninfectious complications of urethral catheters, albeit without the direct patient perspective. For example, Leuck et al7 prospectively reviewed the medical records at a single Veterans Affairs hospital to evaluate both the infectious and traumatic complications associated with the use of an indwelling urethral catheter. When reports of traumatic episodes associated with pain or that required intervention were compared with episodes of symptomatic urinary tract infection (the Centers for Disease Control and Prevention’s current focus for surveillance and treatment), trauma was significantly more common than infection (0.9 per 100 catheter-days vs 0.3 per 100 catheter-days; P < .001).7 Using administrative data on surgical procedures, Aaronson et al8 concluded that catheter-associated complications were reported rarely but were significantly associated with increased length of hospital stay and catheter-associated urinary tract infection rates. Awad and colleagues15 conducted a legal database review of the association of urethral catheters with medical malpractice from 1965 to 2015 and reported that the most common malpractice claim was traumatic insertion of a urethral catheter (48% of claims). These investigators noted that several of the lawsuits involved a spouse who claimed loss of consortium (deprivation of the benefits of a family relationship because of injuries caused by the tort). Finally, Davis and colleagues16 followed up patients at 2 tertiary care hospitals in Ireland for 6 months and reported an incidence of traumatic urethral catheterization of 6.7 per 1000 catheters with an additional length of hospital stay of more than 9 days for those patients coupled with incremental health care costs.

inally, Davis and colleagues16 followed up patients at 2 tertiary care hospitals in Ireland for 6 months and reported an incidence of traumatic urethral catheterization of 6.7 per 1000 catheters with an additional length of hospital stay of more than 9 days for those patients coupled with incremental health care costs. Our findings confirm the importance of noninfectious complications based on reports of a diverse group of patients who received care at 4 different medical centers. We also received patient-reported accounts about their experience with the use of the catheter. Many patients complained about the indwelling catheter. For example, one patient reported, “I never want another Foley catheter. Hurts like hell!” Another reported that the catheter was “uncomfortable, cumbersome, and difficult to sleep with.” Many patients noted that the removal process was very painful: “It felt like it was cutting me coming out.” Not all patients reported negative personal experiences with the indwelling catheter. A few patients contacted viewed the urethral catheter favorably: “It was my first experience with a catheter; it wasn’t too bad.”

y patients noted that the removal process was very painful: “It felt like it was cutting me coming out.” Not all patients reported negative personal experiences with the indwelling catheter. A few patients contacted viewed the urethral catheter favorably: “It was my first experience with a catheter; it wasn’t too bad.” We prospectively followed up a large cohort to identify complications that resulted from urethral catheterization and were described by patients rather than relying on information that may be documented in the medical record or may be available through secondary data sources. This information provided a unique and important perspective on potential complications that may concern patients and may occur outside the hospital setting. For example, urine leaking from the catheter when the device was in place or from the urethra after it had been removed was an issue for many patients. As one patient stated, “I wasn’t given any instruction on how to handle a leak”; this incident resulted in his researching information on the internet. Our study extended understanding of urethral catheter–associated complications by also identifying lifestyle issues—such as sexual problems—that are important to patients. In the words of one patient, “I am mostly dissatisfied with my urinary condition because I have not been able to have sex in a very long time.”

net. Our study extended understanding of urethral catheter–associated complications by also identifying lifestyle issues—such as sexual problems—that are important to patients. In the words of one patient, “I am mostly dissatisfied with my urinary condition because I have not been able to have sex in a very long time.” The analysis also identified sex-related differences in reported complications, which makes some sense given the anatomical differences of the genitourinary tracts. Women are at higher risk for catheter-associated urinary tract infections than men because they have shorter urethras and closer proximity of perineal bacterial colonization to the insertion site of the indwelling catheter.17 For male patients, complications potentially associated with the presence of an enlarged prostate gland (eg, urinary retention) were seen more commonly than the same complications in women. Although male sex was no longer an independent risk factor for noninfectious complications when assessed in a multivariable model including the AUA symptom index score, this is likely linked to collinearity with the AUA symptom index score.

rinary retention) were seen more commonly than the same complications in women. Although male sex was no longer an independent risk factor for noninfectious complications when assessed in a multivariable model including the AUA symptom index score, this is likely linked to collinearity with the AUA symptom index score. Our findings have clinical implications. We should launch efforts to reduce the noninfectious complications of urethral catheters in the same way that we have used collaborative efforts to reduce catheter-associated urinary tract infections in some patient populations.18,19 Attention to the technical aspects of proper catheter insertion and maintenance as well as increased focus on reducing unnecessary catheter insertions and unnecessary days of catheterization duration seem likely to reduce noninfectious and infectious complications. As Meddings and colleagues20 concluded in a systematic review of 30 studies, reminders about an indwelling urinary catheter and stop orders should be used to improve patient safety because these approaches will reduce both types of catheter complications. The indwelling urethral catheter has even been referred to as a “1-point restraint,”21 which underscores the importance of limiting the use of catheters. Strategies for better tracking and surveillance of noninfectious complications, such as major trauma or bleeding, should also be considered. Incorporating what happens after the hospital stay may fully capture possible complications of this invasive procedure.

nderscores the importance of limiting the use of catheters. Strategies for better tracking and surveillance of noninfectious complications, such as major trauma or bleeding, should also be considered. Incorporating what happens after the hospital stay may fully capture possible complications of this invasive procedure. Limitations This study should be interpreted in the context of the following limitations. We conducted this study in the United States at 4 sites within 2 states, and we included only patients who received an indwelling catheter during acute care hospitalization. Thus, the findings may not be generalizable to all patients who receive a urethral catheter, such as those who are evaluated in the emergency department for urinary retention and discharged to home. We might expect worse outcomes in these patients; they often have difficulty obtaining timely follow-up care for removal of the catheter. Also, one of the sites contributed modestly to the overall number of patients enrolled for reasons primarily related to a population of younger patients who had a low baseline need for indwelling urethral catheters. In addition, the information was collected directly from the patient. Although patient-reported complications may not constitute a medically defined complication because of well-known poor documentation of urethral catheter complications,22 we believe that what patients report is important to understand the full scope of potential problems. Without an approach to validate their reports, however, inclusion of patient-reported complications could result in an overestimate of true medical complications.

oor documentation of urethral catheter complications,22 we believe that what patients report is important to understand the full scope of potential problems. Without an approach to validate their reports, however, inclusion of patient-reported complications could result in an overestimate of true medical complications. Conclusions We provided important estimates of the burden of complications due to the use of this device. We also provided a foundation for interventional studies to further ameliorate complications associated with catheter use. Urethral catheters, which are often used to help safely manage care of the patient during hospitalization, often lead to the opposite result. In light of the frequency with which urethral catheters are used, we should consider not only infectious complications but also the noninfectious complications associated with these catheters as key areas of possible harms and thus vital targets for future prevention efforts. Supplement. eTable 1. Categorization of Urethral Catheter–Associated Complications eTable 2. Reason for Catheter Placement and Reported Experience During Catheter Insertion eTable 3. Multivariable Linear Regression Model to Predict Infectious and Noninfectious Urethral Catheter Complications Click here for additional data file.

that are rarely fatal, such as thyroid cancer, in this study was limited as the cohort was followed for mortality outcomes only. Studies with high-quality dosimetry and cancer incidence follow-up are needed to better characterize the risk of thyroid cancer in adults and children after RAI treatment for hyperthyroidism. The lack of evidence of a dose-response for leukemia mortality in the current study was somewhat unexpected, given that an elevated risk of leukemia was observed in patients with thyroid cancer who received much higher levels of administered activity. Low-dose radiation exposure was also found to be associated with leukemia risk in previous studies. However, greater uncertainty in the calculation of red bone marrow activity compared with that of other organ- or tissue-absorbed doses, combined with the low-dose range of exposure and relatively small number of leukemia deaths, limited our ability to detect a dose-response relationship in this setting. Thus, we cannot rule out an association between radiation treatment for hyperthyroidism and leukemia induction.

other organ- or tissue-absorbed doses, combined with the low-dose range of exposure and relatively small number of leukemia deaths, limited our ability to detect a dose-response relationship in this setting. Thus, we cannot rule out an association between radiation treatment for hyperthyroidism and leukemia induction. Strengths and Limitations Strengths of this study included the large cohort size, long-term follow-up, detailed clinical information, and analytic methods to minimize confounding. In addition, we used high-quality, individualized organ- and tissue-specific dose estimates derived from a biokinetic model that was developed and calibrated in a small group of well-characterized patients with hyperthyroidism, with individual radioactivity measurements, calculated S values, and clinical measurements available for all (administered activity) or most (thyroid mass and percentage uptake) of the RAI-treated patients in the full study.

developed and calibrated in a small group of well-characterized patients with hyperthyroidism, with individual radioactivity measurements, calculated S values, and clinical measurements available for all (administered activity) or most (thyroid mass and percentage uptake) of the RAI-treated patients in the full study. Nonetheless, major uncertainties in the organ (particularly non-thyroid) dose estimations may have biased the study findings toward the null. The small number of deaths for many cancers of interest, together with relatively small doses to organs other than the thyroid, limited our power to detect statistically significant associations for those outcomes. Considering the number of statistical tests performed, some results may be because of chance; therefore, the results should be interpreted with caution. As the results were unadjusted for smoking, obesity, alcohol use, or reproductive factors, confounding by these known cancer risk factors is possible. Although the results were adjusted for additional treatment with antithyroid drugs (yes or no), residual confounding may be possible, as this information may not have been complete for all patients, and the information on type, quantity, and dates of use was not recorded in the database. Given that the types of antithyroid drugs administered to patients in the cohort differed from those more commonly prescribed in recent years (eg, methimazole and carbimazole), additional studies that can more precisely evaluate the long-term health implications of current antithyroid drug therapies, including in the context of RAI therapy. Confounding by severity of the underlying hyperthyroid status and other concomitant diseases was a potential source of bias, but we found that the results changed little after adjusting for differences in clinical impression of disease severity.

Introduction In the United States, the prevalence of hyperthyroidism is 1.2% (0.5% overt and 0.7% subclinical), and most cases are due to Graves disease. Radioactive iodine (RAI; sodium iodide I 131, or Na131I) has been extensively used to treat hyperthyroidism since the 1940s and has been the preferred first-line treatment by US physicians for uncomplicated Graves disease. However, in recent decades, preference for RAI therapy as a primary treatment for Graves disease has declined in favor of antithyroid drugs, likely reflecting the increased awareness of an association between RAI and worsening of Graves ophthalmopathy as well as concerns about risks of radiation-induced cancer, whereas preference for initial surgical treatment has remained low.

primary treatment for Graves disease has declined in favor of antithyroid drugs, likely reflecting the increased awareness of an association between RAI and worsening of Graves ophthalmopathy as well as concerns about risks of radiation-induced cancer, whereas preference for initial surgical treatment has remained low. Postsurgical RAI ablation therapy for differentiated thyroid cancer (eg, >100 mCi) has been associated with an increased risk of secondary malignant neoplasms, including bone and soft-tissue sarcomas, salivary gland and digestive tract cancers, and leukemia. However, relatively few cohort studies have evaluated cancer risk after RAI for hyperthyroidism, which involves much lower administered activities (typically 10-15 mCi), and findings have been inconsistent. In previous studies, the amount of administered activity was used as a surrogate measure of radiation exposure; however, radiation absorption varies by organ and can vary substantially across patients for a given administered dose. To date, risks of specific cancers have not been precisely quantified with well-substantiated estimates of absorbed dose to individual organs, which account for administered activity along with anatomic and physiologic patient characteristics associated with these effects.

across patients for a given administered dose. To date, risks of specific cancers have not been precisely quantified with well-substantiated estimates of absorbed dose to individual organs, which account for administered activity along with anatomic and physiologic patient characteristics associated with these effects. The largest and most comprehensive study to date on this topic used data from Cooperative Thyrotoxicosis Therapy Follow-up Study, which had a cohort of more than 35 000 patients with hyperthyroidism (65% of whom were treated with RAI) enrolled between 1946 and 1964 in the United States and United Kingdom and with mortality follow-up through 1990. In that study, total and site-specific cancer mortality rates were not found to be elevated in RAI-treated patients compared with the general population. Although thyroid cancer mortality was elevated 4-fold in RAI-treated patients, this risk did not appear to be associated with radiation exposure from the treatment, given that greater administered activity, used as a proxy measure for absorbed dose to the thyroid, was not associated with thyroid cancer mortality after accounting for a 5-year latency period. Greater estimated red bone marrow–absorbed doses were not associated with death from hematopoietic and lymphoproliferative malignant neoplasms. Absorbed doses to other organs and tissues were calculated according to simplistic dosimetry assumptions but were not evaluated in relation to site-specific cancer death. On the basis of those findings, those authors concluded that RAI appeared to be a safe therapy for hyperthyroidism.

