Association of Medication Adherence with Health-Status Outcomes: Insights from the ISCHEMIA Trial

Abstract

Background:

ISCHEMIA randomized participants with chronic coronary disease (CCD) to guideline-directed medical therapy with or without angiography and revascularization. We examined the association of non-adherence with health status outcomes.

Objectives:

To compare 12-month health status outcomes of adherent and non-adherent participants with CCD with an a priori hypothesis that non-adherent patients would have better health status if randomized to invasive management.

Methods:

Self-reported medication-taking behavior was assessed at randomization with a modified 4-item Morisky-Green-Levine Adherence Scale and participants were classified as adherent or non-adherent. Twelve-month health status was assessed with the 7-item Seattle Angina Questionnaire (SAQ-7) Summary score (SS), which ranges from 0 to 100 (higher=better). The association of adherence with outcomes was evaluated using Bayesian proportional odds models, including an interaction by study arm (conservative vs. invasive).

Results:

Among 4480 randomized participants, 1245 (27.8%) were non-adherent at baseline. Non-adherent participants had worse baseline SAQ-7 SS in both conservative (72.9±19.3 vs. 75.6±18.4) and invasive (71.0±19.8 vs. 74.2±18.7) arms. In adjusted analyses, adherence was associated with higher 12-month SAQ-7 SS in both treatment groups (mean difference in SAQ-7 SS with conservative treatment= 1.6 (95% credible interval [CrI] 0.3, 2.9) vs. with invasive management= 1.9 (95% CrI 0.8, 3.1)), with no interaction by treatment.

Conclusions:

More than 1 in 4 participants reported medication non-adherence, which was associated with worse health status in both conservative and invasive treatment strategies at baseline and 12 months. Strategies to improve medication adherence are needed to improve health status outcomes in CCD, regardless of treatment strategy.

CONDENSED ABSTRACT

Non-adherence could potentially undermine the efficacy of a conservative medical strategy for chronic coronary disease. We evaluated 4480 participants from the ISCHEMIA trial, of whom 27.8% self-reported non-adherent medication taking behavior. Non-adherent participants had worse baseline SAQ-7 Summary Scores in both conservative (72.9±19.3 vs. 75.6±18.4) and invasive (71.0±19.8 vs. 74.2±18.7) arms, with adjusted analyses showing adherence to be associated with higher 12-month scores in both treatment groups (conservative: 1.6 (95% credible interval [CrI] 0.3, 2.9) invasive: 1.9 (95% CrI 0.8, 3.1)), with no interaction by treatment. These data support improving medication adherence regardless of treatment strategy.

INTRODUCTION

The principal goals of treating patients with chronic coronary disease (CCD) are to minimize clinical events (death and myocardial infarction) and to improve patients’ health status (symptoms, function and quality of life).^(1,2) To test the incremental advantage of an invasive strategy (coronary angiography with revascularization, when possible) over and above guideline-directed medical therapy (GDMT) alone, the ISCHEMIA trial randomized 5179 patients with stable CCD and at least moderate ischemia to either an invasive or conservative strategy.⁽³⁾ With a median of 3.2 years of follow-up, the primary and major secondary clinical event rates did not differ statistically according to treatment, but health status was better with an invasive strategy in those who had angina at baseline.^(4,5)

Despite strong evidence that patients who take their medications as prescribed experience significantly better outcomes than those who do not,^(6,7) medication non-adherence remains a significant clinical problem.^(8,9) In a meta-analysis of 376,162 patients taking medication to prevent cardiovascular disease, only 57% took their medications as prescribed.⁽¹⁰⁾ While medication adherence is important in all patients with CCD, we hypothesized that the health status benefits of a conservative strategy emphasizing GDMT would be greater in patients taking medicines as prescribed, and that the health status differences between adherent and non-adherent patients would be smaller because invasively-managed patients would also have the benefits of revascularization. To test this hypothesis, we examined whether medication non-adherence in patients with CCD was associated with worse health-status outcomes in conservatively-treated participants, and whether this effect was attenuated in those randomized to an invasive treatment strategy.

