Under-reporting of harm in clinical trials

Policy Review

Under-reporting of harm in clinical trials Bostjan Seruga, Arnoud J Templeton, Francisco Emilio Vera Badillo, Alberto Ocana, Eitan Amir, Ian F Tannock

Appropriate safety evaluations of anticancer drugs are crucial to assess their benefit–risk ratio. Substantial evidence shows that clinicians under-report harm in clinical trials, and at least three factors contribute to this problem: assessment of harm by clinicians might not represent the experience of patients; harm might be detected within trials, but is not reported appropriately by investigators or reporting is influenced by sponsors; and short-term follow-up might not detect long-term and potentially serious toxicities. Additionally, because of the selection of patients with good functional status in clinical trials, study results might not apply to patients treated in everyday clinical practice. New approaches for the conduct, oversight, and reporting of clinical trials should include patient-reported assessment of side-effects. Effective pharmacovigilance programmes and large-scale observational studies are needed to improve understanding of the tolerability of anticancer drugs in a real world setting.

Introduction The aim of phase 3 randomised controlled trials in oncology is to identify new therapies with a favourable benefit–risk ratio. However, existing approaches for assessment of outcomes (except for overall survival) are usually not patient centred and might not be optimal; specifically, information about side-effects and safety is almost exclusively reported by the trial investigators.1 Drug labels contain information about the safety of new drugs, which is based mainly on laboratory evaluations or clinicians’ impression of patients’ symptoms. However, 40–50% of adverse events reported on US Food and Drug Administration (FDA) labels for drugs for treatment of breast cancer and various non-malignant disorders are symptoms, which can only be assessed accurately by patient self-reporting.1 Furthermore, clinicians usually base their assessment of the benefit– risk ratio of a drug on reports of clinical trial results, which highlights the importance of a balanced presentation of study results by the investigators without influence of the sponsors (often pharmaceutical companies). Once approved, new anticancer drugs can be prescribed to patients with poorer overall health than those who participated in the clinical trials. Therefore, a benefit–risk ratio of new drugs might be less favourable in a real-world setting than in clinical trials. We review the problem of underdetection and under-reporting of harms of new anticancer drugs, and discuss possible solutions to mitigate it.

Reporting of harm Discrepancy Reporting of adverse events in clinical trials is one of the key components of clinical cancer research to ensure patients’ safety and clinicians’ understanding of the toxicity profiles of new anticancer drugs. The Common Terminology Criteria for Adverse Events (CTCAE), developed by the US National Cancer Institute (NCI) is a system of nomenclature to classify adverse events and their associated severity in clinical trials of anticancer drugs.2 This system contains three categories of adverse events: laboratory-based adverse events, observable or measurable adverse events, and symptomatic adverse www.thelancet.com/oncology Vol 17 May 2016

events reported by patients. The CTCAE has been updated continuously via a consensus-based process and has been adopted widely.3 The CTCAE was introduced in 1982 when most anticancer agents were given intermittently and had transient toxic effects. By contrast, continuous use of modern targeted agents can cause recurrent or chronic symptomatic toxicities. Low-grade chronic toxicities might substantially affect patients’ lives, and their effects might not be optimally captured by the existing CTCAE system.4 For example, grade 3 diarrhoea lasting for 1–2 days might be less bothersome for a patient than grade 2 diarrhoea lasting for several weeks. Agreement between different clinicians in reporting of symptomatic adverse events via the CTCAE is suboptimal (table 1).5 Whether the absence of reliability in reporting of adverse events among clinicians is attributable to the CTCAE, or whether it is an inherent limitation of reporting of adverse events by clinicians, is unknown. Moreover, because the process of reporting symptomatic adverse events is not standardised, interpretations of a patient’s symptoms by clinicians and research assistants might be suboptimal.13 Increasing evidence shows that, compared with patients, clinicians (physicians and nurses) under-detect and under-report symptomatic adverse events of anticancer drugs, both in everyday clinical practice1,6,7,9–11 and clinical trials6,8,12 (table 1). Several reasons might explain this effect: insufficient time during patients’ visits to fully discuss symptoms, under-reporting of symptoms by patients because of their desire to remain on therapy, and downgrading of symptoms by clinicians to justify continuation of treatment.5 Some symptoms experienced by patients might be attributed to underlying disorders (ie, non-treatment related) rather than to the investigational product (ie, treatment related), and might therefore not be reported properly by investigators. Furthermore, ascertainment of treatmentrelated and non-treatment-related adverse events in clinical trials might be prone to bias because the toxicity profile of standard therapy given to the control group is likely to be more widely understood than that of the experimental therapy.