roliferative malignant neoplasms. Absorbed doses to other organs and tissues were calculated according to simplistic dosimetry assumptions but were not evaluated in relation to site-specific cancer death. On the basis of those findings, those authors concluded that RAI appeared to be a safe therapy for hyperthyroidism. We have since extended the follow-up of the Cooperative Thyrotoxicosis Therapy Follow-up Study by 24 years (maximum 68 years) and, we now use improved individual estimates of absorbed organ doses of I 131, as described in a previous publication. The objective of the current study was to evaluate the radiation dose-response relationships for site-specific cancer death among the RAI-treated patients with hyperthyroidism in the cohort. Methods Study Population The mortality follow-up of the cohort is approved annually by the Institutional Review Board of the National Cancer Institute. Because the mortality follow-up is based on linkages with available databases and involves no direct contact with study participants, the requirement for informed consent was waived by the Institutional Review Board of the National Cancer Institute. The current analyses were conducted from April 28, 2017, to January 30, 2019.

the mortality follow-up is based on linkages with available databases and involves no direct contact with study participants, the requirement for informed consent was waived by the Institutional Review Board of the National Cancer Institute. The current analyses were conducted from April 28, 2017, to January 30, 2019. The Cooperative Thyrotoxicosis Therapy Follow-up Study included all patients with a hyperthyroidism diagnosis between 1946 and 1964 at 25 US medical centers and 1 UK hospital. Comprehensive clinical data were abstracted from medical records. Patients were initially followed up through June 30, 1968. They were asked to return to the clinic at 2-year intervals for a physical examination, brief history, and additional blood studies. Follow-up information was obtained from the treating physicians or medical records, if available, or directly from the patients via mailed questionnaires. In 1984, investigators from the National Cancer Institute of the National Institutes of Health reassembled the cohort data from printed computer listings, microfiche, microfilm cassettes, and handwritten documents maintained at the collaborating medical centers. The records for 35 630 patients were compiled at 4 regional centers (Harvard University, Boston, Massachusetts; Memorial Sloan-Kettering Cancer Center, New York, New York; University of Southern California, Los Angeles; and Research Triangle Institute; Research Triangle Park, North Carolina). Patients were traced using records from the National Death Index, Social Security Administration, and other resources, and copies of death certificates were obtained and coded by trained nosologists. After excluding duplicates and incomplete records, the final data set included 35 593 patients, of whom 28 719 had complete mortality follow-up through 1990. Mortality follow-up continued for the US patients by linking with various tracing resources, including the Social Security Administration records to determine vital status and National Death Index Plus records to identify causes of death for decedents through December 31, 2014.

omplete mortality follow-up through 1990. Mortality follow-up continued for the US patients by linking with various tracing resources, including the Social Security Administration records to determine vital status and National Death Index Plus records to identify causes of death for decedents through December 31, 2014. We further excluded those with no follow-up information, missing entry or exit dates, or exit dates that occurred on or before study entry (the first patient visit to a participating study clinic during the study enrollment period), resulting in 31 332 patients treated with RAI, surgical procedure, antithyroid drugs, or a combination of these options. From the 19 558 remaining patients who received RAI therapy, alone or in combination with other treatments, we excluded an additional 753 patients with a cancer diagnosis before study entry. After the exclusions, 18 805 patients were eligible for the analysis.

dure, antithyroid drugs, or a combination of these options. From the 19 558 remaining patients who received RAI therapy, alone or in combination with other treatments, we excluded an additional 753 patients with a cancer diagnosis before study entry. After the exclusions, 18 805 patients were eligible for the analysis. Dose Estimation A comprehensive dose assessment was conducted to estimate organ- or tissue-absorbed doses for cohort members who received RAI therapy (described in detail in a previous publication). In brief, an iodine mathematical biokinetic model was developed and calibrated using data from a kinetically well-characterized group of 197 patients with hyperthyroidism who had thyroid, blood, and urine measurements of I 131. The model estimates, along with individual patient data, were then applied to the full cohort of RAI-treated patients with hyperthyroidism and used to calculate the numbers of I 131 disintegrations in the source organs and tissues (regions with increased I 131 avidity). Iodine 131 photon and electron spectra from the International Commission on Radiological Protection were used to compute the S values (mean absorbed dose in a target region per unit disintegration of I 131 in a source region) on the adult reference voxel phantoms adopted by the International Commission on Radiological Protection for all important combinations of source and target regions. Absorbed dose to 26 target organs and tissues were calculated using new S value estimates for each patient and numbers of disintegrations in the source organs derived from the biokinetic model (from both diagnostic tests and therapeutic treatments).

rotection for all important combinations of source and target regions. Absorbed dose to 26 target organs and tissues were calculated using new S value estimates for each patient and numbers of disintegrations in the source organs derived from the biokinetic model (from both diagnostic tests and therapeutic treatments). Statistical Analysis For the current analyses, person-years at risk for each patient were computed from 5 years after the date of the last RAI treatment (if a typical latency period were assumed between radiation exposure and solid cancer occurrence) until the date of death, date last known to be alive for patients lost to follow-up, or end of follow-up (December 31, 2014). Dose-response analyses were conducted among the patients receiving RAI treatment by fitting multivariable linear excess relative risk (ERR) models to disease rates. To directly compare the strength of the radiation dose-response relationship across the specific causes of cancer death, we calculated the ERRs per 100 mGy absorbed dose to the target organ or tissue for that cancer site (with stomach dose as a whole-body dose surrogate) and computed 95% likelihood-based CIs around these estimates. These models have the following form: background(a, s, b, x)[1+βd*f(y)], in which the rate is a function of age(a), sex(s), birth cohort(b), other risk factors(x), and dose(d), and f(y) describes the effects of modifying factors (eg, attained age). Background rates included sex-dependent functions of attained age and birth cohort (continuous), with further adjustment for the following potential confounders, chosen a priori: Graves disease diagnosis (yes or no), additional surgical procedure (yes or no), and additional antithyroid drug treatment (yes or no).

age). Background rates included sex-dependent functions of attained age and birth cohort (continuous), with further adjustment for the following potential confounders, chosen a priori: Graves disease diagnosis (yes or no), additional surgical procedure (yes or no), and additional antithyroid drug treatment (yes or no). Relative risks (RRs) at 100 mGy were calculated by adding 1 to these estimates (RR = 1 + ERR). Baseline clinical impression (eg, mild, moderate, severe hyperthyroidism, or unknown), as recorded by physicians in each center, was not retained in the final models because inclusion of this term did not statistically significantly improve the model fit or affect the estimates. Beginning follow-up 2 years, as opposed to 5 years, after the last treatment did not change the results for non–solid cancer mortality (leukemia excluding chronic lymphocytic leukemia and non-Hodgkin lymphoma). Stratified person-year computations and risk estimations were performed using Epicure, version 2.00.02 (Risk Sciences International). Two-sided hypothesis tests and 95% CIs were based on likelihood ratio tests, with a significance threshold of P = .05.

a excluding chronic lymphocytic leukemia and non-Hodgkin lymphoma). Stratified person-year computations and risk estimations were performed using Epicure, version 2.00.02 (Risk Sciences International). Two-sided hypothesis tests and 95% CIs were based on likelihood ratio tests, with a significance threshold of P = .05. These risk estimates, along with baseline age-specific total and cause-specific death rates obtained from the 2014 Surveillance, Epidemiology, and End Results mortality data, were used to estimate the number of future cancer deaths estimated to be attributed to radiation exposure on the basis of a hypothetical population of patients who received the same organ or tissue dose at the same age (eg, 40 years) and in the same calendar year (eg, 2014). For this analysis, we used the risk of exposure-induced death method, which is the sum of the expected number of exposure-associated deaths at each age, taking into account competing risks. Estimates of the number of future cancer deaths were given per 1000 patients. Results Of the 18 805 patients in the study cohort, 14 671 (78.0%) were women and 17 615 (93.7%) had Graves disease (Table 1). The mean (SD) age at entry was 49 (14) years.

These risk estimates, along with baseline age-specific total and cause-specific death rates obtained from the 2014 Surveillance, Epidemiology, and End Results mortality data, were used to estimate the number of future cancer deaths estimated to be attributed to radiation exposure on the basis of a hypothetical population of patients who received the same organ or tissue dose at the same age (eg, 40 years) and in the same calendar year (eg, 2014). For this analysis, we used the risk of exposure-induced death method, which is the sum of the expected number of exposure-associated deaths at each age, taking into account competing risks. Estimates of the number of future cancer deaths were given per 1000 patients. Results Of the 18 805 patients in the study cohort, 14 671 (78.0%) were women and 17 615 (93.7%) had Graves disease (Table 1). The mean (SD) age at entry was 49 (14) years. Table 1. Selected Characteristics of the Study Population Characteristic Participants, No. (%) Total 18 805 (100) Age at study entry, y <30 1733 (9.2) 30-39 3055 (16.2) 40-49 4816 (25.6) ≥50 9201 (48.9) Sex Male 4134 (22.0) Female 14 671 (78.0) Hyperthyroidism diagnosis Graves disease 17 615 (93.7) Toxic nodular goiter 934 (5.0) Unknown 256 (1.4) Treatment combination RAI only 7182 (38.2) RAI and surgical procedure 694 (3.7) RAI and drugs 8675 (46.1) RAI, surgical procedure, and drugs 2254 (12.0) Vital status Alive or lost to follow-up 3321 (17.7) Deceased 15 484 (82.3) Study site Mount Sinai Hospital (New York, NY) 3042 (16.2) Mayo Clinic (Rochester, MN) 1907 (10.1) Massachusetts General (Boston, MA) 1726 (9.2) Sheffield Hospital (Sheffield, UK) 1378 (7.3) Columbia Presbyterian Hospital (New York, NY) 1190 (6.3) Cedars Sinai Medical Center (Los Angeles, CA) 1093 (5.8) Los Angeles County Hospitals (Los Angeles, CA) 1065 (5.7) University of Michigan (Ann Arbor, MI) 978 (5.2) University of Maryland (Baltimore, MD) 899 (4.8) University of California (San Francisco, CA) 747 (4.0) Beth Israel Hospital (Boston, MA) 678 (3.6) University Hospitals of Cleveland (Cleveland, OH) 630 (3.4) New York Hospital-Cornell (New York, NY) 612 (3.3) Lahey Clinic (Boston, MA) 589 (3.1) Montefiore Med Center (New York, NY) 560 (3.0) White Memorial Hospital (Los Angeles, CA) 518 (2.8) University of Cincinnati (Cincinnati, OH) 492 (2.6) Memorial Sloan-Kettering (New York, NY) 283 (1.5) Cleveland Metropolitan General Hospital (Cleveland, OH) 198 (1.1) Strong Memorial Hospital (Rochester, NY) 125 (0.7) St Louis University (St Louis, MO) 52 (0.3) Ochsner Clinic (New Orleans, LA) 43 (0.2) Abbreviation: RAI, radioactive iodine.

i (Cincinnati, OH) 492 (2.6) Memorial Sloan-Kettering (New York, NY) 283 (1.5) Cleveland Metropolitan General Hospital (Cleveland, OH) 198 (1.1) Strong Memorial Hospital (Rochester, NY) 125 (0.7) St Louis University (St Louis, MO) 52 (0.3) Ochsner Clinic (New Orleans, LA) 43 (0.2) Abbreviation: RAI, radioactive iodine. The treatment combination of RAI and antithyroid drugs was the most common (8675 [46.1%]), followed by RAI alone (7182 [38.2%]); RAI, surgical procedure, and drugs (2254 [12.0%]); and RAI and surgical procedure (694 [3.7%]). By number of RAI treatments, 12 387 patients (65.9%) received 1 treatment, 3629 (19.3%) received 2 treatments, 1317 (7.0%) received 3 treatments, and 1471 (7.8%) received more than 3 treatments. Mean (SD) total administered activity (including diagnostic and therapeutic activity) was 375 MBq (10.1 mCi) for patients with Graves disease (median [interquartile range (IQR)], 269 [187-419] MBq) and 653 MBq (17.6 mCi) for patients with toxic nodular goiter (median [IQR], 488 [301-792] MBq).

d more than 3 treatments. Mean (SD) total administered activity (including diagnostic and therapeutic activity) was 375 MBq (10.1 mCi) for patients with Graves disease (median [interquartile range (IQR)], 269 [187-419] MBq) and 653 MBq (17.6 mCi) for patients with toxic nodular goiter (median [IQR], 488 [301-792] MBq). The highest mean estimated absorbed doses were to the thyroid (130 Gy) (to convert gray to rad, multiply by 100), followed by the esophagus (1.6 Gy); liver, oral mucosa, lung, stomach, red bone marrow, female breast, pancreas, kidney (100-400 mGy); and uterus, brain, bladder, ovary, prostate, and colon or rectum (20-99 mGy). Factors associated with total administered activity included number of RAI treatments, older age at entry, clinical impression of disease severity, toxic nodular goiter diagnosis, previous weight loss, history of heart disease and diabetes, and additional treatment with both surgical procedure and antithyroid drugs (Table 2).