METHODS

Trial Population

Details of the ISCHEMIA trial and proper ethical oversight have been previously described.^(4,5) The protocol was approved by the institutional review board at New York University School of Medicine (the Clinical Coordinating Center) and by institutional review boards and ethics committees at each participating site. Briefly, participants were enrolled after clinically-indicated stress testing revealed moderate or severe ischemia. Blinded coronary computed tomography angiography was performed in participants without severe renal insufficiency to exclude left main coronary artery disease (CAD) and confirm obstructive CAD before randomization. Key exclusion criteria were estimated glomerular filtration rate (eGFR) <30 mL/min/1.73m², an acute coronary syndrome within 2 months of randomization, unprotected left main CAD (stenosis ≥50%), left ventricular ejection fraction (LVEF) <35%, New York Heart Association Class III-IV heart failure, and unacceptable angina despite maximally-tolerated GDMT or patient dissatisfaction with medical management of angina. Eligible participants were randomized to an invasive strategy including GDMT with coronary angiography and, when indicated, revascularization (invasive therapy), or to GDMT alone (conservative therapy) with coronary angiography and revascularization reserved for failure of medical therapy. GDMT consisted of anti-anginal medications and secondary prevention with pharmacologic and lifestyle intervention applied equally to both groups using treat-to-target algorithms.⁽¹¹⁾ Participants were followed at 1.5, 3, 6, and 12 months after randomization and every 6 months thereafter.

As previously described,⁽⁵⁾ of the 5179 participants originally enrolled in ISCHEMIA, 4617 participants completed quality of life assessments at baseline and at least once during follow-up (Figure 1). We further excluded 137 participants with incomplete modified Morisky-Green-Levine Adherence Scale assessments,⁽¹²⁾ yielding a final study population of 4480 randomized participants (2255 conservative, 2225 invasive).

Figure 1.

Consort Diagram

Health-Status Outcomes

Health status outcomes were assessed with the 7-item Seattle Angina Questionnaire (SAQ-7) at randomization, 1.5, 3, 6, and 12 months.^(13–15) The SAQ-7 captures the frequency of angina (Angina Frequency score), and the disease-specific effect of angina on participants’ physical function (SAQ Physical Limitation score) and quality of life (SAQ Quality of Life score), over the preceding 4 weeks. These scores are averaged to obtain a SAQ-7 Summary score, which is an overall measure of participants’ CCD-specific health status. Scores range from 0 to 100, with higher scores indicating less frequent angina, better function, and higher quality of life. Mean differences of 5 points or more in SAQ Summary scores suggest at least a moderate, clinically meaningful impact.⁽¹⁶⁾ Small (<5 points) but clinically significant differences would be determined if there were a significant between-group difference in SAQ Summary scores when comparing those with and without baseline angina.

Medication Adherence

Self-reported medication-taking behavior was assessed at baseline with the modified 4-item Morisky-Green-Levine Adherence Scale.⁽¹²⁾ This is a modification of the original yes/no response scale to allow responses of strongly disagree, disagree, agree, strongly agree, don’t know and refused for the following questions: “I sometimes forget to take my medicines,” “I am sometimes careless about taking my medicines,” “When I feel better, I sometimes stop taking my medicines,” and “If I feel worse when I take my medicine, sometimes I stop taking it.” Responses of disagree or strongly disagree to all four items were classified as being adherent, and all other patterns of response were classified as non-adherent (Table S1).

Statistical Analysis

For descriptive purposes, participants’ baseline characteristics and unadjusted SAQ-7 subscale scores were summarized by adherence status. Between-group comparisons were made using standardized mean differences. The standardized mean difference is the mean difference between groups divided by a pooled estimate of the standard deviation, expressed as a percentage. Standardized mean differences >10% are interpreted as denoting non-trivial imbalances between groups.^(17,18)

The primary outcome was 12-month SAQ-7 Summary score, and the SAQ-7 subscales were secondary outcomes. Proportional odds models, within a Bayesian framework, were used to estimate differences between adherent and non-adherent participants by treatment group.⁽⁵⁾ Analyses did not further stratify the invasive group by revascularization strategy as the allocation to either percutaneous coronary intervention or coronary artery bypass graft surgery was not randomized. Models included terms for adherence, treatment, adherence-by-treatment interaction (the term that defines whether or not there is greater benefit of one treatment vs. another based on adherence), and covariate adjustment for participant age, sex, smoking status, medical history of hypertension, diabetes, myocardial infarction (MI), coronary artery bypass grafting, heart failure, stroke, peripheral vascular disease, imaging modality, degree of ischemia, number of anti-anginal medications at baseline and baseline SAQ-7 score. Uninformative priors were used for all model parameters. To provide insights into health status over time, we used this approach to conduct secondary analyses to further estimate the differences in SAQ-7 Summary scores at 1.5, 3 and 6 months. Finally, given that the ISCHEMIA trial demonstrated better health status outcomes with invasive treatment in only those with baseline angina,⁽⁵⁾ we repeated the analyses in those with and without baseline angina (SAQ Angina Frequency score <100 and 100, respectively).