Lancet Oncol 2016; 17: e209–19 Department of Medical Oncology, Institute of Oncology Ljubljana and University of Ljubljana, Ljubljana, Slovenia (B Seruga MD); Department of Medical Oncology, Kantonsspital St Gallen, St Gallen, Switzerland (A J Templeton MD); Department of Medical Oncology, Queen’s University, Kingston, ON, Canada (F E V Badillo MD); Medical Oncology Department and Translational Research Unit, Albacete University Hospital, Albacete, Spain (A Ocana MD); and Division of Medical Oncology & Hematology, Princess Margaret Cancer Centre, Department of Medicine, University of Toronto, Toronto, ON, Canada (Prof I F Tannock MD, E Amir MD) Correspondence to: Dr Bostjan Seruga, Department of Medical Oncology, Institute of Oncology Ljubljana, 1000 Ljubljana, Slovenia [email protected]

e209

Policy Review

Aim of study

Sample and study design

Results and interpretation

393 patients with cancer treated in a single centre (90 [23%] of whom were in clinical trials); a retrospective reliability study in which each patient was seen and assessed by two clinicians (physician or nurse), who independently rated seven CTCAE symptoms

Agreement between clinicians was moderate: interclass correlation coefficients were 0·46–0·69 for symptoms and were stable over time; different symptom ratings* that would affect treatment decisions occurred in 18% of patients with constipation, 8% of patients with nausea, and 15% of patients with vomiting

Clinician vs clinician symptom reporting Atkinson and To assess the reliability of adverse events reporting by clinicians for colleagues the same patient and same visit (2012)5

Clinician vs patient symptom reporting Di Maio and colleagues (2015)6

To compare reporting of adverse events by patients and physicians

1090 patients from three RCTs (breast cancer and NSCLC, including elderly patients); six symptomatic toxicities were prospectively assessed by investigators with the CTCAE at the first three visits; patients completed EORTC QoL questionnaires at the end of each cycle

Agreement between patients and physicians was low for all toxicities. For patients who reported toxicity of any severity, under-reporting by physicians was 41–74%; with examination of only patients who reported severe toxicity, under-reporting by physicians was 13–50%

Novello and colleagues (2014)7

To compare reporting of adverse events by patients and physicians

116 patients with advanced NSCLC treated with targeted drugs in everyday clinical practice; the survey required monthly compilation of physicians’ and patients’ questionnaires, basing adverse event assessment on CTCAE; physicians and patients assessed toxicity and QoL at three consecutive visits

Results show underestimation of toxicities by clinicians compared with patients; at every visit, a significant difference in perception of targeted therapy-related toxicities of any type and grade was described (p=0·0001 in all cases); the difference between physician and patient reporting was greater for adverse events more strongly associated with daily life and QoL

Quinten and colleagues (2011)8

To assess the extent to which patient and clinician symptom scoring and their agreement could contribute to the prediction of overall survival in patients with cancer

2279 patients with cancer from 14 closed EORTC clinical trials; investigators analysed baseline data for six cancer symptoms, which were assessed by both patients and clinicians; the extent of agreement and potential for clinician-reported or patient-reported symptom scores to improve the accuracy of Cox models to predict overall survival were assessed

Patient-reported scores for some symptoms, especially fatigue, differed from clinician-reported scores; Cox models of overall survival that considered both patient and clinician scores had more predictive accuracy than models that considered clinician’s scores alone for each of four symptoms

Basch (2010)1 To assess cumulative incidence of adverse events over time as reported by patients vs clinicians at successive office visits

467 patients with cancer; patient-reported symptoms collected at 4034 clinic visits; patients and clinicians reported adverse events according to the CTCAE; overall health status assessed by EuroQoL EQ-5D

Patients reported moderate-severity symptoms earlier and more frequently than did clinicians; patients’ reports were more highly concordant with overall health status than were clinicians’ reports

Basch and colleagues (2009)9

To compare how patients’ vs clinicians’ reports relate to sentinel clinical events

163 patients with lung cancer treated with chemotherapy; patients independently reported six CTCAE symptoms and Karnofsky Performance Status longitudinally at sequential office visits

Patients generally reported symptoms earlier and more frequently than did clinicians; significant associations with death and emergency room admissions were recorded for clinician reports of fatigue (p<0·001), nausea (p=0·01), constipation (p=0·038), and Karnofsky Performance Status (p<0·001), but not for patient reports of these items; higher concordance with EuroQoL EQ-5D questionnaire and global question scores was observed for patient-reported symptoms than for clinician-reported symptoms

Pakhomov and colleagues (2008)10

To assess agreement between patient-reported symptoms and documentation of these symptoms by physicians in electronic medical records

1119 patients: chest pain (N=373), dyspnoea (N=391), cough (N=337), multiple symptoms (N=18); three symptoms reported by patients were compared with those identified with language processing of the text of clinical notes from care providers

There was a positive agreement for 74 patients and a negative agreement for 78 patients with chest pain between patient’s report and clinical note; among 391 patients with dyspnoea, there was a positive agreement for 70 patients and a negative agreement for 76 patients; among 337 patients with cough, there was a positive agreement for 63 patients and a negative agreement for 75 patients; κ statistics were 0·50 for chest pain, 0·46 for dyspnoea, and 0·38 for cough

Basch and colleagues (2006)11

To compare reporting of symptom severity by patients and clinicians

400 patients with cancer at one institution; a questionnaire with 11 common CTCAE symptoms (an adapted version of the CTCAE) was given to consecutive outpatients and their clinicians (physicians and nurses); clinicians were aware that comparisons would be made

For most symptoms, agreement between patients and clinicians was high, and most discrepancies were within a grade difference of one point; agreement was lower for subjective symptoms; differences in symptom reporting rarely would have changed treatment decisions or dosing, and patients assigned greater severity to symptoms than did clinicians