associated with total administered activity included number of RAI treatments, older age at entry, clinical impression of disease severity, toxic nodular goiter diagnosis, previous weight loss, history of heart disease and diabetes, and additional treatment with both surgical procedure and antithyroid drugs (Table 2). Table 2. Baseline Demographic and Clinical Characteristics of Patients With Hyperthyroidism Treated With Radioactive Iodine Baseline Characteristics Tertile of Total Administered Activity, MBq 1 (n = 6296) 2 (n = 6248) 3 (n = 6261) Total administered activity, mean (SD), MBqa 160 (45) 289 (46) 724 (500) No. of RAI treatments, mean (SD) 1.3 (1.2) 1.4 (0.9) 2.4 (1.9) Dose, mGy Stomach 65 (20) 120 (25) 320 (240) Breastb 60 (21) 110 (29) 280 (210) Thyroid 65 000 (31 000) 110 000 (41 000) 210 000 (160 000) Study entry age, mean (SD), y 47 (13) 49 (14) 52 (14) Sex, No. (%) Male 1288 (20.5) 1383 (22.1) 1463 (23.4) Female 5008 (79.5) 4865 (77.9) 4798 (76.6) Clinical impression, No. (%) Suspect 111 (1.8) 96 (1.5) 72 (1.2) Mild 1758 (27.9) 1649 (26.4) 1567 (25.0) Moderate 2564 (40.7) 2605 (41.7) 2639 (42.2) Severe 439 (7.0) 596 (9.5) 857 (13.7) Unknown 1424 (22.6) 1301 (20.8) 1126 (18.0) Type of hyperthyroidism, No. (%) Graves disease 6127 (97.3) 5967 (95.5) 5521 (88.2) Toxic nodular goiter 115 (1.8) 207 (3.3) 612 (9.8) Intermediate status 54 (0.9) 74 (1.2) 128 (2.0) Weight loss prior to surgical procedure, No. (%) 4165 (66.2) 3995 (63.9) 3911 (62.5) Amount of weight loss prior to surgical procedure, mean (SD), lb 19 (13) 21 (14) 23 (16) Medical history (prior to surgical procedure), No. (%) Coronary heart disease 257 (4.1) 442 (7.1) 654 (10.5) Hypertensive heart disease 364 (5.8) 564 (9.0) 720 (11.5) Rheumatic heart disease 137 (2.2) 191 (3.1) 222 (3.6) Other heart disease 274 (4.4) 398 (6.4) 718 (11.5) Diabetes 238 (3.8) 303 (4.9) 433 (6.9) Treatment combinations, No. (%) RAI only 2589 (41.1) 2597 (41.6) 1996 (31.9) RAI and surgical procedure 278 (4.4) 225 (3.6) 191 (3.1) RAI and drugs 798 (12.5) 726 (11.6) 739 (11.8) RAI, surgical procedure, and drugs 2640 (41.9) 2700 (43.2) 3335 (53.3) Abbreviations: MBq, megabecquerel; RAI, radioactive iodine.

33 (6.9) Treatment combinations, No. (%) RAI only 2589 (41.1) 2597 (41.6) 1996 (31.9) RAI and surgical procedure 278 (4.4) 225 (3.6) 191 (3.1) RAI and drugs 798 (12.5) 726 (11.6) 739 (11.8) RAI, surgical procedure, and drugs 2640 (41.9) 2700 (43.2) 3335 (53.3) Abbreviations: MBq, megabecquerel; RAI, radioactive iodine. SI conversion factors: To convert pounds to kg, multiply by 0.45. a Includes diagnostic and therapeutic doses. b Women only.

33 (6.9) Treatment combinations, No. (%) RAI only 2589 (41.1) 2597 (41.6) 1996 (31.9) RAI and surgical procedure 278 (4.4) 225 (3.6) 191 (3.1) RAI and drugs 798 (12.5) 726 (11.6) 739 (11.8) RAI, surgical procedure, and drugs 2640 (41.9) 2700 (43.2) 3335 (53.3) Abbreviations: MBq, megabecquerel; RAI, radioactive iodine. SI conversion factors: To convert pounds to kg, multiply by 0.45. a Includes diagnostic and therapeutic doses. b Women only. During follow-up (mean of 26 years; maximum of 68 years), 15 484 deaths (82.3%) were recorded. Malignancy was the primary cause for 2366 deaths (15.3%). After excluding deaths in the first 5 years after the last RAI treatment, positive dose-response relationships were observed for mortality from most of the individual solid cancers evaluated (Table 3), with statistically significant associations observed for female breast cancer (n = 291; RR at 100-mGy dose to the breast = 1.12; 95% CI, 1.003-1.32; P = .04) and all other solid cancers combined (RR at 100-mGy dose to the stomach = 1.05; 95% CI, 1.01-1.10; P = .01). For all solid cancer mortality (n = 1984), the RR at 100-mGy dose to the stomach was 1.06 (95% CI, 1.02-1.10; P = .002). The 100-mGy dose to the stomach corresponded to a mean (SD) administered activity of 243 (35) MBq in patients with Graves disease (median [IQR], 234 [223-263] MBq), whereas the 100-mGy dose to the breast corresponded to a mean (SD) administered activity of 266 (58) MBq (median [IQR], 260 [224-297] MBq). The RR for thyroid cancer mortality was evaluated at 100 Gy thyroid-absorbed dose, a typical high dose received by the thyroid gland; however, the variability around this estimate was wide and based on only 15 deaths (RR = 1.20; 95% CI, <1.00 to 610). Because these estimates were based on linear dose-response models, risks associated with higher (or lower) doses can be directly estimated from these models.

cal high dose received by the thyroid gland; however, the variability around this estimate was wide and based on only 15 deaths (RR = 1.20; 95% CI, <1.00 to 610). Because these estimates were based on linear dose-response models, risks associated with higher (or lower) doses can be directly estimated from these models. Table 3. Relative Risks and 95% CIs for Cancer-Specific Mortality Among Patients With Hyperthyroidism Treated With Radioactive Iodine Cause of Cancer Deatha Absorbed Dose, mGy Dose-Response Relationship Cause-Specific Cancer Death Attributed to Irradiation, No. (%)c Target Organ or Tissue Organ- or Tissue-Absorbed Dose, Mean (SD) No. of Deaths At 100-mGy Organ- or Tissue-Absorbed Dose, RR (95% CI)b P Value Solid cancers Oral cavity Mucosa 320 (320) 31 0.99 (<0.99-1.30) >.50 Esophageal Esophagus 1600 (1500) 38 1.01 (<1.00-1.87) >.50 Stomach Stomach 170 (180) 97 1.03 (<0.98-1.28) >.50 Colon Colon 23 (25) 258 1.19 (<0.80-2.17) >.50 Rectal Rectum 18 (19) 49 1.54 (<0.75-6.53) >.50 Liver Liver 390 (460) 34 0.99 (<0.99-1.12) >.50 Pancreatic Pancreas 110 (120) 132 1.13 (<0.97-1.56) .27 Lung or bronchus Lung 310 (310) 437 1.02 (<0.99-1.07) .31 Bladder Bladder 49 (50) 54 0.96 (<0.96-2.15) >.50 Kidney Kidney 110 (130) 48 1.32 (<0.97-9.34) >.50 Brain or central nervous system Brain 58 (56) 39 1.07 (<0.93-2.98) >.50 Thyroid Thyroid 130 000 (110 000) 15 1.20 (<1.00-6.10)d >.50 Female breast Breast 150 (160) 291 1.12 (1.00-1.32) .04 41.9 (14) Uterine Uterus 63 (69) 63 1.54 (0.98-3.42) .07 Ovarian Ovary 38 (42) 104 1.32 (<0.90-2.46) .30 Prostate Prostate 42 (41) 52 1.04 (<0.86-2.42) >.50 All other solid cancers Stomach 170 (180) 242 1.02 (<0.98-1.16) >.50 Leukemia (excluding CLL) Marrow 160 (160) 59 0.97 (<0.96-1.26) >.50 Non-Hodgkin lymphoma Marrow 160 (160) 70 1.07 (<0.96-1.54) >.50 Multiple myeloma Marrow 160 (160) 30 1.69 (<0.97->6.00) >.50 All solid cancers combined Stomach 170 (180) 1984 1.06 (1.02-1.10) .002 154.7 (8) All solid cancers excluding female breast Stomach 170 (180) 1693 1.05 (1.01-1.10) .01 117.2 (7) Abbreviations: CLL, chronic lymphocytic leukemia; ERR, excess relative risk; Gy, gray; RAI, radioactive iodine; RR, relative risk.

>6.00) >.50 All solid cancers combined Stomach 170 (180) 1984 1.06 (1.02-1.10) .002 154.7 (8) All solid cancers excluding female breast Stomach 170 (180) 1693 1.05 (1.01-1.10) .01 117.2 (7) Abbreviations: CLL, chronic lymphocytic leukemia; ERR, excess relative risk; Gy, gray; RAI, radioactive iodine; RR, relative risk. a Patient deaths that occurred in the first 5 years after last RAI treatment were excluded. b Based on a linear excess RR model using continuous values of organ- or tissue-absorbed dose (per 100 mGy), in which RR = 1 + ERR. Background rates include terms for sex, sex-specific attained age and birth cohort patterns, Graves disease diagnosis (yes or no), additional treatment with surgical procedure (yes or no), and additional treatment with antithyroid drugs (yes or no). To calculate RRs at 100*x mGy organ or tissue dose, use the following equation: [(ERR)*x +1]. c Estimates are shown only when the corresponding RRs were statistically significant at P < .05. d RR at 100 Gy.

b Based on a linear excess RR model using continuous values of organ- or tissue-absorbed dose (per 100 mGy), in which RR = 1 + ERR. Background rates include terms for sex, sex-specific attained age and birth cohort patterns, Graves disease diagnosis (yes or no), additional treatment with surgical procedure (yes or no), and additional treatment with antithyroid drugs (yes or no). To calculate RRs at 100*x mGy organ or tissue dose, use the following equation: [(ERR)*x +1]. c Estimates are shown only when the corresponding RRs were statistically significant at P < .05. d RR at 100 Gy. The dose-response relationships for all solid cancer, breast cancer, and all solid cancer (excluding breast) did not change after restricting to patients receiving a single RAI treatment or RAI-only treatment (no surgical procedure or antithyroid drugs) (eTable 1 in the Supplement) or after excluding patients with toxic nodular goiter. The shapes of the dose-response relationships were consistent with linearity (Figure). The RRs for all solid cancer mortality (but not for breast cancer or all other solid cancer) decreased with increasing attained age. We observed no evidence of a dose-response relationship for mortality from leukemia, excluding chronic lymphocytic leukemia (59 deaths; RR at 100 mGy = 0.97; 95% CI, <0.96 to 1.26), non-Hodgkin lymphoma (70 deaths; RR at 100 mGy = 1.07; 95% CI, <0.96 to 1.54), or multiple myeloma (30 deaths; RR at 100 mGy = 1.69; 95% CI, <0.97 to >6.0) (Table 3). We estimated that 8% of the solid cancer deaths, including 14% of breast cancer and 7% of all other solid cancer deaths, during follow-up were attributed to radiation in these patients.