A total of 286 participants (6.4%) had missing 12-month SAQ-7 Summary scores (143 missed visits, 75 withdrawal/lost follow-up, 50 deaths, 16 incomplete assessments and 2 due to study termination). Covariate data were missing in 2% of participants. Missing outcome and covariate data were assumed to be missing at random, conditional on observed information, and were imputed using multivariate imputation by chained equations as implemented in the R package “mice.”⁽¹⁹⁾ The imputation model included all covariates and SAQ-7 Summary scores through 12 months, and incorporated a treatment-by-adherence interaction term in concordance with the study hypothesis. A total of 64 randomly imputed data sets were generated, models were fit separately on each, and the resulting posterior samples were pooled across all imputations for final inferences. Results are presented as adjusted posterior means with 95% credible intervals (CrI) to demonstrate the effect of adherence status on expected SAQ-7 scores for a “typical” patient (with medians/modes for covariates). Between-group differences with 95% CrI that include 0 indicate that no adherence-by-treatment interaction is a plausible scenario.

In a secondary analysis, we explored a dose-response effect of self-reported adherence on health status outcomes, treating adherence as a continuous variable. We derived a scoring system for the 4-item modified Morisky-Green-Levine scale with a total score for each participant, which ranged from 4 (“Strongly Agree” for all items) to 16 (“Strongly Disagree” for all items). Higher scores denote better self-reported adherence. We assessed the internal validity of the total score by Cronbach’s alpha of the four items (0.82, suggesting good internal consistency) and by exploratory factor analysis, which yielded one factor having an eigenvalue >1.0 (65% of covariation explained) and comparable factor weightings of 0.77 to 0.84 for the four items. For inferential analysis, we refit the primary Bayesian proportional odds model of 12-month SAQ-7 Summary score on adherence-by-treatment, adjusting for patient covariates, but replacing the original binary adherence variable with the continuous adherence score variable. Using restricted cubic splines, we then plotted model-estimated mean SAQ scores by adherence score and treatment group, and the effect of invasive vs. conservative treatment as a function of the adherence score. All analyses were performed with SAS 9.4 (SAS Institute, Cary, NC), R version 3.6.3 and RStan version 2.19.3.⁽²⁰⁾

RESULTS

Participants

Among 4480 randomized participants included in these analyses, the majority were classified as adherent, with 27.8% classified as non-adherent at baseline. Of the 1245 who were non-adherent, 50.0% were randomized to the conservative and invasive arms, respectively. Compared with adherent participants, non-adherent patients were slightly younger (63.6 vs.64.7), more likely to smoke (16.4% vs. 11.9%), and more likely to have lower baseline scores on the SAQ-7 Summary (71.9±19.6 vs. 74.9±18.5), Physical Limitation (76.9±24.4 vs. 80.3±23.1), and Quality of Life (58.9±27.2 vs. 62.9±25.9 vs.) scales (Table 1), indicating worse baseline health status. Other baseline characteristics were similar between adherent and non-adherent groups.

Table 1.