Fromme and colleagues (2004)12

To compare patient reporting of eight symptoms using the EORTC QLQ-C30 questionnaire with physicians’ reporting of the same symptoms in the study’s adverse event log

37 men with mCRPC enrolled in a phase 2 study; a patient-reported symptom was defined as an increase in a symptom score by at least 10 points (on a 0–100 scale), sustained for at least 4 weeks; a physician-reported symptom was judged to be present if it was documented in the adverse event log

49 (new or worsened) symptoms were detected by both physician and patient, 48 symptoms were detected by the physician alone, and 55 symptoms were detected by the patient alone; corrected Cohen’s κ was 0·15, indicating only slight agreement; overall physician sensitivity was 47% and specificity was 68%

CTCAE=Common Terminology Criteria for Adverse Events. RCT=randomised controlled trial. NSCLC=non-small-cell lung cancer. EORTC=European Organisation for Research and Treatment of Cancer. QoL=quality of life. mCRPC=metastatic castrate-resistant prostate cancer. *The frequency of disagreement of two or more grade points between raters on the CTCAE, which might affect treatment decisions.

Table 1: Reporting of harm by clinicians and patients

Clinicians might not record symptoms reported by patients, and might also downgrade the severity of patients’ symptoms, which can lead to new, otherwise preventable, adverse events.10,14 For example, inappropriate management of nausea might lead to e210

serious fluid and electrolyte imbalances, which can result in kidney failure or arrhythmia. Furthermore, patient reporting seems to detect symptoms attributable to potentially serious adverse events earlier than does clinician reporting.1,15 Patients’ and clinicians’ www.thelancet.com/oncology Vol 17 May 2016

Policy Review

perspectives of adverse events seem to be complementary, together providing a more complete picture of the toxic effects of treatments than either perspective alone.15,16 Existing approaches to reporting of adverse events in clinical trials do not provide sufficient insight into the time course and duration of adverse events (eg, recurrent or chronic adverse events) and patients’ perspectives. Reporting patients’ health-related quality of life (HRQoL) with standardised and validated self-assessment questionnaires (eg, the European Organisation for the Research and Treatment of Cancer [EORTC] Quality of Life Questionnaire C30 [QLQ-C30]) could provide useful information, with the caveat that both the underlying cancer and its treatment will affect HRQoL. Unfortunately, reporting of HRQoL in clinical trials remains highly variable, both with regard to quality of reporting and data analysis and presentation, and such reports might not be interpretable by clinicians.17 For example, mean or median values of a summary measure of HRQoL do not capture individual variations in patient wellbeing. In contrast, a presentation of the proportion of patients with predefined improvement (or deterioration) of HRQoL can be much more informative for clinicians than presentation of mean or median values of a summary measure of HRQoL. Furthermore, some investigators and sponsors incorporate assessment of HRQoL into the study protocol and others do not. When a new experimental drug is expected to be toxic, assessment of HRQoL might be omitted. For example, nivolumab and cabozantinib showed similar efficacy to everolimus in patients with advanced renal cell carcinoma in two recent randomised controlled trials: in the CheckMate 025 trial,18 HRQoL of the less toxic drug nivolumab was assessed,18 but not in the Meteor trial19 of the more toxic cabozantinib. Furthermore, although data for HRQoL analyses are collected in some trials, results of such analyses might not be presented by investigators and sponsors.20 We are not aware of published data from pivotal randomised controlled trials that formally show that any recently approved anticancer drugs impair the HRQoL of patients with cancer, although their known toxicities make this very likely. This situation suggests bias in the conduct and reporting of HRQoL in randomised controlled trials.

Summary results Even if detection and reporting of individual adverse events in clinical trials are not compromised, poor-quality reporting of summary results of clinical trials at conferences and in journals might have an unfavourable effect on clinical practice. Bias in reporting of clinical trials and selective presentation of results can create false perceptions of drug efficacy and safety, not only in the trial itself but also in everyday clinical practice and in subsequent systematic reviews and meta-analyses.21,22 Harm might be detected in clinical trials but not reported appropriately by the investigators.23 To improve reporting www.thelancet.com/oncology Vol 17 May 2016