RR at 100 mGy = 1.07; 95% CI, <0.96 to 1.54), or multiple myeloma (30 deaths; RR at 100 mGy = 1.69; 95% CI, <0.97 to >6.0) (Table 3). We estimated that 8% of the solid cancer deaths, including 14% of breast cancer and 7% of all other solid cancer deaths, during follow-up were attributed to radiation in these patients. Figure. Relative Risks for Solid Cancer, Female Breast Cancer, and Solid Cancer Without Female Breast Cancer Mortality Among Patients With Hyperthyroidism Patients were cancer free at the time of radioactive iodine (RAI) treatment. Relative risks were compared across organ- or tissue-absorbed dose (in grays [to convert to the conventional unit rad, multiply by 100]). Solid horizontal lines represent the relative risk reference value (1). Solid blue lines represent the estimated log-linear dose-response relationships. Dashed black lines represent the smoothed dose-response relationships, and dashed gray lines represent 95% CIs. Black dots represent the relative risk at each organ dose category. Background rates include terms for sex, sex-specific attained age and birth cohort patterns, Graves disease diagnosis, additional treatment with surgical procedure, and additional treatment with antithyroid drugs.

rrent antithyroid drug therapies, including in the context of RAI therapy. Confounding by severity of the underlying hyperthyroid status and other concomitant diseases was a potential source of bias, but we found that the results changed little after adjusting for differences in clinical impression of disease severity. Conclusions Hyperthyroid treatment decisions should take into consideration the balance of risks with advantages of each available treatment option as well as patient preference, health status, and access to these options. We believe the results of this study provide quantitative estimates of the risks of radiation-associated cancer deaths in RAI-treated patients with hyperthyroidism, which were previously not well understood, and suggest that the risk of death from solid cancer (including breast cancer) increases with the greater absorbed dose to exposed organs and tissues. Additional studies are needed to fully weigh the risks and advantages of RAI and other major treatment options available to patients with hyperthyroidism. Supplement. eTable 1. Relative Risks (RRs) and 95% Confidence Intervals (CIs) for Cancer-Specific Mortality at 100 mGy Organ/Tissue Absorbed Dose Among Hyperthyroid Patients Treated With RAI eTable 2. Estimates of Radiation-Associated Excess Solid Cancer Deaths (95% CIs) per 1,000 Hyperthyroidism Patients (80% Women) treated With Radioactive Iodine, by Age at treatment and Estimated Stomach Dose Click here for additional data file.

d dashed gray lines represent 95% CIs. Black dots represent the relative risk at each organ dose category. Background rates include terms for sex, sex-specific attained age and birth cohort patterns, Graves disease diagnosis, additional treatment with surgical procedure, and additional treatment with antithyroid drugs. Combining these RR estimates with current US mortality rates, we estimated that 13 (95% CI, 2-27) excess solid cancer deaths, including 3 breast cancer deaths (95% CI, 0.1-7), would occur for every 1000 patients (80% women), receiving 100 mGy absorbed dose to the stomach or breast at age 40 years. For every 1000 patients currently treated at age 50 years, we estimated an excess of 12 (95% CI, 2-26) radiation-associated solid cancer deaths, including 3 breast cancer deaths (95% CI, 0.1-7) (eTable 2 in the Supplement). Of the excess solid cancer deaths, only 8% in the age 40 years group and 25% in the age 50 years group would be expected to occur in the first 20 years after treatment. Because higher administered activities are now currently recommended for the treatment of patients with Graves disease (370-555 MBq), we also calculated excess solid cancer deaths at 150-mGy, 200-mGy, and 250-mGy dose to the stomach, which would be more typical of current treatment practices (eTable 2 in the Supplement). At these dose levels, we would expect between 19 and 32 excess solid cancer deaths per 1000 patients treated at age 40 years and between 18 and 31 excess solid cancer deaths per 1000 patients treated at age 50 years.

o the stomach, which would be more typical of current treatment practices (eTable 2 in the Supplement). At these dose levels, we would expect between 19 and 32 excess solid cancer deaths per 1000 patients treated at age 40 years and between 18 and 31 excess solid cancer deaths per 1000 patients treated at age 50 years. Discussion Using data from the world’s largest cohort, to date, of RAI-treated patients with hyperthyroidism, we investigated the association between organ- or tissue-absorbed dose and site-specific cancer death. To our knowledge, this is the first study to characterize the dose-response relationship between RAI treatment and site-specific cancer mortality in patients with hyperthyroidism using reliable estimates of absorbed dose to exposed organs or tissues. We observed that the RR of death from solid cancer, including breast cancer and all other solid cancers combined, increased with greater doses to organs or tissues. We estimated that RAI treatment could result in a small long-term increase in the expected number of solid cancer deaths associated with radiation exposure among patients with hyperthyroidism.

solid cancer, including breast cancer and all other solid cancers combined, increased with greater doses to organs or tissues. We estimated that RAI treatment could result in a small long-term increase in the expected number of solid cancer deaths associated with radiation exposure among patients with hyperthyroidism. Smaller previously studied European cohorts showed evidence of an increased risk of total and/or site-specific cancers following RAI treatment, including cancers in organs that took up radioiodine or were exposed along the I 131 distribution pathways (eg, salivary and digestive organs). However, results from these earlier studies were inconsistent, and it remains unclear, particularly in studies relying on an external (eg, general population) comparison group, whether those findings represent consequences from the treatment or the underlying disease, or if they were biased owing to confounding. Confounding by indication, which occurs when the reasons for choosing a specific treatment option are associated with the outcome under study, is of particular concern in observational studies that compare cancer or other disease risks by treatment type, which was the main approach used in the previous analysis of this cohort. By focusing the present analysis on patients treated with RAI, we were able to minimize this bias.

iated with the outcome under study, is of particular concern in observational studies that compare cancer or other disease risks by treatment type, which was the main approach used in the previous analysis of this cohort. By focusing the present analysis on patients treated with RAI, we were able to minimize this bias. To our knowledge, this study is the first to provide direct evidence of an association between internal exposure from I 131 and breast cancer risk. A similar study of a small cohort of RAI-treated patients with hyperthyroidism in Finland found an elevated risk for breast cancer; however, the dose-response relationship was not evaluated in that study. Ecological data from the Chernobyl radiation accident in Russia previously suggested such an association, with a 2-fold higher breast cancer risk in the most (vs least) contaminated districts starting approximately 10 years after the accident. The dose-response estimate for breast cancer mortality in the present study (RR at 100 mGy = 1.12; 95% CI, 1.003-1.32) was consistent with similar estimates from other populations exposed to ionizing radiation in adulthood, including the Life Span Study of atomic bomb survivors in Japan (RR at 100 mGy at age 30 years with attained age of 70 years = 1.09; 95% CI, 1.06-1.13); the US Radiologic Technologists Study (RR at 100 mGy = 1.07; 95% CI, 0.99-1.19); and smaller environmentally, occupationally, and medically exposed populations. This consistency lends credibility to the estimated doses and findings reported in this study.

years with attained age of 70 years = 1.09; 95% CI, 1.06-1.13); the US Radiologic Technologists Study (RR at 100 mGy = 1.07; 95% CI, 0.99-1.19); and smaller environmentally, occupationally, and medically exposed populations. This consistency lends credibility to the estimated doses and findings reported in this study. Despite the high thyroid doses received (mean, 130 Gy), we found no evidence of a dose-response relationship with thyroid cancer death. Previous studies have found that children undergoing high-dose radiotherapy for a first primary cancer were at an increased risk of developing secondary thyroid cancer, but the risk was attenuated at higher doses (>30 Gy), presumably because of cell killing. Whether malignant transformation of residual thyroid cells is possible at the higher thyroid doses received by patients with hyperthyroidism, who are generally exposed in adulthood when the thyroid gland is less susceptible to radiation exposure, remains unclear. The ability to evaluate dose-response relationships for cancers that are rarely fatal, such as thyroid cancer, in this study was limited as the cohort was followed for mortality outcomes only. Studies with high-quality dosimetry and cancer incidence follow-up are needed to better characterize the risk of thyroid cancer in adults and children after RAI treatment for hyperthyroidism.

Introduction Previous studies suggested that the presence of myocardial ischemia during stress testing and ambulatory electrocardiographic monitoring indicated an increased risk of cardiac events. Some studies also suggested that the severity and extent of abnormalities demonstrated during resting myocardial perfusion single-photon emission computed tomography as well as the amount of stress-induced ischemia were associated with adverse cardiovascular outcomes. Consequently, the severity of ischemia is generally accepted to be one of the indications for revascularization. However, contemporary data from the Clinical Outcomes Utilizing Revascularization and Aggressive Drug Evaluation (COURAGE) trial substudy indicated that the presence of ischemia did not alter treatment effectiveness. In COURAGE, patients with stable coronary artery disease (CAD) received optimal medical therapy (OMT) with or without percutaneous coronary intervention (PCI). Of note, in the nuclear substudy of COURAGE, at least moderate ischemia at baseline was not associated with reduction of death or myocardial infarction (MI) from PCI added to OMT.

s with stable coronary artery disease (CAD) received optimal medical therapy (OMT) with or without percutaneous coronary intervention (PCI). Of note, in the nuclear substudy of COURAGE, at least moderate ischemia at baseline was not associated with reduction of death or myocardial infarction (MI) from PCI added to OMT. Moreover, a recent meta-analysis that included patients with stable CAD and ischemia documented by stress testing or fractional flow reserve and compared hard end points (death, nonfatal MI, unplanned revascularization, or angina) from PCI and OMT vs OMT alone did not show any benefits from PCI in this subset of patients with CAD. Consequently, whether the presence of ischemia documented during stress testing is associated with major cardiovascular events, regardless of the treatment applied, remains unproven. To our knowledge, few studies have assessed the association of documented ischemia with long-term cardiovascular outcomes in patients with stable CAD, especially considering OMT. The ongoing National Heart, Lung, and Blood Institute–sponsored International Study of Comparative Health Effectiveness with Medical and Invasive Approaches (ISCHEMIA) trial is currently investigating this issue. In addition, no study evaluated the possible effects of chronic ischemia on the evolution of left ventricular function.

National Heart, Lung, and Blood Institute–sponsored International Study of Comparative Health Effectiveness with Medical and Invasive Approaches (ISCHEMIA) trial is currently investigating this issue. In addition, no study evaluated the possible effects of chronic ischemia on the evolution of left ventricular function. The second Medical, Angioplasty, or Surgery Study (MASS II) is a randomized clinical trial designed to compare the long-term effects of OMT, PCI, or coronary artery bypass graft (CABG) surgery in patients with stable multivessel CAD and preserved systolic ventricular function who are appropriate candidates for any of the 3 therapies. The present study is a post hoc analysis of MASS II to evaluate the association of stress-induced documented ischemia with the occurrence of major adverse cardiovascular events (MACEs) and the evolution of the left ventricular ejection fraction (LVEF) in patients with multivessel CAD and initially preserved ventricular function, 10 years after aggressive OMT with or without PCI or CABG. Methods Study Design This was a prospective cohort study using data from the randomized clinical trial MASS II (isrctn.org identifier: ISRCTN66068876). Patients were enrolled in MASS II between May 1, 1995, and May 31, 2000; the present analysis used data collected at the 10-year follow-up.

The second Medical, Angioplasty, or Surgery Study (MASS II) is a randomized clinical trial designed to compare the long-term effects of OMT, PCI, or coronary artery bypass graft (CABG) surgery in patients with stable multivessel CAD and preserved systolic ventricular function who are appropriate candidates for any of the 3 therapies. The present study is a post hoc analysis of MASS II to evaluate the association of stress-induced documented ischemia with the occurrence of major adverse cardiovascular events (MACEs) and the evolution of the left ventricular ejection fraction (LVEF) in patients with multivessel CAD and initially preserved ventricular function, 10 years after aggressive OMT with or without PCI or CABG. Methods Study Design This was a prospective cohort study using data from the randomized clinical trial MASS II (isrctn.org identifier: ISRCTN66068876). Patients were enrolled in MASS II between May 1, 1995, and May 31, 2000; the present analysis used data collected at the 10-year follow-up. MASS II Patient Selection Patients with angiographically documented, proximal multivessel coronary stenosis of more than 70% by visual assessment and documented ischemia were considered for inclusion. Ischemia was documented by either exercise stress testing (EST) or the typical stable angina assessment of the Canadian Cardiovascular Society (class II or III). Patients were enrolled and randomized if the surgeons, attending physicians, and interventional cardiologists agreed that revascularization could be attained by either PCI or CABG.

d by either exercise stress testing (EST) or the typical stable angina assessment of the Canadian Cardiovascular Society (class II or III). Patients were enrolled and randomized if the surgeons, attending physicians, and interventional cardiologists agreed that revascularization could be attained by either PCI or CABG. The trial was approved by the ethics committee of the Heart Institute (InCor) of the University of São Paulo Medical School in São Paulo, Brazil, and all procedures were performed in accordance with the Declaration of Helsinki. Written informed consent was obtained from patients, who were randomly assigned to a treatment group. The clinical criteria for exclusion were mandatory revascularization, left main coronary artery stenosis of 50% or more, ventricular aneurysm that required surgical repair, LVEF less than 40%, a history of PCI or CABG, congenital or valvular heart disease, or another coexisting condition that was a contraindication for CABG or PCI. Treatment Interventions In MASS II, all patients were placed on an OMT until the end of the follow-up. Patients were randomized to either continue with aggressive OMT alone or undergo PCI or CABG in addition to OMT. The use of drugs with known cardiovascular benefits was adjusted throughout the study according to treatment guidelines. Patients were also encouraged to stop smoking, improve their dietary behaviors, and exercise regularly.

domized to either continue with aggressive OMT alone or undergo PCI or CABG in addition to OMT. The use of drugs with known cardiovascular benefits was adjusted throughout the study according to treatment guidelines. Patients were also encouraged to stop smoking, improve their dietary behaviors, and exercise regularly. Investigators were required to perform optimum coronary revascularization. For patients assigned to PCI, the procedure was available within 3 weeks after randomization and was performed according to a standard protocol. Techniques used for catheter-based therapy included bare metal stent placement, laser angioplasty, directional atherectomy, and balloon angioplasty. For patients assigned to CABG, the procedures were available within 12 weeks after randomization, and revascularization was performed with standard surgical techniques using saphenous vein grafts, internal mammary arteries, and other arterial conduits. No off-pump CABG was performed. Follow-up Adverse and other clinical events were tracked from randomization. Patients were evaluated and angina symptoms were graded, with follow-up visits every 6 months for at least 10 years at the Heart Institute. Patients underwent regular electrocardiography and routine blood tests. Myocardial infarction was defined as the presence of significant new Q waves in at least 2 electrocardiogram leads or symptoms compatible with MI associated with creatine kinase MB fraction concentrations that were more than 3 times the upper limit of the reference range.