Patient Baseline Characteristics

	Adherent	Non-Adherent	Standard Difference
	N = 3235	N = 1245	(%)
Treatment group			1.0
Conservative	1633 (50.5%)	622 (50.0%)
Invasive	1602 (49.5%)	623 (50.0%)
Age	64.7 ± 9.5	63.6 ± 9.4	10.8
Sex			4.9
Male	2500 (77.3%)	936 (75.2%)
Female	735 (22.7%)	309 (24.8%)
Region			16.3
Asia	724 (22.4%)	326 (26.2%)
Europe	1219 (37.7%)	392 (31.5%)
Latin America	312 (9.6%)	152 (12.2%)
North America	899 (27.8%)	354 (28.4%)
Other	81 (2.5%)	21 (1.7%)
Diabetes	1275 (39.4%)	502 (40.3%)	1.9
Smoking Status			14.8
Never Smoked	1358 (42.0%)	457 (36.7%)
Former Smoker	1491 (46.1%)	584 (46.9%)
Current Smoker	384 (11.9%)	204 (16.4%)
Missing	2
eGFR	83.1 ± 22.6	82.5 ± 22.8	2.7
Hypertension	2440 (75.7%)	971 (78.3%)	6.3
Missing	10	5
Heart Failure	118 (3.6%)	81 (6.5%)	13.0
Peripheral Vascular Disease	134 (4.2%)	57 (4.6%)	2.1
Missing	1
Prior Stroke	99 (3.1%)	43 (3.5%)	2.2
Missing	1
Prior MI	647 (20.1%)	282 (22.7%)	6.5
Missing	9	3
Prior CABG	148 (4.6%)	46 (3.7%)	4.4
Imaging Modality			15.8
Nuclear	1694 (52.4%)	725 (58.2%)
Echo	743 (23.0%)	281 (22.6%)
CMR	148 (4.6%)	59 (4.7%)
ETT	650 (20.1%)	180 (14.5%)
Degree of Ischemia			8.9
None	176 (5.5%)	60 (4.9%)
Mild	219 (6.9%)	102 (8.3%)
Moderate	1122 (35.1%)	467 (37.8%)
Severe	1680 (52.5%)	605 (49.0%)
Missing	38	11
Number of Diseased Vessels			9.8
(Vessels with stenosis ≥70%)
0	199 (13.1%)	88 (13.9%)
1	620 (40.9%)	260 (41.2%)
2	434 (28.6%)	157 (24.9%)
3	262 (17.3%)	126 (20.0%)
Missing	1720	614
Number of Anti-Anginal Medications			6.2
0	308 (9.5%)	133 (10.7%)
1	1404 (43.3%)	533 (42.8%)
2	1099 (34.0%)	436 (35.0%)
SAQ-7 Summary Score	74.9 ± 18.5	71.9 ± 19.6	15.8
Missing	1	0
SAQ-7 Physical Limitation Score	80.3 ± 23.1	76.9 ± 24.4	14.3
Missing	253	94
SAQ-7 Angina Frequency Score	82.0 ± 19.5	80.0 ± 20.0	9.9
Missing	3	1
SAQ-7 Quality of Life Score	62.9 ± 25.9	58.9 ± 27.2	15.2
Missing	8	1

Abbreviations: CABG = Coronary artery bypass grafting. CMR = cardiac magnetic resonance. eGFR = Estimated glomerular filtration rate. ETT = exercise tolerance test. MI = Myocardial infarction. SAQ-7= 7-item Seattle Angina Questionnaire.

Unadjusted Health-Status Outcomes Over Time

Figure 2 shows the health status outcomes for conservative and invasive participants over 12 months. Additional figures of the SAQ subscales (Angina Free, Physical Limitation, and Quality of Life) are provided in the Supplementary Appendix (Figure S1). Among conservatively-treated participants, those who reported being adherent had a higher SAQ-7 Summary score at baseline than those that were non-adherent cohort (75.6±18.4 vs 72.9±19.3), which was also observed at 12 months (85.4±15.0 vs 81.2±18.3). A similar pattern was observed among participants randomized to the invasive arm, both at baseline (74.2±18.7 vs. 71.0±19.8) and at 12 months (88.2±13.9 vs. 84.3±17.4; Table 2).

Figure 2.

Unadjusted Mean SAQ-7 Summary Scores Over Time for Conservative vs Invasive Treatment Groups

Table 2:

