Background

Adverse drug reaction (ADR) rates in clinical trials often understate those in practice.[1] Postmarketing surveillance is employed to help overcome the limitations in ADR identification in clinical trials. These arise from many factors including: inadequate sample size, limited trial duration, restricted range of assessed outcomes, and relative exclusion of precisely those patients most likely to experience adverse effects, such as the elderly and those with polypharmacy and co-morbidities.[2]

Because of such factors, “…the challenge of early detection [of ADRs] has largely shifted to … postmarketing systems,”[3] which seek to capture ‘real-world’ ADR experiences and are a major mechanism undergirding drug withdrawals.[4,5] Postmarketing surveillance “provides vital information of clinical importance,”[6] it may lead to the identification of new ADRs, and may modify knowledge about known ADRs.

But detection is not always ‘early’.[7] Both physician-and patient-initiated postmarketing ADR reporting are inherently subjective, and physicians may not report a putative ADR because they are unfamiliar with it; because for other reasons they do not attribute the patient’s symptoms to a prescribed drug; or because of time pressures and other factors. Indeed, physician surveys suggest reporting is infrequent even when an ADR is recognised.[8] Underreporting compromises the effectiveness of post-marketing surveillance[9] and may contribute to lags in identification of problems.[7] Although patients may also report ADRs, the US FDA and industry are both “dependent on practicing clinicians to actively participate in national postmarketing drug surveillance.”[10] However, patients have been proposed as a reliable supplemental source of reports.[11]

We sought to determine whether a patient-targeted ADR surveillance approach may foster identification of additional patients with possible or probable ADRs that might not be captured through efforts directed at physicians. (This paper does not present data that characterise specific ADRs. These will continue to be presented in other venues.[12,13]) We therefore focus on patients’ representation of the doctor-patient interaction and physicians’ attribution, when patients report perceived ADRs. We used a commonly prescribed drug class as a demonstration case, the HMG-CoA reductase inhibitors (‘statins’).

Methods

Patients with perceived ADRs associated with statin use gave informed consent and completed a University of California, San Diego (UCSD) Institutional Review Board (IRB)-approved survey. Recruitment was passive, as is physician-targeted post-marketing surveillance. Of those offering source information (n = 433), patients learned of the study through Internet (72%), newspaper (25%) and television news (3%) reports that discussed the UCSD Statin Effects Study and its intent to learn about patient experiences with statin adverse effects. A total of 94% of subjects were from the US. The remaining 6% were from 11 nations spanning five continents. Patients were well educated on average, with a mean and median 16 years of education (range 7–25 years). The most commonly reported ADRs (all buttressed by a supportive literature) related to muscle,[14,15] cognition[16–18] and neuropathy.[19–21] For consenting patients who completed the main ADR survey and reported one or more of these symptoms, we probed their symptom experience with additional IRB-approved surveys targeting these areas.

These surveys elicited information on patient characteristics, drug, dose, ADR character, time course of onset with drug, severity, quality-of-life impact, recovery with discontinuation, and recurrence with reinitiationFootnote 1(these data are reported else-where[12,13]).

This paper focuses on patient reports of patient-physician interactions in relation to possible adverse effects; and the implications of these interactions for patient-targeted reporting. Thus, patients were asked if a drug contribution was discussed with their physician. They were asked whether the physician acknowledged/dismissed the possibility of a connection; whether patient or physician initiated discussion of a possible drug relation; and for specific surveys, whether the patient perceived their physician to appreciate the quality-of-life impact of the ADR (muscle and cognition survey). Optional narrative comments were elicited regarding the physician interaction. Surveys were abstracted for presumptive ADR causality criteria, based on factors such as whether symptoms arose on drug, abated with discontinuation and recurred with rechallenge.[22]

Results

A total of 650 patients, with a mean age of 63 years, completed a general ADR questionnaire. Of these patients, 54% of respondents were male. In total, 207, 113 and 85 patients completed supplemental targeted muscle, cognitive and neuropathy ADR surveys, respectively (48–58% male). Of the patients completing the muscle survey, 79% met presumptive literature criteria for probable or definite ADRs, whereas 77% of patients completing the cognition survey met presumptive literature criteria for probable or definite ADRs.

According to patients, physicians reportedly more commonly dismissed than acknowledged the possibility of a statin link to evaluated symptoms (table I). The likelihood of dismissal varied by ADR (p < 0.05) and was most common for neuropathy (51%). Physician response did not differ significantly by patient sex or age (<70 vs ≥70 years). Reportedly, the patient rather than the physician most often initiated discussion of a possible connection between drug and symptom (98% vs 2% cognition survey, 96% vs 4% neuropathy survey, 86% vs 14% muscle survey; p < 10[−8] for each). The physician initiated discussion more often for the best recognised ADR, muscle adverse effects, than for cognitive or neuropathy symptoms (p < 0.01). Thus, patients may consider a drug connection when a doctor does not and the disparity may be greater for lesser known adverse effects —where reporting may be most important. Many patients perceived that their physician did not appreciate the impact of their symptoms (61% muscle survey; 48% cognition survey).

Table I
figure Tab1

Reported physician response by adverse drug reaction (ADR) category

Full size image

A number of patients stated their physician attributed their problems to age or imagination, dismissed the importance of their symptoms or stated that a connection of symptoms to statins was not possible. Optional narrative comments on physicians’ response are illustrated in table II.