Follow-up Adverse and other clinical events were tracked from randomization. Patients were evaluated and angina symptoms were graded, with follow-up visits every 6 months for at least 10 years at the Heart Institute. Patients underwent regular electrocardiography and routine blood tests. Myocardial infarction was defined as the presence of significant new Q waves in at least 2 electrocardiogram leads or symptoms compatible with MI associated with creatine kinase MB fraction concentrations that were more than 3 times the upper limit of the reference range. The predefined primary composite end point was the incidence of total mortality, Q-wave MI, or refractory angina that required revascularization. Exercise Stress Testing A symptom-limited treadmill EST according to the modified Bruce Protocol was performed on participants in MASS II before randomization to one of the proposed treatments (OMT, PCI, or CABG) and thereafter every year until the end of the study unless contraindicated. We considered stress-induced ischemia as the presence of exertional angina or an ST-segment depression (horizontal or downsloping of 1 mm for men and 2 mm for women) at 0.08 second after the J point. LVEF Assessment Participants in the MASS II trial underwent transthoracic echocardiography in 2 different periods: before randomization and after 10 years of follow-up. All echocardiographic factors assessed were predefined, and images were analyzed in a core laboratory by expert physicians.

Exercise Stress Testing A symptom-limited treadmill EST according to the modified Bruce Protocol was performed on participants in MASS II before randomization to one of the proposed treatments (OMT, PCI, or CABG) and thereafter every year until the end of the study unless contraindicated. We considered stress-induced ischemia as the presence of exertional angina or an ST-segment depression (horizontal or downsloping of 1 mm for men and 2 mm for women) at 0.08 second after the J point. LVEF Assessment Participants in the MASS II trial underwent transthoracic echocardiography in 2 different periods: before randomization and after 10 years of follow-up. All echocardiographic factors assessed were predefined, and images were analyzed in a core laboratory by expert physicians. Left ventricular ejection fraction was measured by the biplane method (also known as the Simpson method) when the endocardial border of the left ventricle was well defined and whenever regional wall-motion abnormalities were present or, alternatively, by the Teichholz method. Statistical Analysis Data analysis was performed from February 1, 2016, to April 1, 2017. Baseline characteristics were summarized for all patients as percentages for categorical variables and as means with SDs for continuous variables. Means were compared using the unpaired t test for parametric variables and the Mann-Whitney test for nonparametric variables. Tests were 2-sided. The homogeneity between proportions was evaluated using the χ2 or the Fisher exact test.

percentages for categorical variables and as means with SDs for continuous variables. Means were compared using the unpaired t test for parametric variables and the Mann-Whitney test for nonparametric variables. Tests were 2-sided. The homogeneity between proportions was evaluated using the χ2 or the Fisher exact test. The event-free survival time was defined as the interval between random assignment and the occurrence of the first of the components of a primary end point or the latest follow-up. Event-free survival was estimated by the Kaplan-Meier method, and differences among groups were assessed with the log-rank test. Finally, multivariable Cox regression models were calculated to assess the relationship between ischemia and the primary composite end point by adjusting for sex, age, history of MI, number of diseased vessels, and initial CAD treatments. All data were analyzed according to the intention-to-treat principle, and values of P < .05 were considered statistically significant. Statistical analysis was performed with SPSS, version 17.0 for Windows (IBM Corp). Results Baseline Clinical Characteristics Between May 1, 1995, and May 31, 2000, 611 eligible patients who met all entry criteria were randomly assigned to 1 of 3 therapeutic strategies: PCI, OMT, or CABG. The vital status of all randomly assigned patients was determined on February 28, 2011. For patients still alive, the minimum length of follow-up was 10 years, and the maximum was 15 years (mean [SD], 11.4 [4.3] years). No patient was lost during follow-up.

assigned to 1 of 3 therapeutic strategies: PCI, OMT, or CABG. The vital status of all randomly assigned patients was determined on February 28, 2011. For patients still alive, the minimum length of follow-up was 10 years, and the maximum was 15 years (mean [SD], 11.4 [4.3] years). No patient was lost during follow-up. In all, 535 patients were randomized. Of these, 373 (69.7%) were men, 162 (30.3%) were women, and the mean (SD) age for the entire cohort was 59.7 (9.2) years. Randomization created balanced treatment groups (176 in OMT, 180 in PCI, and 179 in CABG) with respect to important prognostic characteristics, except for the occurrence of previous MI (more frequent in the PCI group) and total cholesterol levels (higher in the PCI group; eTable in the Supplement). Before randomization, 535 patients underwent the EST, among whom 270 (50.5%) had stress-induced ischemia whereas 265 (49.5%) did not. None of the characteristics were significantly different between the 2 groups, except for previous MI (more frequent in patients without ischemia) (Table 1). Treatment allocation was similar among those with and without ischemia (80 patients [30.0%] with stress-induced ischemia vs 95 [35.8%] without for OMT, 91 [33.7%] vs 89 [33.6%] for PCI, and 98 [36.3%] vs 81 [30.6%] for CABG, P = .26).

roups, except for previous MI (more frequent in patients without ischemia) (Table 1). Treatment allocation was similar among those with and without ischemia (80 patients [30.0%] with stress-induced ischemia vs 95 [35.8%] without for OMT, 91 [33.7%] vs 89 [33.6%] for PCI, and 98 [36.3%] vs 81 [30.6%] for CABG, P = .26). Table 1. Characteristics of the Population Stratified by Baseline Stress Test Results Characteristic Stress-Induced Ischemia (n = 270)a No Stress-Induced Ischemia (n = 265)a P Value Demographic variables Age, mean (SD), y 59.6 (9.3) 59.6 (8.9) .89 Age ≥65 y 79 (29.3) 94 (35.5) .15 Male 193 (71.5) 180 (68.0) .42 Female 77 (28.5) 85 (32.0) .53 Clinical history and status Current or past smoker 89 (33.0) 95 (35.8) .54 Hypertension 164 (60.7) 150 (56.6) .38 Diabetes 95 (35.2) 103 (38.9) .43 History of MI 102 (37.8) 131 (49.4) .008 Laboratory values, mean (SD), mg/dL Total cholesterol 225.1 (46.8) 222.5 (47.8) .32 LDL cholesterol 150.7 (41.3) 145.3 (42.5) .17 HDL cholesterol 38.2 (10.4) 36.7 (10.2) .04 Triglycerides 186.8 (102.0) 198.4 (115.2) .34 Angiographic profile Double-vessel disease 113 (41.8) 111 (41.9) >.99 Triple-vessel disease 157 (58.1) 154 (58.1) Proximal LAD disease 243 (90.0) 238 (89.8) CAD treatment OMT 81 (30.0) 95 (35.8) .26 PCI 91 (33.7) 89 (33.6) CABG 98 (36.3) 81 (30.6) Abbreviations: CABG, coronary artery bypass graft; CAD, coronary artery disease; HDL, high-density lipoprotein; LAD, left anterior descending coronary artery; LDL, low-density lipoprotein; MI, myocardial infarction; OMT, optimal medical therapy; PCI, percutaneous coronary intervention.

9 (33.6) CABG 98 (36.3) 81 (30.6) Abbreviations: CABG, coronary artery bypass graft; CAD, coronary artery disease; HDL, high-density lipoprotein; LAD, left anterior descending coronary artery; LDL, low-density lipoprotein; MI, myocardial infarction; OMT, optimal medical therapy; PCI, percutaneous coronary intervention. SI conversion factors: To convert total, LDL, and HDL cholesterol to millimoles per liter, multiply by 0.0259; triglycerides to millimoles per liter, multiply by 0.0113. a Values are expressed as number (percentage) unless otherwise indicated. In addition, 472 patients (88.2%) had a history of exertional angina and 63 (11.8%) did not. Of the 472 patients with angina symptoms, 246 (52.1%) had positive EST results, 88 (18.6%) had negative results, and 138 (29.2%) had inconclusive results. Of the 63 patients without any angina symptoms, 24 (38.1%) had positive EST results, 14 (22.2%) had negative results, and 25 (39.7%) had inconclusive results. Exercise Stress Testing Analysis The overall major cardiac events at the 10-year follow-up of patients stratified by the results of the baseline stress tests are provided in Table 2. No association was found between the presence of ischemia at baseline and the occurrence of MACE.

In addition, 472 patients (88.2%) had a history of exertional angina and 63 (11.8%) did not. Of the 472 patients with angina symptoms, 246 (52.1%) had positive EST results, 88 (18.6%) had negative results, and 138 (29.2%) had inconclusive results. Of the 63 patients without any angina symptoms, 24 (38.1%) had positive EST results, 14 (22.2%) had negative results, and 25 (39.7%) had inconclusive results. Exercise Stress Testing Analysis The overall major cardiac events at the 10-year follow-up of patients stratified by the results of the baseline stress tests are provided in Table 2. No association was found between the presence of ischemia at baseline and the occurrence of MACE. Table 2. Major Adverse Cardiac Events at 10 Years Stratified by Baseline Stress Test Results Outcomes Stress-Induced Ischemia (n = 270)a No Stress-Induced Ischemia (n = 265)a Hazard Ratio (95% CI) P Valueb Overall mortality 80 (29.6) 84 (31.7) 0.80 (0.58-1.11) .18 Acute myocardial infarction 33 (12.2) 41 (15.5) 0.85 (0.54-1.36) .51 Additional intervention 87 (32.2) 68 (25.7) 1.26 (0.92-1.74) .15 Primary end pointc 153 (56.7) 145 (54.7) 1.00 (0.80-1.27) .95 a Values are expressed as number (percentage) unless otherwise indicated. b Cox proportional hazards models adjusted for sex, age, 2- or 3-vessel coronary artery disease, coronary artery disease treatments, and previous myocardial infarction. c The primary end point was the occurrence of the first clinical event (acute myocardial infarction, additional intervention, or overall mortality).

Table 2. Major Adverse Cardiac Events at 10 Years Stratified by Baseline Stress Test Results Outcomes Stress-Induced Ischemia (n = 270)a No Stress-Induced Ischemia (n = 265)a Hazard Ratio (95% CI) P Valueb Overall mortality 80 (29.6) 84 (31.7) 0.80 (0.58-1.11) .18 Acute myocardial infarction 33 (12.2) 41 (15.5) 0.85 (0.54-1.36) .51 Additional intervention 87 (32.2) 68 (25.7) 1.26 (0.92-1.74) .15 Primary end pointc 153 (56.7) 145 (54.7) 1.00 (0.80-1.27) .95 a Values are expressed as number (percentage) unless otherwise indicated. b Cox proportional hazards models adjusted for sex, age, 2- or 3-vessel coronary artery disease, coronary artery disease treatments, and previous myocardial infarction. c The primary end point was the occurrence of the first clinical event (acute myocardial infarction, additional intervention, or overall mortality). Figure 1 shows the Kaplan-Meier survival curves of the occurrence of cardiovascular events in patients stratified by the presence of stress-induced ischemia. No significant differences were found between patients with or without ischemia regarding survival free of the combined cardiovascular end points (hazard ratio [HR], 0.96; 95% CI, 0.76-1.21; P = .73). After adjusting for baseline prognostic variables (sex, age, previous MI, number of diseased vessels, and initial CAD treatments), the presence of stress-induced ischemia was not found to be associated with the occurrence of cardiovascular events (HR, 1.00; 95% CI, 0.80-1.27; P = .95).

