Drug adherence monitoring in clinical trials

Evidence-based medicine is the actual paradigm that drives medical practice [1]. This approach decreases the impact of medical intuition and unsystematic clinical experience and favors the evidence gathered from well designed clinical researches such as randomized controlled trials or meta-analyses of randomized controlled trials. In the field of hypertension, international guidelines, such as those of the European Society of Hypertension and European Society of Cardiology, periodically revise the evidence for the diagnosis and management of hypertension [2] and provide useful recommendations on the adequate use of antihypertensive drugs and therapeutic strategies having the best chances to control blood pressure (BP) in the majority of patients. In general, evidence regarding the efficacy and tolerability profile of drugs or therapeutic strategies is obtained from small and large randomized clinical trials designed by pharmaceutical companies or clinical research organizations to address the requirements and regulations of national regulatory agencies such as the US Food and Drug Administration or the European Medicines Agency. These trials usually have a well defined design, very strict criteria of inclusion and exclusion, enroll large numbers of patients, and are conducted with motivated physicians and patients. For these reasons, these studies undoubtedly provide the best level of evidence. Nonetheless, they are sometimes criticized because of their risk of overestimation of clinical benefits and underestimation of side-effects or drug toxicity leading to a potential misinformation of physicians, patients, and policy makers as recently reported in the field of psychiatry [3]. Overestimation of the clinical benefits of drugs may be because of several factors including the underrepresentation of some patients’ category such as sicker and older individuals, the publication bias (negative studies are less likely to be published), the early termination of trials [4], the outcome reporting bias whereby only the positive outcomes of the study are published and not the negative one [5], or the spin phenomenon when the apparent benefits of treatments are not supported by the published data [6]. In recent years, the risk of overestimation of clinical results obtained in large clinical trials has caught a lot of attention, but is it really the major problem of these studies? Should we not consider the underestimation of the drugs’ clinical effectiveness and safety issues as a greater matter of concern? Indeed, in all controlled clinical trials assessing the efficacy and tolerability of new therapeutic agents or treatment strategies, it is assumed that adherence to the prescribed regimen is nearly ideal. There is, however, good evidence that it is not the case. Thus, the analysis of a cohort of 16 907 patients with different medical conditions enrolled in 95 clinical studies has clearly demonstrated that the number of patients taking the prescribed study medication(s) decreased progressively during the study so that at day 100, about 20% of patients had discontinued treatment and 12% of those still engaged with the dosing regimen were omitting a fraction of the prescribed doses. Consequently, less than 70% of patients were actually following the protocol-specified treatment [7]. Therefore, partial adherence or nonadherence can be a major confounding factor in the interpretation of clinical trial results, with consequences that include an underestimation of the drug efficacy as well as of the incidence of adverse effects [8]. An important consequence of this misleading information may be an overestimated dosing requirement for marketed pharmaceuticals and a need to lower the drug doses later on, when the compound is on the market [9]. The article published in this issue of the Journal of Hypertension is an illustration of the relevance of gathering adequate information on drug adherence while evaluating the results of an intervention in hypertension [10]. In 2012, Bobrie et al.[11] published a randomized controlled study in which sequential nephron blockade (SNB) using different diuretics was compared with a sequential blockade of the renin–angiotensin system (RAS) in patients with resistant hypertension. All patients were hypertensive despite a 4-week treatment with irbesartan 300 mg/day, hydrochlorothiazide 12.5 mg/day, and amlodipine 5 mg/day. Then patients were randomly assigned to receive first spironolactone 25 mg/day in the diuretic group or ramipril 5 mg/day in the RAS group for 4 weeks. Thereafter, treatment was increased at weeks 4, 8, or 10 if home BP was at least 135 mmHg and/or at least 85 mmHg by sequentially administering furosemide 20 mg/day, furosemide 40 mg/day, and amiloride 5 mg/day in the SNB group or ramipril 10 mg/day, bisoprolol 5 mg/day, and bisoprolol 10 mg/day in the sequential RAS blockade group. At week 12, the mean between-group difference in daytime ambulatory BP was 10/4 mmHg in favor of the SNB. The BP goal was achieved in 58% of patients with the combination of diuretics and only 20% with the intensive RAS blockade. In the predefined post-hoc analysis of Beaussier et al.[10] published in this issue of the Journal of Hypertension, the influence of medication adherence on BP control and target organ damage was investigated. Owing to the fact that it is difficult to measure drug adherence reliably, the authors used three different approaches. First, they measured plasma irbesartan and hydrochlorothiazide concentrations by high-performance liquid chromatography; second, they assessed the adherence to ramipril with the determination of N-acetyl-serylaspartyl-lysyl-proline/creatinine ratio in spot urines, a very sensitive marker of angiotensin converting enzyme inhibition [12]. At last, they used the classical pill count method. With these three assessments of drug adherence, an adherence score was calculated for each study participant. In this 12-week study, 18.0% of patients did not follow the prescribed regimen and were considered as poorly adherent. In contrast to the initial analysis, 64% of patients of the SNB group reached the BP target and the difference between the SNB and the sequential RAS blockade was significant only in those patients who were adherent. In addition, the assessment of target organ damage evaluated by the changes in pulse wave velocity and left ventricular hypertrophy were greater in patients assigned to the SNB group but only in those with an acceptable adherence. At last, the incidence of hyperkalemia in the SNB group was significantly higher in adherent patients. These results perfectly illustrate the importance of having reliable information on drug adherence when analyzing the data of studies investigating the efficacy and tolerability of new drugs, devices, or treatment strategies as in the case of Bobrie et al.'s study [11] and its post-hoc analysis by Beaussier et al.