Table II
figure Tab2

Comments attributed to physicians

Full size image

Discussion

Among a sample of patients with possible to definite statin ADRs who completed surveys related to their experience, physicians were commonly reported to deny the possibility of a relationship of symptoms to drug, and to do so even when the patient noticed and reported an apparent association, when there was literature support for the ADR and when cases met literature criteria for probable or definite ADR causality. According to patients, physicians seldom initiated the conversations regarding a possible relation of the drug to the symptom (in fact, literature suggests that idiopathic polyneuropathy —i.e. after excluding other recognised causes —occurs with 16-fold elevated odds in statin users[19]). Since these patients considered a possible connection between a drug and new symptoms even when their physicians did not, ADR reporting by physicians may be complemented by patient reporting.

This study focused on known ADRs. Numbers for potential ‘new’ ADRs would be smaller, limiting inferences; thus dismissals of a possible drug connection to such symptoms might be thought justified. The finding that physicians (reportedly) dismiss the possibility of a relationship to the drug even for known ADRs meeting presumptive causality criteria provides concern they may do so, a fortiori, for ‘new’ hitherto unrecognised adverse effects.

Since this study is aimed at identifying whether patient targeting may add yield to postmarketing surveillance, by its nature it has limitations inherent to surveillance approaches: there is no defined base population or control group precluding generation of rates or risk ratios —thus, we draw no inferences regarding these. However, only patients who sustained a putative ADR are relevant to the goal of understanding the experience arising during an ADR, including patients’ reported interaction with their physician regarding the ADR. Self-selection to participate may impinge on external generalisability of findings, as in all studies with volunteer participants (physicians who elect to report ADRs to the US FDA MedWatch system are also self-selected and may also be non-representative). Moreover, patients with milder problems may lack the motivation to participate (also for physician-based reporting), while those with severe or fatal problems may be unable to participate. Patients whose physicians dismissed a connection may be more or less likely to participate, depending on whether physician dismissal dissuades patients from inferring a connection (stalling additional inquiry and study participation), or spurs dissatisfaction and additional inquiry. It is possible that findings may differ for statins than for other drugs: physician acknowledgement of symptoms as possible ADRs may either be greater than for many drugs due to more overall knowledge about this class of drugs; or may be lesser, if widespread reports of statins’ favourable benefit-risk profile[23–26] influence physicians to dismiss a possible connection to symptoms. Pharmacist reporting of ADRs is not addressed in the present study. It is unlikely to replace patient reporting since not all patients speak with their pharmacist about their ADR; however, it may serve an important role, augmenting physician and patient reporting.

As in all survey designs, data rely on self-report; therefore, recall and reporting bias may operate. Patients may falsely impute causality. However, for surveillance approaches, the goal of sensitivity over-rides that of specificity, since the approach should serve as an early signal of possible unusual patterns of reporting. Moreover, data here suggest physician inferences may also be erroneous, and may do so in a fashion that may perpetuate under-reporting. ADR causality evaluation algorithms enable presumptive external reappraisal of causal likelihood.

A systematic comprehensive active postmarketing surveillance effort targeted to patients as well as clinicians would address many concerns about ADR under-reporting, and about representativeness of ADR reporting. These concerns affect current passive surveillance approaches —targeted to healthcare providers, pharmacists and/or patients as here. However, until routine active surveillance is undertaken, the more reporting the better. Currently, regulatory agencies differ across nations in utilisation and acceptance of patient reporting.[27–30] These data suggest that patient-targeted reporting may complement provider-targeted reporting, adding yield, at little added cost.

Conclusion

It has been said that “pharmacovigilance —the science and activities relating to the detection, assessment, understanding, and prevention of adverse effects or any other drug-related problems … is one of the weakest links in drug regulation in the United States”[31] and that “finding potential drug-safety problems requires skillful observation by clinicians who are attuned to the possibility of drug-related adverse events and aware of the need to report them.”[32] Physicians commonly fail to report, even when aware of ADRs, contributing to a “gaping hole in the drug safety net”[8] —particularly given the importance of this reporting as “the most effective source of new ADR reports leading to changes in labeling.”[33,34] The data here, drawn from one drug class, suggest that awareness itself may also be a major issue: in at least some cases, physicians may fail to contemplate a possible drug connection for symptoms arising in their patients even when there is reason to consider one —and this failure may not be rare. These data suggest that patient-targeting may identify additional putative ADR cases, speeding accrual of reports. Less sparse accrual of ADR reports may facilitate more timely regulatory responses when these are merited, such as notification of potential adverse effects to physicians and the public, and may enable conduct of studies to identify vulnerabilities to mechanisms of and protections against these ADRs, which may improve patient care and outcomes.

Additional reporting of already-known ADRs is desirable if it can amplify awareness of ADRs, which may improve care for patients. Recognition of hitherto acknowledged ADRs may enable substitution of other, better tolerated medications for a patient, which may reduce morbidity and increase compliance, improving the quality of medication management. However, boosting recognition via patient involvement cannot readily lead to improved treatment unless the physician acknowledges the possibility of a connection.

Future efforts should consider collecting patient-targeted information using population-based, or pharmacy management-based approaches to ADR identification; such approaches will enhance generalisability of findings and may in addition permit determination of ADR rates and risk ratios. But until the resources are devoted to this task, patient-supplemented reporting may boost ADR reporting yield and foster more timely recognition of, and response to, ADRs.