of adverse events in randomised controlled trials, the Consolidated Standards of Reporting Trials (CONSORT) guidance was extended to provide a set of ten specific and comprehensive guidelines for reporting harms.24 However, despite this extension of CONSORT guidelines, under-reporting of harm remains prevalent and is not improving over time25,26 (table 2). Furthermore, lack of standardisation in reporting of adverse events beyond the CONSORT guidelines leads to challenges in interpretation of safety data for anticancer drugs across studies. Whether all adverse events should be reported or only those occurring in a specific proportion of patients, whether grade 1–2 and grade 3–4 adverse events should be pooled or presented individually, or whether only the highest grade of a specific adverse event experienced by an individual patient should be reported is not clear. For example, although a patient with two or three episodes of grade 3 diarrhoea is likely to have suffered more than a patient with one episode of grade 3 diarrhoea during the course of a specific treatment, these two patients usually contribute equally to the safety report according to present reporting practice. Several studies have assessed reporting of harm in oncology randomised controlled trials using various methods, and are unanimous in finding that reporting of harm is inadequate (table 2).25–31 Reporting of harm is not only often incomplete but is also biased and inconsistent with protocols.28 Under-reporting of detected harm might occur at different levels of study data processing. In trials sponsored by the NCI, substantial discrepancies exist between high-grade adverse events reported in the sponsor’s database (ie, the NCI Clinical Data Update System) and those reported in published articles.31 Although 111 (27%) of 413 high-grade adverse events reported in published articles could not be matched to agent-attributable adverse events in the sponsor’s database, 118 (28%) of 423 high-grade adverse events listed in the sponsor’s database could not be matched to adverse events in the corresponding article.31 Worryingly, even key adverse events are under-reported in clinical trials.31 An analysis of randomised controlled oncology trials showed that only 17% described the reporting threshold above which grade 3–4 adverse events were included for reporting, 27 raising the possibility that important but less frequent grade 3–4 adverse events might be under-reported. Additionally, the frequency and causality of grade 5 adverse events (ie, those leading to death) were adequately reported in only 161 (50%) of 325 trials assessed in this study,27 adverse events leading to study withdrawal in 61 (19%), and adverse events leading to dose reduction in 43 (13%).27 In a previous study, we analysed reports of randomised controlled trials for women with breast cancer published in major oncology journals between 1995 and 2011 and found that 67% of reports were biased in the reporting of toxicity, and only 32% of articles reported the frequency of grade 3 and 4 toxicities in the abstract. In this study,29 a positive primary endpoint was associated with e211

Policy Review

Aim of study

Sample and study design

Results and interpretation Only 56 (17%) of 325 RCTs described the reporting threshold above which grade 3–4 adverse events were included; deaths were adequately reported in 161 (50%), adverse events leading to study withdrawal in 19%, and adverse events leading to dose reduction in 43 (13%) of RCTs; in contrast, for each of these categories (deaths, adverse events leading to study withdrawal, and adverse events leading to dose reduction) 201 (96%), 181 (86%), and 148 (70%) of 210 participating EORTC members expected comprehensive reporting, respectively

Maillet and colleagues (2015)27

To analyse reporting of key adverse 325 RCTs published between 2007 and 2011 analysed for reporting of key adverse events* and compared events in oncology RCTs and with EORTC membership expectations compare it with expectations of members of the EORTC

Sivendran and colleagues (2014)25

To assess the extent to which CONSORT recommendations are followed in publications

175 RCTs in metastatic solid malignancies published between 2009 and 2011 were analysed for 14 adverse events—reporting elements derived from the CONSORT harms extension statement (completeness score range 0–14)

Median completeness score was 8 (3–12); most publications (168 [96%] of 175) reported only adverse events occurring above a threshold rate or severity, 65 (37%) did not specify the criteria used to select which adverse events were reported, and 154 (88%) grouped together adverse events of varying severity; trials without a stated funding source and with an earlier year of publication had significantly lower completeness scores than trials with a stated funding source and later trials

Mhaskar and colleagues (2014)28

To investigate how often the assessment of treatment-related harms is prespecified in protocols and reported in final publications, and to assess the effect of data collection guidelines on the quality of reporting of harms

Data from 426 consecutive NCICOG RCT protocols and matching publications for 1955–2006 were extracted and analysed for reporting of harm

Although the NCICOG trials were well planned, the quality of reporting of harms was poor: 115 (27%) of 429 trials had good quality reporting for treatment-related harms data compared with 90% of trials for treatment-related benefits data; only 201 (46%) of RCTs reported data collection instrument or methods

Péron and colleagues (2013)26

325 oncology RCTs published in 2007–11 were analysed The most common items poorly reported were: the methods of adverse event To assess adherence to the collection (adequately reported in only 33 [10%] of 325 studies), the extended CONSORT guidelines on by the 16-point adverse event reporting quality score description of adverse event characteristics leading to withdrawals (49 [15%]), reporting of adverse events in RCTs based on the CONSORT extension and whether adverse events were attributed to trial interventions (124 [38%]); the quality of adverse event reporting did not improve significantly over time and was not better in articles published in journals with a high impact factor than in lesser known journals

Vera-Badillo To assess the quality of reporting of Bias in the reporting of the primary endpoint and of and colleagues the primary endpoint and of toxicity was predefined and assessed in 164 RCTs for (2013)29 breast cancer using predesigned algorithms; toxicity in RCTs for breast cancer association of bias with some key characteristics was investigated

Among 164 RCTs, 54 (33%) showed bias in reporting of the primary endpoint and 110 (67%) in the reporting of toxicity; 54 (59%) of 92 trials with a negative primary endpoint used secondary endpoints to suggest benefit of experimental therapy; only 52 (32%) of articles indicated the frequency of grade 3–4 toxicities in the abstract; a positive primary endpoint was associated with under-reporting of toxicity

Pitrou and colleagues (2009)30

To describe the reporting of harm-related results from RCTs

133 RCTs (including 16 oncology RCTs) assessed for the reporting of harm by standardised data extraction developed on the basis of a review of the literature and the CONSORT statement guidelines