SAQ-7 Summary Scores for Treatment Strategy By Medication Adherence

	Adherent		Non-Adherent
Month	N	Mean ± SD (95% CI)	N	Mean ± SD (95% CI)
Conservative Arm
0	1632	75.6 ± 18.4 (74.7, 76.5)	622	72.9 ± 19.3 (71.4, 74.4)
1.5	1566	80.6 ± 17.6 (79.8, 81.5)	595	76.7 ± 19.2 (75.1, 78.1)
3	1562	83.1 ± 16.3 (82.3, 83.9)	595	78.6 ± 17.8 (77.2, 80.1)
6	1563	84.2 ± 16.4 (83.3, 85.0)	594	80.0 ± 18.1 (78.5, 81.4)
12	1523	85.4 ± 15.0 (84.7, 86.2)	591	81.2 ± 18.3 (79.7, 82.7)
Invasive Arm
0	1602	74.2 ± 18.7 (73.3, 75.1)	623	71.0 ± 19.8 (69.4, 72.5)
1.5	1481	81.5 ± 17.4 (80.6, 82.4)	575	78.7 ± 19.5 (77.1, 80.3)
3	1507	85.6 ± 14.9 (84.9, 86.4)	582	82.2 ± 17.8 (80.8, 83.7)
6	1522	87.6 ± 13.9 (86.9, 88.3)	594	83.0 ± 17.4 (81.6, 84.4)
12	1500	88.2 ± 13.9 (87.5, 88.9)	580	84.3 ± 17.4 (82.9, 85.7)

Adjusted 12-month Health-Status Outcomes and Association with Medication Adherence

After adjusting for patient characteristics, there remained an independent association between baseline medication adherence and 12-month health-status outcomes. The differences between adherent and non-adherent participants were similar for the conservative (mean difference in 12-month SAQ-7 Summary score = 1.6 [95% CrI 0.4 to 2.8]) and invasive groups (1.9 [95% CrI of 0.8 to 3.0]; Figure 3). Importantly, there was no difference observed in the association of the effect of medication adherence between conservative and invasive treatment (difference between treatment groups in SAQ-7 Summary score = 0.3 [95% CrI of −1.3 to 1.9]; Table 3). Consistent results were found for the SAQ-7 domains of Angina Frequency, Physical Limitation, and Quality of Life (See Supplementary Appendix, Figures S2–4). Secondary analyses on the evaluation of the dose-response effect of self-reported adherence on health status outcomes suggested there is a graded effect of adherence scores, with higher scores of adherence being associated with higher SAQ-7 Summary scores. These analyses similarly suggested that there does not appear to be an appreciable differential effect of adherence by treatment group (Figures S5–6).

Figure 3. The Posterior Distribution of the Adjusted Effect of Medication Adherence on Mean SAQ-7 Summary Scores at 12 Months.

These figures represent Bayesian posterior distributions of the estimated and adjusted association of medication adherence with 12-Month SAQ-Summary scores (SS) within each treatment strategy. The x-axis represents possible values of the effect of medication adherence on SAQ score, and the y-axis represents the density or relative likelihood of those values. Higher y-values correspond to more likely effect sizes. Numerical results include the mean of these distributions (i.e., the mean difference between adherent and non-adherent participants’ 12-month SAQ SS) and 95% of the highest posterior density credible intervals (CrI), (i.e., 95% of the most likely values of the effect).

Table 3: Adjusted Effects of Adherence on Mean SAQ-7 Summary Scores at 12 Months.

This table demonstrates the posterior-adjusted mean difference in 12-month SAQ-7 Summary scores (SS) between adherent and non-adherent participants. Mean differences (i.e., the mean difference between adherent and non-adherent participants’ 12-month SAQ SS) and 95% credible intervals (i.e., 95% of the most likely values of the effect), for both the conservative and invasive treatment groups, are reported for each of the SAQ-7 subscales, as well as the difference in means between groups (invasive minus conservative). Between-group differences with 95% credible intervals (CrI) that include 0 indicate that no adherence-by-treatment interaction is a plausible scenario.

SAQ-7 Subscale Scores	Association of Medication Adherence with 12-Month Score*
	Conservative Arm (95% CrI)	Invasive Arm (95% CrI)	Difference (95% CrI)
Summary Score	1.6 (0.4, 2.8)	1.9 (0.8, 2.9)	0.3 (−1.3, 1.9)
Angina Frequency Score	1.1 (0.1, 2.0)	1.2 (0.5, 1.9)	0.1 (−1.1, 1.3)
Quality of Life Score	2.4 (0.4, 4.4)	2.9 (1.0, 4.8)	0.6 (−2.1, 3.3)
Physical Limitation Score	1.7 (0.4, 3.1)	2.0 (0.8, 3.2)	0.3 (−1.5, 2.0)

Abbreviations: CrI = Credible interval. SAQ-7 = 7-item Seattle Angina Questionnaire

^*For a typical patient with medians/modes for age, sex, smoking status, hypertension, diabetes, prior MI, prior coronary artery bypass grafting, heart failure, prior stroke, peripheral vascular disease, imaging modality, degree of ischemia, number of anti-anginal medications and baseline score.