.96; 95% CI, 0.76-1.21; P = .73). After adjusting for baseline prognostic variables (sex, age, previous MI, number of diseased vessels, and initial CAD treatments), the presence of stress-induced ischemia was not found to be associated with the occurrence of cardiovascular events (HR, 1.00; 95% CI, 0.80-1.27; P = .95). Figure 1. Kaplan-Meier Curves for Survival Free of Cardiovascular Events According to Stress-Induced Ischemic Status at Baseline HR indicates hazard ratio. In addition, among patients with and without stress-induced ischemia, survival free of cardiovascular events was more adverse in OMT and PCI groups, compared to that in the CABG group (eFigure 1 in the Supplement). The pairwise treatment comparisons of the occurrence of the primary composite end point demonstrated no differences between the OMT and PCI groups in patients with (HR, 0.81; 95% CI, 0.55-1.18; P = .27) and without (HR, 0.87; 95% CI, 0.60-1.28; P = .48) stress-induced ischemia. LVEF Assessment Of the 535 patients who underwent EST, 320 had their ventricular function assessed with echocardiography at baseline and after the 10-year follow-up. The echocardiographic assessment identified similarities in LVEF evolution among patients with and without stress-induced ischemia. Irrespective of the ischemic status, both groups exhibited a slight decline in LVEF, which is represented by the difference in reduction (median [SD], −4.9% [18.7%] vs −6.6% [20.0%], respectively, for groups with and without ischemia; P = .97) (Figure 2).

olution among patients with and without stress-induced ischemia. Irrespective of the ischemic status, both groups exhibited a slight decline in LVEF, which is represented by the difference in reduction (median [SD], −4.9% [18.7%] vs −6.6% [20.0%], respectively, for groups with and without ischemia; P = .97) (Figure 2). Figure 2. Changes in Left Ventricular Ejection Fraction (LVEF) According to the Presence of Stress-Induced Ischemia The delta LVEF (difference in reduction of LVEF) is calculated as 100 × [(LVEF at 10 years − LVEF at baseline)/LVEF at baseline]. The vertical line in the middle of each box indicates the median; left and right borders of the box, the interquartile ranges; whiskers, the maximum and minimum values excluding the outliers; and data points beyond the whiskers, outliers. The outcomes of the different treatment strategies were analyzed in the subgroups of patients with and without stress-induced ischemia. In both conditions, regardless of the treatment applied, patients experienced the same pattern and magnitude of ventricular function variation, with a slight decline in LVEF represented by the difference in reduction for those without stress-induced ischemia (eFigure 2 in the Supplement).

thout stress-induced ischemia. In both conditions, regardless of the treatment applied, patients experienced the same pattern and magnitude of ventricular function variation, with a slight decline in LVEF represented by the difference in reduction for those without stress-induced ischemia (eFigure 2 in the Supplement). Discussion This study evaluated the occurrence of cardiovascular events and changes in left ventricular function in patients (participants in the MASS II randomized clinical trial) with multivessel CAD who underwent 1 of 3 therapeutic strategies (alone or with PCI or CABG) according to the presence or absence of exercise stress-induced ischemia at baseline. The results show that the presence or absence of documented myocardial ischemia appears to have no association with long-term cardiovascular outcomes and changes in LVEF in patients with stable multivessel CAD and preserved ventricular function. Results of previous trials presumed the presence of myocardial ischemia based on symptoms, ischemic test results, and coronary anatomy findings. Their findings suggest that myocardial ischemia might not play a role as an additional risk factor for events in patients with stable CAD irrespective of the treatment applied. However, the influence of objective, documented ischemia on treatment outcomes is not well established.

results, and coronary anatomy findings. Their findings suggest that myocardial ischemia might not play a role as an additional risk factor for events in patients with stable CAD irrespective of the treatment applied. However, the influence of objective, documented ischemia on treatment outcomes is not well established. One of the first studies that assessed the comparative results of CAD treatments in patients with documented ischemia was conducted by Hachamovitch et al. In their analysis of registry data, the authors compared OMT with revascularization strategies in patients who underwent myocardial perfusion stress tests. Their findings showed that patients with no or mild ischemia in the OMT group had a survival benefit, whereas those with moderate to severe ischemia had better survival outcomes if they were revascularized. Despite the interesting results, major methodological concerns, such as the differences in baseline variables between groups, no defined medical therapy, the lack of an indication for revascularization procedures, and selection bias, compromised the interpretation of the study findings.

mes if they were revascularized. Despite the interesting results, major methodological concerns, such as the differences in baseline variables between groups, no defined medical therapy, the lack of an indication for revascularization procedures, and selection bias, compromised the interpretation of the study findings. On the other hand, contemporary randomized studies have shown different results regarding the association of ischemia with outcomes and treatment effects. The COURAGE nuclear substudy performed serial rest/stress myocardial perfusion scintigraphy before treatment and 6 to 18 months after randomization in 314 of the 2287 participants in the original trial. The authors observed that the reduction in ischemia achieved more frequently in PCI added to the OMT group was not associated with a lower risk of death or MI in the adjusted analysis. Another substudy of the COURAGE trial including 1381 participants found similar event rates with both treatments irrespective of the extent and severity of ischemia at baseline. Therefore, results from the COURAGE trial are similar to those of the present analysis.

a lower risk of death or MI in the adjusted analysis. Another substudy of the COURAGE trial including 1381 participants found similar event rates with both treatments irrespective of the extent and severity of ischemia at baseline. Therefore, results from the COURAGE trial are similar to those of the present analysis. In the Fractional Flow Reserve vs Angiography for Multivessel Evaluation (FAME) 2 trial, which studied patients with functionally significant stenosis, as determined by a measurement of fractional flow reserve less than 0.80, the addition of PCI to OMT was evaluated for the prevention of the primary composite end point (death, MI, or urgent revascularization). The researchers found that fractional flow reserve–guided PCI plus OMT, compared with OMT alone, was associated less frequently with the primary composite end point. However, contributing to the result was the decreased need for urgent revascularization in the PCI group. In addition, the unblinded nature of FAME 2 may have biased the results in favor of more urgent revascularizations in patients randomized to OMT. Of note, in the subset of patients with left ventricular dysfunction, a substudy of the Surgical Treatment for Ischemic Heart Failure (STICH) trial also demonstrated that inducible myocardial ischemia did not identify patients with worse prognosis or those with greater benefit from CABG over OMT. The STICH trial randomized patients with CAD and an ejection fraction less than or equal to 35% to CABG or OMT.

urgical Treatment for Ischemic Heart Failure (STICH) trial also demonstrated that inducible myocardial ischemia did not identify patients with worse prognosis or those with greater benefit from CABG over OMT. The STICH trial randomized patients with CAD and an ejection fraction less than or equal to 35% to CABG or OMT. Supporting the findings of these later studies, a meta-analysis of contemporary trials compared the clinical outcomes of PCI and OMT vs those of OMT alone exclusively in patients with stable CAD and documented myocardial ischemia. The authors concluded that, in patients with documented ischemia, PCI was not able to reduce cardiovascular events. A prior study of the MASS group among patients with multivessel CAD, evaluated the evolution of LVEF after 10 years of follow-up and found no difference among the 3 treatment groups. Also, the subgroup of patients with and without stress-induced ischemia demonstrated by EST at the end of follow-up had the same evolution of ventricular function. In the present study, the presence of ischemia in EST at baseline was not associated with worse evolution of LVEF.

difference among the 3 treatment groups. Also, the subgroup of patients with and without stress-induced ischemia demonstrated by EST at the end of follow-up had the same evolution of ventricular function. In the present study, the presence of ischemia in EST at baseline was not associated with worse evolution of LVEF. Although the present study did not have data about high-risk findings in EST, more than 50% of the studied patients had triple-vessel disease and almost 90% had proximal left anterior descending coronary artery involvement, which characterize a higher-risk population. Despite this high risk, the results of this study suggest that, in patients with stable CAD, documented ischemia observed by ischemic changes during EST was not associated with different 10-year cardiovascular outcomes and worsening of ventricular function compared with that in patients with nonischemic EST. Although the presence of documented ischemia has been identified as a possible marker of a higher-risk population and an indication for myocardial revascularization procedures to protect the myocardium from the chronic, deleterious effects of ischemia over time, the present study’s findings do not support this assumption. The delicate imbalance between oxygen supply and demand at stress is a consequence and does not seem to be a factor for impairment of ventricular function or coronary events during a long-term follow-up.

the chronic, deleterious effects of ischemia over time, the present study’s findings do not support this assumption. The delicate imbalance between oxygen supply and demand at stress is a consequence and does not seem to be a factor for impairment of ventricular function or coronary events during a long-term follow-up. A possible physiopathological explanation for this finding could be that the functional information about myocardial stress-induced ischemia may not indicate atherosclerotic plaque instability, which is the major factor responsible for clinical events. Since myocardial ischemia is associated with the degree of stenosis and atherosclerotic burden, we must consider that its presence, even if significant, may remain stable for a long time while the plaque stability persists. Thus, the results of such functional tests should be interpreted cautiously, and not isolated from other clinical information when revascularization strategies are being considered for the treatment of CAD patients. The results from the ongoing ISCHEMIA trial are expected to introduce new information regarding this subject.

the results of such functional tests should be interpreted cautiously, and not isolated from other clinical information when revascularization strategies are being considered for the treatment of CAD patients. The results from the ongoing ISCHEMIA trial are expected to introduce new information regarding this subject. Limitations and Strengths This study has limitations. First, because this was a retrospective analysis, all aspects covered are related to this type of study. However, all data were derived from a randomized clinical trial, with predictors and outcome variables collected prospectively in a detailed database. Second, medical therapy changed during the evolution of the trial, which may have influenced the findings; such evolution is inherent in long-term follow-up studies. The changes noted occurred in all study patients, irrespective of the group they were placed in later. Last, this was a single-center study, which may limit the external validation of the analysis. However, the homogeneity of medical conduct, medical therapy, and indications for coronary interventions, especially during the long-term follow-up, reduce the limitations of the present study. Conclusions In this study, regardless of the therapeutic strategy applied, the presence of documented ischemia did not appear to be associated with the occurrence of major adverse events or changes in left ventricular function among the participants selected from the MASS II trial. Supplement. eTable. Baseline Characteristics of the 535 Patients With Exercise Stress Tests According to Treatment in MASS II Trial

Conclusions In this study, regardless of the therapeutic strategy applied, the presence of documented ischemia did not appear to be associated with the occurrence of major adverse events or changes in left ventricular function among the participants selected from the MASS II trial. Supplement. eTable. Baseline Characteristics of the 535 Patients With Exercise Stress Tests According to Treatment in MASS II Trial eFigure 1. Survival Free of Cardiovascular Events According to Initial Assigned Treatment in Patients With (1A) and Without (1B) Stress Induced Ischemia eFigure 2. Changes in Left Ventricular Ejection Fraction According to the Presence or Absence of Myocardial Ischemia in Each Treatment Group After 10-Year Follow-up Click here for additional data file.

icines, and income); the overall test of including effect modification was not statistically significant; and the estimate of the effect of the intervention on adherence was similar without adjustment for baseline characteristics (odds ratio, 1.7; 95% CI, 1.3-2.3) and with adjustment (odds ratio, 1.7; 95% CI, 1.2-2.3). Table 3. Primary and Secondary Outcome Results by Group Outcome Free Distribution Group (n = 395) Usual Access Group (n = 391) Difference, % (95% CI) P Value Primary outcome, No. (%) Participants appropriately adherent to all medicines 151 (38.2) 104 (26.6) 11.6 (4.9 to 18.4) <.001 Secondary outcomes, % Mean % of medicines adhered to by each participanta 66.1 56.4 7.2 (1.1 to 14.0) .02 Mean % of medicines potentially inappropriately prescribed to each participanta 0.17 0.85 −0.66 (−0.79 to −0.33) .007 a Differences estimated from rate ratios and estimated mean percentage in control group. Rate ratio for medicines adhered to was 1.13 (95% CI, 1.02-1.25). Rate ratio for potentially inappropriate prescriptions was 0.22 (95% CI, 0.064-0.60).

Introduction Lifesaving medicines such as treatments for cardiovascular disease and HIV and AIDS are often not accessible because of the cost to patients, among other reasons. Global estimates of medication nonadherence are between 40% and 60% for treatments for chronic diseases. A systematic review of interventions to improve adherence identified few effective interventions and underscored the need for additional high-quality studies. Providing certain medicines at no charge for people with specific conditions such as myocardial infarction, hypertension, HIV and AIDS, and schizophrenia can improve surrogate and direct health measures related to those conditions. The World Health Organization recommends that countries develop a list of essential medicines “that satisfy the priority health care needs of the population” for the purpose of increasing access and quality of care, and facilitating medicine access is one way to help achieve universal health coverage.

those conditions. The World Health Organization recommends that countries develop a list of essential medicines “that satisfy the priority health care needs of the population” for the purpose of increasing access and quality of care, and facilitating medicine access is one way to help achieve universal health coverage. Previous studies such as the RAND Health Insurance Experiment did not isolate the effect of medicine access from that of health care access. Canada is a suitable setting to measure the effects of medicine distribution because physician care and hospitalizations are universally publicly funded, while medicines are not. Public coverage of medicines used outside of Canadian hospitals varies by province, but is most frequently offered for specific groups, such as people receiving social assistance and those older than 65 years. Approximately 55% of people in Ontario, Canada, have employer-based private insurance plans that typically cover medicines with copayments or deductibles, while approximately 28% have public insurance that also involve copayments and deductibles. We conducted a randomized clinical trial examining the effect of providing Canadian outpatients who reported not being able to afford medicines with free distribution of essential medicines. Because the provided medicines included only treatments that have been well established as effective—such as those for cardiovascular disease, HIV and AIDS, and pneumonia—adherence was chosen as the primary outcome, and surrogate health outcomes were the secondary outcomes.