Adverse events were mentioned in 118 (89%) of 133 reports; no information on severity of adverse events and withdrawal of patients owing to an adverse event was given in 36 (27%) and 63 (47%) of articles, respectively; restrictions in reporting of harm-related data were noted in 43 (32%) articles, with a description of the most common adverse events only (n=17), severe adverse events only (n=16), statistically significant adverse events only (n=5), and a combination of restrictions (n=5); the population considered for safety analysis was clearly reported in tables of 59 (66%) of 90 reports, in which safety results were presented in tables

Scharf and colleagues (2006)31

To assess whether published adverse event data differ from those in the sponsor’s database and from the data collection requirements stated in study protocols

22 studies identified in the NCI’s database (CDUS) that used the CTC version 2.0 and for which a final study publication was available; protocol information pertaining to adverse event collection and reporting methods was compared with methods cited in the article; comparison of the adverse event data in the trial publication with the adverse event data submitted by the investigators to CDUS was made

Substantial inconsistency was found between adverse event collection and reporting methods cited in the protocols vs final publications; 27% of high-grade adverse events in articles could not be matched to agent-attributable adverse events in the CDUS, and 28% of CDUS high-grade adverse events could not be matched to adverse events in the corresponding article; in 14 of 22 articles, the number of high-grade adverse events in CDUS differed from the number in the articles by 20% or more

EORTC=European Organisation for Research and Treatment of Cancer. RCT=randomised controlled trial. CONSORT=Consolidated Standards for Reporting Trials. NCICOG=National Cancer Institute’s Co-Operative Group. NCI=US National Cancer Institute. CTC=Common Toxicity Criteria. CDUS=Clinical Data Update System. *Grade 3–4 adverse events, grade 5 adverse events, and adverse events resulting in study withdrawal or in dose reduction.

Table 2: Studies assessing reporting of harm in oncology clinical trials

under-reporting of toxicity, which contrasts with a previous observation.32 Furthermore, our study showed that, in 59% of randomised controlled trials with a negative primary endpoint, investigators used a secondary endpoint to suggest some benefit of the experimental intervention.29 Investigators, sponsors, or both might focus on efficacy as the basis of registration and downplay toxicity to make the results seem more positive than they actually are. e212

First and updated reports Clinicians are often enthusiastic when the first results of positive randomised controlled trials are presented. However, initial analyses might be based on immature data, and therefore efficacy of new anticancer therapies might be overestimated and toxicity under-reported. Updated reports of randomised controlled trials based on more mature data might better inform oncologists about the benefit–risk ratio of new anticancer therapies. www.thelancet.com/oncology Vol 17 May 2016

Policy Review

For example, non-final analyses of oncology studies, which are often presented as abstracts, might be discordant with subsequent publications, and discrepancies have been found to be large in 17 (10%) of 167 studies.33 Randomised controlled trials that are stopped early for benefit (truncated randomised controlled trials) are becoming more common.34 Early in the progress of the trial, random fluctuations in the estimated treatment effect can occur, and there is concern that the treatment effect of new therapies in such trials is overestimated. Bassler and colleagues35 showed that truncated randomised controlled trials overestimate the relative size of benefit compared with trials that are not stopped early; the pooled ratio of relative risks (RRs) for truncated versus non-truncated randomised controlled trials was 0·71 (95% CI 0·65–0·77). The estimated pooled difference in relative benefits between truncated and non-truncated randomised controlled trials was 26% in this study (pooled RR 0·56 [0·52–0·61] for truncated trials and 0·82 [0·78–0·86] for non-truncated trials; p<0·001), suggesting that pooled treatment effect size was substantially higher in truncated studies than in non-truncated studies.35 Differences between truncated and non-truncated studies were independent of the presence of statistical stopping rules and were largest in smaller studies.35 Treatment effect (ie, ratio of RRs <0·75) was overestimated when fewer than 500 events occurred in truncated studies.35 Furthermore, for 39 (62%) of 63 research questions, the pooled effects of the non-truncated randomised controlled trials failed to show the significant benefit shown in truncated trials. Such overestimation of treatment effect might mislead clinicians and patients when they try to assess the benefit–risk ratio of new drugs. These findings also have implications for interpretation of interim analyses, despite adjustment of the significance level, since results are based on a small number of events. Updated analyses might improve our understanding of the benefit–risk ratio of anticancer drugs in randomised controlled trials. In a pooled analysis, 311 published reports of randomised controlled trials in breast cancer, non-small-cell lung cancer, and prostate cancer, we identified updated reports of efficacy and safety for only 64 (21%) and 39 (12%) of 311 trials, respectively.36 Independent predictors for an updated analysis were cancer site, conduct of an interim analysis, larger sample size, and larger treatment effect. Investigators (and sponsors) might be more inclined to update the results of randomised controlled trials with large sample size and large treatment effect because benefit in such trials is more likely to remain stable over time. Overall, according to updated reports, more patients experienced low-grade and high-grade adverse events, and the magnitude of the treatment benefit decreased when results were www.thelancet.com/oncology Vol 17 May 2016

updated.36 Therefore, in initial reports of randomised controlled trials, efficacy of anticancer drugs might be overestimated and safety under-reported.