To examine the trajectory of health status improvement from randomization to 12 months, we performed within-group comparisons that demonstrated higher mean SAQ-7 Summary scores in adherent patients throughout the first year of treatment. We observed qualitatively larger differences during earlier follow-up within the conservative arm than in the invasive arm, but the differences between treatment arms were not, in general, statistically significant (Table S3). Furthermore, when restricting the analyses to just those with baseline angina, no significant differences between the primary analyses were observed (Table S4).

DISCUSSION

Medication non-adherence is common and associated with worse clinical outcomes.^{(6,7,10,21–29)} Given that the conservative management of CCD may be particularly dependent upon adherence to evidence-based pharmacologic and lifestyle interventions, we tested whether self-reported non-adherence to medications would be associated with even greater health status benefits with an invasive treatment strategy. We found that among participants with quality-of-life assessments in the ISCHEMIA trial, non-adherent participants experienced worse 1-year health-status outcomes than those who were adherent with their medications. However, contrary to our hypothesis that medication non-adherence would be associated with worse health-status outcomes among participants randomized to conservative management, as compared with those randomized to invasive management, we found a similar detrimental effect of medication non-adherence in both treatment arms (Central Illustration). These data suggest that non-adherence is not a reason to preferentially consider an invasive strategy and that strategies to support medication adherence should be pursued in all patients with CCD, regardless of treatment strategy.

Central Illustration.

Association of Medication Adherence With Health Status Outcomes

Abbreviations: CrI = credible interval

These findings support prior research on the association of medication adherence with outcomes and extend this work to provide a deeper understanding of the health status outcomes when managing patients with CCD. The rate of self-reported adherence (72.2%) in ISCHEMIA is similar to the proportion of medication adherence observed in the CRUSADE (Can Rapid risk stratification of Unstable angina patients Suppress ADverse outcomes with Early implementation of the American College of Cardiology/American Heart Association Guidelines) and ACTION (Acute Coronary Treatment and Intervention Outcomes Network) registries (72%).^(30,31) Our results are consistent with previous research demonstrating that patients who do not adhere to medications have worse outcomes than those who are adherent;^{(6,7,10,21–29)} an effect seen in multiple chronic diseases,⁽²⁷⁾ as well as cardiovascular disease.^{(6,7,10,21–26,28,32)} However, this is the first report that we are aware of to evaluate these associations for disease-specific health status outcomes.

Importantly, in both this report and prior studies, it was not possible to determine whether poorer adherence is a mediator or a marker of worse outcomes. Nonadherent patterns for example, that are related to prescription costs, have been shown to be associated with higher all-cause and disease-specific mortality.⁽³³⁾ Additional forms of cost-related nonadherence (e.g., insurance coverage), and other factors such as social determinants of health, have been shown to similarly influence adherence and subsequent outcomes.^(34,35) Moreover, it has been postulated that greater self-reported adherence reflects a “healthy adherer” effect, where differences in outcomes are not due to treatment but, rather, an overall pattern of healthier patient behavior.⁽³²⁾ Rasmussen and colleagues investigated the relationship between adherence to statins, β-blockers, and calcium-channel blockers and mortality one year after an acute MI.⁽²⁸⁾ They found that statin and β-blocker users categorized as high adherers had the lowest mortality rate as compared with low and intermediate adherers, but this effect was not observed with calcium-channel blockers, which have not shown a survival advantage post-MI,^(36–40) supporting a true clinical benefit rather than merely a healthy-adherer effect. Other studies, however, have also demonstrated improved health outcomes in relation to adherence even among those randomized to placebo.^(41–43) One study found similar rates of good adherence to both candesartan and placebo, as well as lower risk of death for both treatment groups in those who were adherent. These data support the argument that medication adherence is a marker, rather than a mediator, of improved outcomes. To better resolve this uncertainty, future clinical trials of strategies to improve adherence should be tested to define the benefits in participants’ disease-specific health status.