[10]. Without knowing how patients followed the prescription, there is a definite risk of underestimating the efficacy but also the incidence of side-effects of the investigated treatment. In patients with resistant hypertension, the recent development of renal denervation and, to a lesser degree, baroreceptor stimulation has emphasized the need to improve the characterization of patients in order to be sure that these patients are indeed resistant before enrolling them in a clinical trial. The initial recommendations did not consider the need to assess drug adherence in the selection process [13]. However, with time and experience, physicians suddenly discovered that two aspects of the clinical management of patients with resistant hypertension were crucial: first, to ascertain that patients were hypertensive not only in the office but also on a 24-h BP recording, and second, to be sure that the prescribed drugs were correctly taken as reported in the early 2000 already [14]. Thus, several recent studies have found that total nonadherence or poor drug adherence (i.e. <50% drug intake) can occur in nearly half of patients with resistant hypertension recruited in renal denervation programs [14–17][14–17][14–17][14–17]. Today, the assessment of drug adherence in programs of renal denervation has made its way, but even in most recent recommendations on design considerations for clinical trials in hypertension, the interest of measuring drug adherence is unfortunately still not considered [18]. In fact, monitoring drug adherence in resistant hypertension is necessary not only before an intervention such as renal denervation but also after the intervention. Indeed, patients with resistant hypertension are asked to continue their medical treatment, which generally consists of three to five pills per day. Therefore, the true BP effect of an intervention such as renal denervation depends as much on the efficacy of the denervation procedure as on drug adherence after the intervention. If patients markedly reduce their adherence to drug therapy after the denervation, BP will probably remain unchanged or decrease only slightly. A contrario, if drug adherence improves after the intervention, the decrease in BP after denervation might be very pronounced when compared with the baseline. With a combination of both, the intervention may have no effect as reported recently in the SIMPLICITY HTN3 trial [19]. Thus, fluctuations in drug adherence after renal denervation may in part account for the high variability of the BP response to the procedure in addition to technical aspects and issues linked to the patients’ selection [20]. The results published by Beaussier et al.[10] would also support the idea that because of the major problems of drug adherence, the efficacy and tolerability of new therapeutic agents or devices investigated in large clinical trials are rather underestimated than overestimated. Indeed, extensive underdosing due to a variable adherence can affect the results of these trials and confound the selection of the optimal dosing regimen. Thus, one could question what clinical results should be made available for physicians and authorities. Is it really useful to publish drug-induced BP decreases in populations including a large proportion of poorly adherent patients? Although it might represent the clinical reality, would it be more accurate to get information on the decrease in BP that can be achieved when the treatment is taken as prescribed (Fig. 1)? In this respect, one should really start to think about including an effective monitoring of drug adherence in clinical trials that goes beyond the traditional pill count that is known to grossly overestimate drug adherence and gives no real information on the drug dosing history. In the actual context of drug and device developments, including drug adherence in early phase 1 as well as in phases 2 and 3 studies could contribute to improve the determination of the efficacy and safety profile of the tested drugs or concomitant interventions. Indeed, conducting adherence-informed clinical trials as recently suggested by Vrijens and Urquhart [21] gives the opportunity to perform much more sensitive and complete analyses of study data. With the availability of adherence data, it will be possible to improve our understanding of the gap between the drugs’ efficacy during their development phases and their true efficacy while on the market. This gap nicely illustrated by Blaschke et al.[7] could also be minimized (Fig. 1).FIGURE 1: Idealized schematic projection of changes in adherence and changes in treatment outcomes during successive phases of clinical drug development. The gap between use-effectiveness and method-effectiveness increases as adherence declines during progressively less-strict management of patients’ medication and progressively more varied severity of disease and comorbidities of treated patients. In a perfect world, use-effectiveness and method-effectiveness would be equal [7].Today, the important question is how to measure drug adherence reliably in clinical trials? Several simple and cheap techniques are available and have been used such as the pill count method, but we know that they are imprecise and poorly informative. Today, there is much enthusiasm for the detection and determination of drugs in plasma or urines as performed by Beaussier et al.[10]. This approach indeed proves the ingestion of the drug but it has some obvious limitations essentially due to the costs and the white-coat adherence frequently observed in chronic diseases [22] including hypertension [23]. This is particularly important in hypertension because BP measured in the morning depends markedly on drug intake during the days before the medical term. At last, measuring drug concentrations does not provide a dosing history and the availability of results is often delayed. Today, few reliable methods of measuring drug adherence are available in addition to electronic monitoring or direct observed treatment or plasma drug monitoring [23], and new approaches are being developed such as electronic detection of package entry, ingestible smart sensors, or electronic diaries [22]. These new technical developments might completely change our perspective on measuring drug adherence in clinical trials and perhaps enable to design adherence-informed clinical trials. ACKNOWLEDGEMENTS Conflicts of interest There are no conflicts of interest.