Real-world experience of harm Narrow eligibility criteria decrease the heterogeneity of patients enrolled in trials and increase statistical power, but might result in differences between study populations and the general population of patients with cancer, and therefore reduce the generalisability of study results.37 In the era of personalised medicine, eligibility criteria for participation in clinical trials might become even more restrictive than in the past. A substantial proportion of patients treated with new anticancer therapies in everyday clinical practice do not fulfil the eligibility criteria for participation in the corresponding pivotal randomised controlled trials in which these therapies are assessed (table 3).38–42 Limited evidence shows that trial results are generalisable only for patients treated in everyday clinical practice who meet the standard eligibility criteria used in a specific clinical trial.42 Patients not participating in clinical trials might have worse survival and experience increased toxicity compared with patients eligible for clinical trials (table 3). For example, in our study41 of men with metastatic castration-resistant prostate cancer, patients who received docetaxel in clinical practice had poorer median survival than did those treated with docetaxel in clinical trials (13·6 months [95% CI 12·1–15·1] vs 20·4 months (17·4–23·4); p=0·007) and experienced more febrile neutropenia (9·6% vs 0; p=0·035). These findings seem to be mainly a consequence of patients’ characteristics rather than trial participation (eg, more uniform rules for dose adjustment or toxicity management). Thus, when anticancer drugs are used in an unselected population of patients, marginal benefits reported in randomised controlled trials might decrease or even disappear, possibly when toxic anticancer drugs are used in health systems with scarce resources.44 This problem can be addressed by the use of wider eligibility criteria for clinical trials and confirmation of benefit seen in randomised controlled trials in population-based health outcome studies. Early access and other similar programmes bridge the gap between completion of clinical trial accrual and approval of anticancer drugs. Often such programmes claim to represent real-world experience.45,46 However, patients participating in these programmes usually have to fulfil eligibility criteria similar to those stated in the protocols of randomised controlled trials (NCT01254279) and the occurrence of toxicity in patients in such programmes is often not captured systematically. The independent validity of results of these programmes should be interpreted cautiously because results probably reflect randomised controlled trials and might similarly overestimate benefit and under-report toxicity of anticancer drugs. e213

Policy Review

Setting and type of therapy

Sample size and source of real-world data

Methods

Patients’ eligibility for participation in trials

Outcome

Mitchell and colleagues (2015)38

Advanced RCC treated with first-line antiangiogenic drugs or temsirolimus

N=438; retrospective registry of patients with advanced RCC

Baseline characteristics of registry patients compared with those participating in pivotal phase 3 clinical trials

171 (39%) of patients ineligible

Not reported

Treewek and colleagues (2015)39

Early breast cancer; endocrine therapy and radiotherapy

N=4811; women diagnosed with early breast cancer in the Tayside region (Scotland, UK) over a period of 16 years

Eligibility criteria from 12 major trials were applied to women with early breast cancer

1276 (27%) of women not eligible for any trial (eligibility for an individual trial rarely exceeded 45%)

Not reported

Heng and colleagues (2014)40

Advanced renal cell carcinoma; antiangiogenic therapy

N=2210; IMDC

Patients retrospectively judged to be ineligible for clinical trials according to commonly used inclusion and exclusion criteria found in the phase 3 registration trials (not a comprehensive list)

768 (35%) of patients ineligible

Median overall survival: 12·5 months for trial-ineligible patients vs 28·4 months for trial-eligible patients (p<0·0001); when adjusted for the IMDC prognostic factors, overall survival of ineligible patients still worse than that of eligible patients (HR 1·55, 95% CI 1·37–1·75; p<0·0001)

Templeton mCRPC; first-line and colleagues docetaxel (2013)41

N=438; men with mCRPC treated with docetaxel at a single institution

Comparison of overall survival and toxicity in men with mCRPC who were treated with docetaxel every 3 weeks in everyday clinical practice and within trials

Not reported

Median overall survival 13·6 months (95% CI 12·1–15·1; clinical practice) vs 20·4 months (17·4–23·4; clinical studies; p=0·007), vs 19·3 months (17·6–21·3; in the TAX327 trial; p<0·001); incidence of febrile neutropenia: 9·3% (clinical practice) vs 0 (clinical studies; p=0·035) vs 3% (TAX327; p<0·001); deaths within 30 days of final dose were 4% (clinical practice) vs 0 (clinical studies; p=0·24) vs 3% (TAX327; p=0·68)

Mol and colleagues (2013)42

Metastatic colorectal cancer; chemotherapy

CAIRO trial treated patients (N=162) and patients treated concurrently in everyday clinical practice (N=396 in 29 Dutch hospitals (identified by the NCR)

Comparison of the outcome of CAIRO trial patients and patients treated in everyday clinical practice

224 (43·5%) of patients ineligible

Overall survival of eligible non-trial patients was similar to that of trial patients (median overall survival 17·0 months vs 15·7 months; HR 1·03 [95% CI 0·87–1·23], p=0·70); non-eligible non-trial patients had a significantly worse overall survival than did trial patients (median overall survival 17·0 months vs 9·3 months; HR 1·70 [1·33–2·17], p<0·01)

Sorbye and colleagues (2009)43

Metastatic colorectal cancer, chemotherapy

N=760; patients referred to three hospitals in Scandinavia for their first oncology consultation

To investigate whether patients enrolled in trials are representative of a general population of patients with cancer for patient characteristics and survival

Of 295 who received chemotherapy but did not participate in trials, 202 (69%) were ineligible

Trial patients had better prognostic characteristics and significantly longer overall survival than did non-trial patients: 21·3 months (18·2–24·4) vs 15·2 months (13·6–16·8; p=0·005)

RCC=renal cell carcinoma. MDC=International Metastatic RCC Database Consideration. mCRPC=metastatic castration-resistant prostate cancer. NCR=Netherlands Cancer Registry.