Extending studies on adherence to focus on health status outcomes is important because it has not been convincingly demonstrated that efforts to improve adherence improve clinical event outcomes. For example, despite testing multiple interventions to improve adherence – including patient education, technological-reminders, behavioral, motivational, or medication-cost interventions – no clear impact on reducing clinical events has been observed.⁽⁴⁴⁾ This may, however, reflect the difficulty in improving adherence. The latest Cochrane Systemic Review examining adherence interventions in 182 randomized controlled trials across many disease conditions found that medication adherence improved clinical outcomes in only five studies, resulting in no recommendations about best practices to improve adherence and outcomes.⁽⁴⁵⁾ Similar results were found in a systematic review evaluating interventions for the secondary prevention of cardiovascular disease, with text messaging, a polypill, and a community health care worker-based intervention being the only interventions that improved both adherence and clinical outcomes.⁽⁴⁶⁾ While adherence interventions vary widely, most focus on changing patient behavior. One possible strategy is for physicians to simply ask their patients about adherence. In a study examining the frequency of cardiologists’ discussion with their patients about adherence to cardiovascular medications, 61% of patients reported that this topic was rarely or never discussed.⁽⁴⁷⁾ Similarly, another report demonstrated that a provider’s under-recognition of their patients’ actual adherence status was the strongest predictor of lower rates of medication optimization.⁽⁴⁸⁾ Thus, although not routinely done, asking patients about adherence has multiple potential benefits as it fosters greater insight to the patient’s behavior to better guide discussions on therapies best-suited to improve care and outcomes.

Our findings should be interpreted within the context of several limitations. First, medication adherence was assessed only at baseline using self-report. Self-report has correlated with more objective measures, such as pill counts⁽⁴⁹⁾ and electronic monitoring of pill bottles,⁽⁵⁰⁾ and is much more feasible in clinical practice than these alternative methods. Second, the modified Morisky-Green-Levine Adherence Scale was expanded from its original version to include more possible response choices. However, these response choices were then collapsed to be concordant with the original scale for defining those classified as adherent and non-adherent. Moreover, although our primary analysis of non-adherence dichotomized patients into being adherent or not, a secondary analysis suggested worse health status in both treatment groups in those who were more non-adherent. It is also important to note that adherence patterns can change over time, but our analytic perspective was that of a clinician considering an invasive treatment strategy for a patient with CCD who would not know how adherence patterns might change over time and would only have access to a baseline assessment. Finally, the ISCHEMIA trial enrolled a carefully screened population of patients with CCD. Patients unwilling to take medications or who were very dissatisfied with medical therapy and those with unacceptable angina were excluded. It is possible that patients with higher levels of non-adherent behavior toward GDMT might have shown greater benefits from invasive management, which we had postulated but did not demonstrate.

CONCLUSIONS

More than 1 in 4 participants with CCD enrolled in the ISCHEMIA trial self-reported non-adherence to their medications. Non-adherence was associated with worse symptoms and decreased health status in both conservative and invasive arms compared with participants reporting medication adherence, without evidence that non-adherent patients received an even greater health status benefit from an invasive strategy than adherent patients. This suggests that non-adherence is not a justification to prefer an early invasive strategy in CCD patients with significant ischemia. Assessing and encouraging good adherence to GDMT may improve health-status outcomes regardless of whether initial treatment of patients with CCD is conservative or invasive.

CLINICAL PERSPECTIVES.

Competency in Medical Knowledge:

Non-adherence could potentially undermine the efficacy of a more conservative medical strategy, for which poor medication-taking behavior would have better health-status outcomes if randomized to invasive strategies.

Translational Outlook #1:

Although non-adherence is associated with worse outcomes, equivalent detrimental effects between conservative and invasive strategies suggest that identification of a high propensity for non-adherence is not a justification to prefer an early invasive strategy in CCD patients with significant ischemia.

Translational Outlook #2:

In patients with chronic coronary disease, screening for, and encouragement of, adherence to guideline-directed medical therapy may improve health-status outcomes, regardless of the selected treatment strategy.

ABBREVIATIONS:

CAD

Coronary artery disease

CCD

Chronic coronary disease

CrI

Credible intervals

eGFR

Estimated glomerular filtration rate

GDMT

Guideline-directed medical therapy

ISCHEMIA

International Study of Comparative Health Effectiveness with Medical and Invasive Approaches

LVEF

Left ventricular ejection fraction

MI

Myocardial infarction

SAQ

Seattle Angina Questionnaire

SS

Summary score