ith free distribution of essential medicines. Because the provided medicines included only treatments that have been well established as effective—such as those for cardiovascular disease, HIV and AIDS, and pneumonia—adherence was chosen as the primary outcome, and surrogate health outcomes were the secondary outcomes. Methods Study Design The Carefully Selected and Easily Accessible at No Charge Medicines (CLEAN Meds) trial is a multicenter, unblinded, parallel, 2-group, superiority, outcomes assessor–blinded, individually randomized clinical trial with 1:1 allocation conducted at 9 primary care sites in Ontario, Canada, that enrolled patients between June 1, 2016, and April 28, 2017. The trial protocol is available in Supplement 1. The trial was registered with ClinicalTrials.gov (NCT02744963) and a protocol has been published. The trial is reported in accordance with the 2010 CONSORT statement and the intervention is described using the TIDieR (Template for Intervention Description and Replication) checklist. After registration with ClinicalTrials.gov, publication of the protocol, and initiation of the study, the duration of the trial was extended from 12 to 24 months when additional funding was obtained (trial protocol in Supplement 1). For the analysis of the primary outcome, we originally planned to use electronic pill bottle cap devices in one-seventh of participants to confirm adherence measurements, but owing to a large amount of missing data in both groups we removed this measure from the definition of the primary outcome; we report the results when the available electronic pill bottle cap device data were used in eTable 1 in Supplement 2. A data and safety monitoring board met to ensure that medication incidents were properly addressed and that the intervention was not harming participants. Ethics approval for the conduct of this study was obtained from the St Michael’s Research Ethics Board, the Huron Shores Family Health Team Research Ethics Committee, and the Laurentian University Research Ethics Board. All enrolled participants provided written informed consent.

ervention was not harming participants. Ethics approval for the conduct of this study was obtained from the St Michael’s Research Ethics Board, the Huron Shores Family Health Team Research Ethics Committee, and the Laurentian University Research Ethics Board. All enrolled participants provided written informed consent. Patients Patients 18 years or older who self-reported medication nonadherence owing to cost in the last 12 months were eligible for inclusion. After potentially eligible participants were identified by clinicians at routine visits, study personnel asked a question adapted from the Commonwealth Fund International Health Policy Survey to confirm cost-related nonadherence: “In the last twelve months, did you not fill a prescription or do anything to make a prescription last longer because of the cost?”(p30) We excluded family members living at the same address as participants already enrolled in the study to avoid contamination and excluded patients who joined the clinic within the last 6 months to deter patients from joining the practice to enroll in the study. Trial Procedures Patients who met the eligibility criteria were randomly allocated to 1 of 2 groups. The intervention group received free distribution of essential medicines. The control group accessed medications as usual. Randomization and allocation concealment were achieved through a web-based tool that was accessed through the REDCap electronic case report forms application and was stratified by center and blocked using permuted blocks of 2 to 4.

ee distribution of essential medicines. The control group accessed medications as usual. Randomization and allocation concealment were achieved through a web-based tool that was accessed through the REDCap electronic case report forms application and was stratified by center and blocked using permuted blocks of 2 to 4. Patients in the intervention group received free distribution of medicines on a list of essential medicines as well as otherwise usual care. The list of 128 essential medicines was adapted from the 2013 World Health Organization Model List of Essential Medicines based on Canadian clinical practice guidelines, suggestions from clinicians and patients, prescribing volumes, and evidence syntheses (eAppendix 1 in Supplement 2). The medicines included treatments for acute conditions (eg, antibiotics and analgesics) and chronic conditions (eg, antipsychotics, antiretrovirals, glucose-lowering medicines, and antihypertensives). Patients could be prescribed medicines that were not on the essential list and could access them in the usual way (eg, by paying for them). Participants could be switched from a medicine not on the list to an equivalent that was on the list. Medicine distribution was primarily through the mail to an address of the participant’s choice. A community pharmacist (N.U.; with a Bachelor of Science in Pharmacy and more than 15 years of experience) contacted patients by telephone from a pharmacy established for the study. Medicines that needed to be started in a timely fashion (eg, anti-infectives, analgesics, diuretics, bronchodilators, antihypertensives, and antipsychotics) were also stored at each study site for dispensing by clinicians. Controlled substances (eg, opioids, sedatives, and stimulants) were not included in the intervention; patients accessed these medications in the usual fashion.

nti-infectives, analgesics, diuretics, bronchodilators, antihypertensives, and antipsychotics) were also stored at each study site for dispensing by clinicians. Controlled substances (eg, opioids, sedatives, and stimulants) were not included in the intervention; patients accessed these medications in the usual fashion. Participants allocated to the control group had their usual access to medicines. Typical annual out-of-pocket costs to participants in the control group were the full cost of medicines for those with no insurance (eg, approximately $800, as well as dispensing fees of approximately $10 per prescription for a patient taking oral diabetes medicines), a deductible for an older adult with public drug coverage (eg, $100 plus copayments of $4 for each prescription filled), or a percentage of total medicine costs for those with private insurance (eg, $160 = $800 × 20%, assuming 80% coverage). Medicines were generally dispensed by community pharmacies, some of which offer local delivery services for medicines on request.

verage (eg, $100 plus copayments of $4 for each prescription filled), or a percentage of total medicine costs for those with private insurance (eg, $160 = $800 × 20%, assuming 80% coverage). Medicines were generally dispensed by community pharmacies, some of which offer local delivery services for medicines on request. Outcomes The follow-up period was 12 months. The prespecified primary outcome was adherence to all appropriately prescribed medicines. The primary outcome was determined at 12 months by assessing whether each prescription was both appropriate (based on explicit criteria) and taken as prescribed for at least 80% of expected doses. A participant was either classified as receiving only appropriate prescriptions and being adherent to all of them (evidence of taking at least 80% of expected doses), or as having either received at least 1 potentially inappropriate prescription or being nonadherent to at least 1 medicine. We reviewed primary care prescribing records to determine whether each prescription was potentially inappropriate using established criteria based only on the prescribed medicines (eAppendix 2 in Supplement 2). We considered as appropriate all prescriptions that did not meet criteria for being potentially inappropriate. Two adjudicators independently applied the explicit criteria in a blinded fashion; there were no disagreements. For adherence, we used the lowest estimate from the 2 methods used: reviews of primary care records for prescription renewal intervals during the 12-month study period and patient report of the number of doses missed during the last week, as reported by telephone interview or email survey between 9 and 12 months. Blinded abstraction of medical records was done by 1 adjudicator from a team of 5 and verified by another after a training period for each adjudicator.

12-month study period and patient report of the number of doses missed during the last week, as reported by telephone interview or email survey between 9 and 12 months. Blinded abstraction of medical records was done by 1 adjudicator from a team of 5 and verified by another after a training period for each adjudicator. The prespecified secondary outcomes were the proportion of medicines that were appropriately prescribed, the proportion of medicines that met adherence criteria, hemoglobin A1c levels in patients treated for type 1 or 2 diabetes (adjusted for baseline), blood pressure in patients treated with an antihypertensive (adjusted for baseline), and low-density lipoprotein cholesterol levels in patients treated with a statin (adjusted for baseline). Hemoglobin A1c, blood pressure, and low-density lipoprotein cholesterol levels were obtained by review of medical records during the baseline period (up to 3 months prior to randomization) and at follow-up (9-12 months after randomization). Clinicians ordered investigations as usual in both groups; no instructions related to clinical care were provided by the trial team. To compare experiences between groups, after a 9- to 12-month follow-up period we asked patients 14 questions about their care, medicine dispensing, and social circumstances (eTable 2 in Supplement 2). Serious adverse events, including hospitalizations and deaths, were ascertained through clinician reports, patient reports, and reviews of the primary care medical records.

2-month follow-up period we asked patients 14 questions about their care, medicine dispensing, and social circumstances (eTable 2 in Supplement 2). Serious adverse events, including hospitalizations and deaths, were ascertained through clinician reports, patient reports, and reviews of the primary care medical records. Statistical Analysis The sample size was calculated to detect a 10% absolute difference (80% power, type I error of 5%) in the primary outcome of appropriate adherence and assuming that 40% to 60% of patients in the control group would be appropriately adherent to all medications. A sample size of 392 per group was required. The primary analysis was performed using an intention-to-treat principle. Appropriate adherence was compared using a χ2 test and the unadjusted treatment effect was expressed as the absolute risk difference. For the primary analysis, P < .05 was used to reject the null hypothesis of no difference. We did not correct for multiple comparisons because comparisons other than for the prespecified primary and secondary outcomes are hypothesis generating. For the exploratory subgroup analysis, we fit a logistic regression model to determine whether the following characteristics modified the effect of the intervention: age, sex, urban vs rural site, number of baseline medicines, and income.

an for the prespecified primary and secondary outcomes are hypothesis generating. For the exploratory subgroup analysis, we fit a logistic regression model to determine whether the following characteristics modified the effect of the intervention: age, sex, urban vs rural site, number of baseline medicines, and income. Results Patients Between June 1, 2016, and April 28, 2017, 1130 individuals identified as potentially eligible by clinicians were assessed for eligibility and 786 were randomly allocated (Figure). The characteristics of participants in the 2 groups are summarized in Table 1. For the 22 of 786 participants (2.8%) who withdrew consent—9 of 395 (2.3%) in the free distribution group and 13 of 391 (3.3%) in the usual access group—data collected prior to withdrawal were included in the analysis.

ly allocated (Figure). The characteristics of participants in the 2 groups are summarized in Table 1. For the 22 of 786 participants (2.8%) who withdrew consent—9 of 395 (2.3%) in the free distribution group and 13 of 391 (3.3%) in the usual access group—data collected prior to withdrawal were included in the analysis. Figure. Participant Flow Diagram Table 1. Baseline Participant Characteristics by Group Characteristic Participants, No. (%) Free Distribution Group (n = 395) Usual Access Group (n = 391) Female sex 220 (55.7) 219 (56.0) Age, mean (SD), y 51.0 (14.2) 50.4 (14.3) Age ≥65 y 71 (18.0) 64 (16.4) Race/ethnicity White 256 (64.8) 260 (66.5) Black 35 (8.9) 39 (10.0) Southeast or East Asian (including Korean, Japanese, Filipino, and Chinese) 28 (7.1) 19 (4.9) South Asian 25 (6.3) 24 (6.1) Latin American 10 (2.5) 15 (3.8) Indigenous 12 (3.0) 14 (3.6) West Asian (including Arab) 6 (1.5) 5 (1.3) Mixed or other 22 (5.6) 8 (2.0) Declined to provide 1 (0.3) 7 (1.8) Main income source Wages and salaries (including self-employed) 218 (55.2) 221 (56.5) Pension 50 (12.7) 42 (10.7) Social support (eg, welfare or disability) 36 (9.1) 47 (12.0) Unemployment insurance 15 (3.8) 9 (2.3) Other 56 (14.2) 51 (13.0) Declined to provide 20 (5.1) 21 (5.4) Household income, Can$a <30 000 205 (51.9) 182 (46.5) 30 000-70 000 92 (23.3) 99 (25.3) >70 000 21 (5.3) 22 (5.6) Declined to provide 77 (19.5) 88 (22.5) No. of medicines prescribed at baseline, mean (SD) 5.3 (3.6) 5.6 (4.0) Site Urban 269 (68.1) 267 (68.3) Rural 126 (31.9) 124 (31.7) Prescribed Diabetes treatment 89 (22.5) 91 (23.3) Antihypertensive 122 (30.9) 114 (29.2) Statin 81 (20.5) 81 (20.7) a The median Canadian household income in 2015 was $70 336, and the poverty line was $37 542.

cribed at baseline, mean (SD) 5.3 (3.6) 5.6 (4.0) Site Urban 269 (68.1) 267 (68.3) Rural 126 (31.9) 124 (31.7) Prescribed Diabetes treatment 89 (22.5) 91 (23.3) Antihypertensive 122 (30.9) 114 (29.2) Statin 81 (20.5) 81 (20.7) a The median Canadian household income in 2015 was $70 336, and the poverty line was $37 542. The categories of medicines prescribed were similar between groups (Table 2). Commonly prescribed medicines during the study period included analgesics, diabetes treatments, proton pump inhibitors, treatments for hypertension or vascular disease, and treatments for asthma or chronic obstructive pulmonary disease.