Table 3: Eligibility of real-world patients with cancer for participation in clinical trials and their outcome

Pharmacovigilance for anticancer drugs Investigators in large randomised controlled trials usually detect and report common adverse events, but not rare adverse events or those that occur after a latent period. Furthermore, rare but potentially serious adverse events associated with anticancer drugs might not be detected because of the small sample size of randomised controlled trials, because patients with selected characteristics are enrolled (eg, those with minimal comorbidities or few concurrent medications), and because of the quite short e214

follow-up. By contrast with classic chemotherapy, continuous use of targeted anticancer drugs is more likely to chronically modify the function of normal tissues and produce unexpected long-term toxic effects. Furthermore, adverse events classified as mild in randomised controlled trials might present as severe and potentially life threatening in less selected populations. After approval of an anticancer drug by a regulatory agency, information about safety is collected in the post-marketing period from various sources, including www.thelancet.com/oncology Vol 17 May 2016

Policy Review

spontaneous reporting, post-marketing phase 4 studies, and ongoing clinical trials.47 By contrast with adverse events reported in randomised controlled trials, adverse drug reactions are events for which a causal link is strong enough to warrant attribution of the event to the drug. More than half of all serious adverse drug reactions are identified 7 years or more after a drug is approved by the FDA.48 For example, osteonecrosis of the jaw caused by bisphosphonates was not detected in pivotal randomised controlled trials, but was discovered only in the post-marketing period and appeared in the updated drug label of zoledronic acid about 2·5 years after its approval by the FDA.49 In our 2011 study, we analysed 76 serious adverse drug reactions reported in updated drug labels of 12 targeted drugs, of which 38 (50%) were potentially fatal.50 30 (39%) of all these serious adverse drug reactions and 15 (39%) of 38 potentially fatal adverse drug reactions were not described in any published report of randomised controlled trials, and 37 (49%) of serious adverse drug reactions and 22 (58%) of potentially fatal adverse drug reactions were not described in initial drug labels. Worryingly, first drug labels contained less information about adverse drug reactions than did reports of randomised controlled trials. For example, cardiopulmonary arrest or sudden death in patients with head and neck cancer treated with cetuximab and radiotherapy appeared as a black box warning (ie, the highest level of alert) only in the updated drug label of cetuximab 13 months after its approval.51 In our study, after a median of 4·3 years (range 1·3–11·9 years) between initial drug approval and update of drug labels, 4 (42%) of 12 targeted cancer drugs acquired one or more boxed warnings.50 Cases have been documented in which pharmaceutical companies failed to provide mortality data to the FDA in a timely manner, thereby minimising estimates of the risk of mortality and creating a false impression of the safety of some non-cancer drugs.52,53 Furthermore, regulatory agencies might not present some confidential information about safety in publicly available drug safety reports. Regulatory documents outlining the clinical basis for approval of new drugs do not necessarily disclose all the relevant information required to assess safety.54 Voluntary reports of adverse drug reactions can be submitted directly by health professionals or patients to regulatory agencies or through drug manufacturers. However, reporting of adverse drug reactions by health professionals and patients has been identified and might result in delayed detection of serious safety signals. A systematic review showed that a substantial proportion of adverse drug reactions identified in published reports was never reported to local, regional, or national spontaneous reporting systems in the UK;55 the median incidence of under-reporting for all adverse drug reactions was 94% (IQR 82–98), and for serious adverse drug reactions was 85% (IQR not available).55 The need www.thelancet.com/oncology Vol 17 May 2016

for timely access to new anticancer drugs has led to a need for rapid review by regulatory authorities for new drug approval, but more safety problems can occur once these new drugs are used in everyday clinical practice.56 Health professionals and patients can make important contributions to drug safety by spontaneous reporting.57 However, analysis of adverse drug reaction patient reporting systems from 50 countries showed that only 9% of all safety reports are from patients, with the rest of reports being from health-care professionals.58 In view of the underdetection and under-reporting of harm in clinical trials, development of more effective and systematic pharmacovigilance programmes deserves high priority.

Effect of under-reporting harm Under-reporting of harm affects many stakeholders, including clinicians, patients, drug developers, regulators, and payers.15 Patients discussing possible treatment options with health professionals during the process of informed decision making might receive incomplete or inaccurate information about the safety of new therapies. When a disease is incurable, prolongation of life by aggressive and potentially toxic treatments might not be a priority for patients with cancer.59 In early-phase clinical trials, drug developers try to identify the maximum tolerable doses of anticancer drugs. However, if the subjective symptoms of patients enrolled in early-phase clinical trials are neglected, estimates of the maximum tolerable dose might be too high. Symptoms reported by patients but not recognised by clinicians affect treatment decisions in a substantial proportion of patients treated in everyday clinical practice (table 1).5 Adverse effects of drugs that seem to be tolerable in a highly selected population of patients in clinical trials might not be tolerated by patients in routine practice, and adherence to such treatment is likely to be poor. Regulatory approval is based on objective evidence of efficacy and safety. However, assessment of the benefit–risk ratio is a complex process during which there is an attempt to balance several efficacy and safety outcomes based on expert opinion rather than explicit criteria.60 Under-reporting of harm might have an unfavourable effect on licensing and approval: it is concerning that regulatory agencies often seem to prioritise drug approval by emphasising efficacy outcomes with less attention to safety.61 Economic considerations are usually the responsibility of health technology assessment organisations and payers, who base their decision to pay for new therapies on economic analyses.62 Economic analyses are based on clinical trial data, and under-reporting of harm in clinical trials, and underestimates of the costs of managing harm, will decrease the validity of such economic analyses when applied to patients in general oncology practice. e215

