Patient adherence: Clinical pharmacology's embarrassing relative

Much of a clinical pharmacologist's time is spent studying the clinical response to drugs in humans, and what factors contribute to that response. In particular, those of us involved in the building of models describing the exposure-response relationship for drugs are frequently pre-occupied with the search for that unique combination of body weight, age, sex, renal/hepatic function, race/ethnicity, and genetics which will explain either the variation in the pharmacodynamic response of the drug, or the variability in the pharmacokinetics underlying that response. This covariate model building can take considerable amounts of time, frequently results in passionate disagreements between modelers over things like objective functions and functional forms for covariates, and, in the end, seldom results in explaining more than a small fraction of the variability in the response under study. Many times it seems hardly worth the effort. Ironically, there exists a covariate which may be quantifiably measured and interpreted, is capable of decreasing the residual variability in population pharmacokinetic studies by upwards of 50%,1 is positively correlated with efficacy,2 but is almost never measured in clinical trials. That covariate, of course, is patient adherence. By simply quantifying when and if a given patient takes his medicine, we remove an enormous amount of noise from both the pharmacokinetic and pharmacodynamic signals of the drug under investigation. Yet, despite the obvious connection between a patient taking a medicine as prescribed and its effect, adherence is not discussed much within the drug development community, much like that embarrassing relative on your spouse's side of the family. We collect body weight, race, age, genomics, and renal function data routinely in all trials, but seldom, if ever, collect data on whether our patients actually take the medicine we work so hard to develop. In speaking to numerous clinicians about adherence in clinical trials, the prevailing feeling seems to be that, although adherence is a serious issue with approved drugs, on the research side of things we do not have to worry so much. “I'm confident that our patients are adherent,” said one project leader to me recently, with no evidence to support the statement. There is a belief that, because patients are willing volunteers in our studies, and we study serious diseases, they will obviously take their medicine as prescribed. Unfortunately, one does not have to look very far for evidence that patient adherence in clinical studies is abysmal. Consider Figure 1, which depicts the results of an 8-week patient study of DOV 21947 on weight loss in 28 obese patients.3 In this study, adherence was measured with a very crude approach; at each of 3 clinic visits, a plasma sample was drawn and analyzed for the presence of DOV 21947. Patients were designated as adherent for that visit if any drug was detected in plasma. Those patients who were adherent at all 3 clinic visits met the criterion as adherent for the trial. Using this approach, 70% of the patients were deemed adherent. Using this information in the analysis of the study data clearly shows that those patients who were adherent lost weight, whereas those who were not actually gained weight (Figure 1, left panel). Although a combined analysis (adherent and nonadherent patients together) was not performed, it does not take much effort to see that such an intent-to-treat analysis would have yielded an overall conclusion of no effect. Notably, adherence via “pill” counts was estimated at 92%, far higher than the actual measured value.3 This example is important because it is the type of proof-of-concept study we in industry frequently perform in early development. Such studies provide a reason to believe that further investment in the study compound is worthwhile. Without measuring adherence, a potentially important drug may have been mistakenly dropped from development. One wonders how many potentially effective compounds the pharmaceutical industry has dropped from further clinical development simply because patients did not take the medicine as directed, and this was not taken into account in the analysis? “In summary, a substantial proportion of patients with CML treated with imatinib for more than 2 years fail to take a drug that can unequivocally prolong their life, and may, in some instances, cure their leukemia…”4 Those who cling to the notion that patients in clinical trials for serious diseases such as cancer will be adherent to their experimental therapy should note this example well. The final example illustrates the point that often, nonadherence is deliberate. The Lung Health Study was a 5-year clinical trial to evaluate the effect of concentrated smoking cessation counseling and bronchodilator use on the progression of COPD in 101 smokers.5 The metered-dose inhalers used in the trial were equipped to record the date and time of actuation without the patients' knowledge. During the first year of the trial, 30% of the patients actuated their MDIs repeatedly (over 100 times) in an effort to make themselves look adherent to therapy. This usually took place just before a clinic visit, and strongly indicates that these patients were deliberately trying to conceal their nonadherence from the investigators. Strangely, self-reported inhaler usage did not reflect this dose dumping. The conclusion of the researchers is that deliberate deception may occur frequently in clinical trials, and is not predictable unless data on actual use patterns are collected via electronic means.5 A couple of truths may be gleaned from this handful of examples. First, patient diaries and “pill” counts are worthless as measurements of adherence. We in the industry should stop performing these activities as they are of little value. Secondly, some quantitative measurement of adherence should be part of all Phase II and Phase III trials, as we may be throwing away important therapies for the wrong reasons. There are many ways to measure adherence today, including package entry detection, ingestible sensors which detect ingestion of the dosage form, and photographic documentation. All of these methods have advantages and disadvantages, which are detailed in a recent review by Vrijens and Urquhart.6 There is also a cost associated with using these methods, but compared with the expense of a failed Phase III trial, the cost is likely to be trivial. We also need to prospectively analyze the data resulting from these trials using adherence as a covariate. Accounting for smaller sources of variability (age, weight, genomics) without accounting for whether or not the patient is actually taking the study drug is not a fruitful endeavor, as no covariate could possibly have as big of an influence on efficacy and safety as not taking the drug.7 Ignoring adherence in obedience to “intent-to-treat” dogma, particularly in Phase II trials, is not likely to be beneficial to anyone. Finally, we must come to grips with the fact that lack of adherence may often be deliberate, and not just because of forgetfulness. Thus, it will be very important to determine why patients are nonadherent with their therapy early on in the development of our compounds. This knowledge could then be used postmarketing to improve adherence in patients taking our medications, which would benefit both patients and the industry. As clinical pharmacologists, we understand the link between dosing regimen, plasma concentrations, and clinical response/toxicity better than any other discipline involved in drug development. I suggest that those of us in the field take the initiative in encouraging the measurement of adherence in all Phase II and III clinical trials. I call on societies involved in quantitative pharmacology and drug development (ACCP, ASCPT, AAPS, IQ Consortium, and ISOP) to develop position statements on adherence and encourage its use in clinical trials. Simulations performed as part of clinical trial design should include the effect of adherence on study performance. Reviewers for journals which publish clinical pharmacology/pharmacometrics studies should specifically ask for adherence data as part of their peer review of these studies, and those which use adherence as part of their analysis should receive special consideration. Finally, and most importantly, regulatory agencies should strongly encourage companies to not only quantify adherence in all clinical trials, but also encourage the use of this important covariate in the analysis of those trials. It is long past time to bring patient adherence back into the clinical pharmacology family. The author is an employee and stockholder of GlaxoSmithKline. The opinions reflected in this paper are the author's and not necessarily those of GlaxoSmithKline.