cribed at baseline, mean (SD) 5.3 (3.6) 5.6 (4.0) Site Urban 269 (68.1) 267 (68.3) Rural 126 (31.9) 124 (31.7) Prescribed Diabetes treatment 89 (22.5) 91 (23.3) Antihypertensive 122 (30.9) 114 (29.2) Statin 81 (20.5) 81 (20.7) a The median Canadian household income in 2015 was $70 336, and the poverty line was $37 542. The categories of medicines prescribed were similar between groups (Table 2). Commonly prescribed medicines during the study period included analgesics, diabetes treatments, proton pump inhibitors, treatments for hypertension or vascular disease, and treatments for asthma or chronic obstructive pulmonary disease. Table 2. Participants Prescribed Medicines by Anatomical Therapeutic Chemical Classification System Main Groups Anatomical Therapeutic Chemical Main Group (Examples of Medicines Commonly Prescribed) Prescriptions, No. (%) Free Distribution Group (n = 2071) Usual Access Group (n = 2183) Nervous system (gabapentin, sertraline, venlafaxine, and acetaminophen) 424 (20.5) 450 (20.6) Alimentary tract and metabolism (metformin, pantoprazole, rabeprazole, and insulin) 381 (18.4) 403 (18.5) Cardiovascular system (atorvastatin, ramipril, rosuvastatin, amlodipine, and hydrochlorothiazide) 326 (15.7) 366 (16.8) Respiratory system (albuterol, fluticasone, and tiotropium) 274 (13.2) 264 (12.1) Dermatologic (hydrocortisone and betamethasone) 161 (7.8) 159 (7.3) Blood and blood-forming organs (acetylsalicylic acid and ferrous fumarate) 125 (6.0) 140 (6.4) Musculoskeletal system (naproxen and ibuprofen) 117 (5.6) 128 (5.9) Genitourinary system and sex hormones (estradiol) 116 (5.6) 124 (5.7) Anti-infectives for systemic use (amoxicillin) 86 (4.2) 88 (4.0) Systemic hormonal preparations (levothyroxine) 40 (1.9) 37 (1.7) Other 21 (1.0) 24 (1.1) Adherence Free distribution increased the number of participants in the free distribution group who were appropriately adherent to all medicines, compared with those in the usual access group (151 of 395 [38.2%] vs 104 of 391 [26.6%]; difference, 11.6%; 95% CI, 4.9%-18.4%; P < .001) (Table 3). We performed sensitivity analyses using less stringent definitions and found that the difference was similar in magnitude and remained statistically significant (eTable 1 in Supplement 2). There was little indication of subgroup effects: no effect modification terms were statistically significantly different (age, sex, urban vs rural site, number of baseline medicines, and income); the overall test of including effect modification was not statistically significant; and the estimate of the effect of the intervention on adherence was similar without adjustment for baseline characteristics (odds ratio, 1.7; 95% CI, 1.3-2.3) and with adjustment (odds ratio, 1.7; 95% CI, 1.2-2.3).

opriately prescribed to each participanta 0.17 0.85 −0.66 (−0.79 to −0.33) .007 a Differences estimated from rate ratios and estimated mean percentage in control group. Rate ratio for medicines adhered to was 1.13 (95% CI, 1.02-1.25). Rate ratio for potentially inappropriate prescriptions was 0.22 (95% CI, 0.064-0.60). Secondary Outcomes The proportion of medicines each participant was adherent to was higher in those receiving free distribution than those with usual access (66.1% vs 56.4%; difference, 7.2%; 95% CI, 1.1%-14.0%; P = .02) and the proportion of potentially inappropriately prescribed medicines was lower in those with free distribution than those with usual access (0.17% vs 0.85%; difference, −0.66%; 95% CI, −0.79% to −0.33%; P = .007) (Table 3). Free distribution improved systolic blood pressure among those prescribed an antihypertensive drug compared with those receiving usual access (−7.2 mm Hg; 95% CI, −11.7 to −2.8 mm Hg; P = .002), but did not have a statistically significant effect on the other surrogate health outcomes such as low-density lipoprotein cholesterol level (−2.3 mg/dL; 95% CI, −14.7 to 10.0 mg/dL; P = .70 [to convert to millimoles per liter, multiply by 0.0259]), although there was a nonsignificant decrease in hemoglobin A1c levels in participants in the free distribution group who were prescribed a diabetes treatment (−0.38%; 95% CI, −0.76% to 0.00%; P = .05 [to convert to proportion of total hemoglobin, multiply by 0.01]) (Table 4).

millimoles per liter, multiply by 0.0259]), although there was a nonsignificant decrease in hemoglobin A1c levels in participants in the free distribution group who were prescribed a diabetes treatment (−0.38%; 95% CI, −0.76% to 0.00%; P = .05 [to convert to proportion of total hemoglobin, multiply by 0.01]) (Table 4). Table 4. Secondary Surrogate Health Outcome Results by Group Outcome Free Distribution Group Usual Access Group Hemoglobin A1c, % No. 73 68 Baseline, mean (SD) 8.20 (1.86) 8.15 (1.85) Follow-up, mean (SD) 7.69 (1.50) 8.04 (1.58) Difference (95% CI) −0.38 (−0.76 to 0.00) NA P value .05 NA Systolic blood pressure, mm Hg No. 105 88 Baseline, mean (SD) 137 (19) 135 (17) Follow-up, mean (SD) 132 (16) 139 (19) Difference (95% CI) −7.2 (−11.7 to −2.8) NA P value .002 NA Diastolic blood pressure, mm Hg No. 105 88 Baseline, mean (SD) 81 (13) 81 (11) Follow-up, mean (SD) 78 (12) 80 (13) Difference (95% CI) −2.0 (−5.0 to 1.0) NA P value .19 NA LDL cholesterol level, mg/dL No. 48 40 Baseline, mean (SD) 88.9 (38.7) 77.3 (34.8) Follow-up, mean (SD) 81.2 (34.8) 81.2 (42.5) Difference (95% CI) −2.3 (−14.7 to 10.0) NA P value .70 NA Abbreviations: LDL, low-density lipoprotein; NA, not applicable. SI conversion factors: To convert LDL cholesterol to millimoles per liter, multiply by 0.0259; hemoglobin A1c to proportion of total hemoglobin, multiply by 0.01.

P value .19 NA LDL cholesterol level, mg/dL No. 48 40 Baseline, mean (SD) 88.9 (38.7) 77.3 (34.8) Follow-up, mean (SD) 81.2 (34.8) 81.2 (42.5) Difference (95% CI) −2.3 (−14.7 to 10.0) NA P value .70 NA Abbreviations: LDL, low-density lipoprotein; NA, not applicable. SI conversion factors: To convert LDL cholesterol to millimoles per liter, multiply by 0.0259; hemoglobin A1c to proportion of total hemoglobin, multiply by 0.01. There were statistically significant differences between groups in 10 of the 14 patient-oriented outcomes (eTable 2 in Supplement 2). Compared with the time of enrollment, participants in the free distribution group were more likely than those in the usual care group to report receiving their medicines before the previous prescription ran out (217 of 261 [83.1%] vs 157 of 237 [66.2%]; difference, 16.9%; 95% CI, 9.0%-24.8%), more likely to report that their care improved (123 of 266 [46.2%] vs 47 of 251 [18.7%]; difference, 27.5%; 95% CI, 19.4%-35.6%), and more likely to report being able to “make ends meet” or afford necessities (230 of 266 [86.5%] vs 79 of 238 [33.2%]; difference 53.3%; 95% CI, 45.6%-60.9%). There was no statistically significant difference between groups in the following 5 of 14 patient oriented outcomes: receiving medicines in good condition, receiving new medicines quickly, reported medicine adverse effects, having unanswered questions about medicines, information changing the way medicines were taken, and the information from the pharmacist and prescriber matching.

in the following 5 of 14 patient oriented outcomes: receiving medicines in good condition, receiving new medicines quickly, reported medicine adverse effects, having unanswered questions about medicines, information changing the way medicines were taken, and the information from the pharmacist and prescriber matching. Safety There was no substantial difference between participants in the free distribution group and those in the usual access group in serious adverse events (33 of 395 [8.4%] vs 35 of 391 [9.0%]; P = .80). Hospitalizations (26 of 395 [6.6%] vs 25 of 391 [6.4%]; P > .99), deaths (8 of 395 [2.0%] vs 8 of 391 [2.0%]; P > .99), and other serious adverse events (7 of 395 [1.8%] vs 4 of 391 [1.0%]; P = .55) were also similar between the free distribution group and the usual access group. There were 37 medication incidents involving 32 of 395 participants (8.1%) receiving free distribution.

, deaths (8 of 395 [2.0%] vs 8 of 391 [2.0%]; P > .99), and other serious adverse events (7 of 395 [1.8%] vs 4 of 391 [1.0%]; P = .55) were also similar between the free distribution group and the usual access group. There were 37 medication incidents involving 32 of 395 participants (8.1%) receiving free distribution. Discussion In our multicenter randomized trial, distributing a comprehensive set of essential medicines at no charge improved adherence. Free distribution also lowered systolic blood pressure and there was a suggestion of better diabetes control, although results did not reach statistical significance. There was no effect on low-density lipoprotein cholesterol levels. There was no increase in potentially inappropriate prescribing and there was no substantial difference in serious adverse events. Participants receiving free medicine distribution were more likely to report being able to make ends meet; the hypothesis that medicine access allows people to afford other necessities can be tested in future studies.

tentially inappropriate prescribing and there was no substantial difference in serious adverse events. Participants receiving free medicine distribution were more likely to report being able to make ends meet; the hypothesis that medicine access allows people to afford other necessities can be tested in future studies. To our knowledge, no previous trial has assessed the effect of providing a wide range of medicines for free, including treatments for chronic noncommunicable disease (eg, diabetes and rheumatoid arthritis), chronic infectious disease (eg, HIV and AIDS), and acute conditions (eg, pneumonia) and symptoms (eg, pain). A trial of free access to secondary prevention medicines after a myocardial infarction found greater adherence in a higher risk population (38.9% in usual access patients, compared with 26.6% in our study) and modest improvements with free access (absolute increase, 5.4%). Offsetting copayments for clopidogrel bisulfate or ticagrelor after an acute myocardial infarction increased adherence slightly (absolute increase, 3.3%) but did not affect major adverse cardiovascular events. Although improving medicine adherence is difficult—only 5 of 17 complex interventions involving frequent patient contact modestly improved adherence—even relatively small increases in adherence to effective treatments seem to improve health outcomes such as cardiovascular events or reduce mortality such as HIV- and AIDS-related mortality in higher risk individuals.

ult—only 5 of 17 complex interventions involving frequent patient contact modestly improved adherence—even relatively small increases in adherence to effective treatments seem to improve health outcomes such as cardiovascular events or reduce mortality such as HIV- and AIDS-related mortality in higher risk individuals. Health improvements with free distribution are supported by the changes in some surrogate health outcomes. The mean systolic blood pressure reduction of 7.2 mm Hg observed here is likely large enough to reduce mortality at some baseline levels. The observed small reduction in hemoglobin A1c of 0.38%, if true, may be enough to reduce microvascular complications, based on trials of intensive control. At the same time, many patients did not see improvements in surrogate outcomes despite free distribution of medicines, emphasizing that cost is only one of several contributors to nonadherence and that medicines are just one part of care. The setting and design of the trial allows inferences about the causal effects of essential medicine access to be drawn because all participants had access to publicly funded health care services. The study population included people with a range of income levels and sources and different ethnicities who lived in urban and rural settings.

gn of the trial allows inferences about the causal effects of essential medicine access to be drawn because all participants had access to publicly funded health care services. The study population included people with a range of income levels and sources and different ethnicities who lived in urban and rural settings. Limitations Findings from one high-income country should be applied with caution in jurisdictions with different health care services and disease burdens. All the participants self-reported cost-related nonadherence, and the effect of free medicine distribution will presumably be smaller where adherence is better. At the same time, all participants had the option to apply for public insurance (which would have capped drug payments at approximately 4% of annual income) and others had public or private insurance, so the free distribution may be beneficial even in populations with some insurance. We did not measure adherence at baseline and so cannot determine if free distribution depends on the baseline level of adherence. Unblinded allocation to free medicine distribution could have motivated participants to exaggerate their adherence or resulted in different care from clinicians. The respective contributions of the different aspects of this multifaceted intervention, including free access and mailing of prescriptions, were not assessed in this 2-group trial, although mailing medicines is associated with improved adherence. There is no ideal method for determining medicine adherence because patient reports can overestimate adherence and objective measures such as medical record reviews can underestimate it. We could not confirm adherence with electronic pill bottle cap devices in a subset of participants as planned because most participants did not return the devices. There were missing data for the surrogate health outcomes as we did not provide clinicians with any instructions about checking blood pressure or laboratory indexes because doing so might have affected the outcomes. The cost-effectiveness of the intervention will be assessed after the 24-month trial is complete; a previous study has estimated that purchasing these medicines in bulk for the entire Canadian population would save approximately $4 billion per year, with an incremental government cost of approximately $1 billion per year. This study was not designed or powered to assess effects on mortality or morbidity, although the provided medicines include highly effective ones.

medicines in bulk for the entire Canadian population would save approximately $4 billion per year, with an incremental government cost of approximately $1 billion per year. This study was not designed or powered to assess effects on mortality or morbidity, although the provided medicines include highly effective ones. Conclusions In this randomized clinical trial, distributing essential medicines at no charge improved adherence to treatment with effective medicines. These results could help inform policy changes such as publicly funding a list of essential medicines as recommended by the Canadian Advisory Council on the Implementation of National Pharmacare. Supplement 1. Trial Protocol Click here for additional data file. Supplement 2. eAppendix 1. Essential Medicines Provided to Patients With Free Distribution eAppendix 2. Criteria for Potentially Inappropriate Prescriptions eTable 1. Alternative Definitions of Primary Outcome Results Table eTable 2. Patient Oriented Outcome Results Table Click here for additional data file. Supplement 3. Data Sharing Statement Click here for additional data file.