Policy Review

Possible solutions Several possible methods exist to mitigate underdetection and under-reporting of harm (figure). A patient-centred approach should be the main focus of clinical trials, and clinical trials should mimic the real-world setting as closely as possible. A group of multidisciplinary investigators and patients collaborating under the NCI has developed a patient-reported outcomes version of the CTCAE (PRO-CTCAE), which integrates patient perspective into adverse event reporting in clinical trials.13 Of the 790 adverse events reported in the CTCAE, 78 were appropriate for patient self-reporting. For each of these adverse events, a PRO-CTCAE plain-language term in English and one to three items characterising the frequency, severity, and activity interference of the adverse events were created, producing a library of 124 PRO-CTCAE items. After refinement, a software platform was built to administer PRO-CTCAE items to clinical trial participants. A recent study63 showed favourable validity, reliability, and responsiveness of PRO-CTCAE in a large, heterogeneous sample of US patients undergoing cancer treatment. In this study, primary comparison was made to the clinician-reported Eastern Cooperative Oncology Group (ECOG) performance status and patient-reported EORTC QLQ-30. Results of a randomised controlled trial64

Existing and potential problems

Possible solutions

Clinical trial Patients Investigators under-detect harm Investigators

Presentation of summary study results by investigators and sponsors might be biased

Sponsor (often pharmaceutical industry)

Patient-centred approach to detection of harm (eg, PRO-CTCAE), broader eligibility criteria, and trials in vulnerable populations of patients

Unbiased and standardised presentation of study results, including HRQoL data (eg, required by journal editors)

Safety and HRQoL data in clinical trials overseen and reported to regulators by independent academic clinicians

Sponsors might convey incomplete information about harm to regulators Regulatory agency Regulatory agencies do not disclose all relevant information about harm to the public

Pharmacovigilance programmes are suboptimum and true real-world data are scarce

Regulatory agencies are required to disclose all safety information about harm to the public

Post-marketing setting (ie, real-world)

Clinicians and patients are encouraged to voluntarily report harm, and governments to support conduct of large-scale observational studies

Figure: Under-reporting of harms during drug development, drug approval, and post-marketing, and suggested solutions to mitigate it HRQoL=health-related quality of life. PRO-CTCAE=patient-reported outcomes version of the Common Terminology Criteria For Adverse Events.

e216

assessing self-reporting of toxicity showed that self-assessment of symptoms by patients is associated with improved HRQoL and prolonged quality-adjusted survival in patients receiving outpatient chemotherapy for solid tumours. The PRO-CTCAE has already been integrated into some ongoing oncology clinical trials.13 However, although detection of harm by the PRO-CTCAE is an important development, its reporting by investigators and sponsors might be suboptimal if reporting of adverse events is not standardised beyond the CONSORT requirements. Patients should have an active role in reporting adverse events not only in clinical trials but also in everyday clinical practice. Multidisciplinary working groups are developing a standardised set of patient-centred outcomes to be followed during routine care.65 Standardised measurement and reporting of patient-reported outcomes in everyday clinical practice and their inclusion in population-based datasets could lead to improved assessment of quality of health care, and could accelerate the uptake of best practices. Furthermore, clinicians and patients should be encouraged to participate actively and systematically in pharmacovigilance programmes.66 Improvement and better resourcing of the surveillance systems are needed. Investigators and pharmaceutical companies need to modify the design, conduct, analysis, and presentation of randomised controlled trials. Eligibility criteria should be broadened, and specific trials should be designed to address the unmet needs of patients who were not eligible to participate in drug registration trials. Such studies are especially relevant for drugs for which there is a substantial expectation of toxicity.67,68 Since early reports of randomised controlled trials might be discordant with those published later, investigators and sponsors should be required to update analyses and report their results as a condition for registration of new therapies. Updated reports of randomised controlled trials and results of post-marketing studies should be a requirement for permanent approval of anticancer drugs. Additionally, HRQoL should be assessed routinely in all potentially practice-changing trials, and standardised reporting of HRQoL should be an absolute requirement for journal editors and regulators.69 Furthermore, because HRQoL data are important for assessment of the benefit–risk ratio, such analyses should be presented in the original publication and not in a separate report. Space limitation in most journals might make inclusion of HRQoL data difficult, but we suggest that space dedicated to subset and other secondary analyses, which have low statistical validity, should be replaced by HRQoL data in the original publication of pivotal trials. Less relevant clinical data can be presented in accompanying online supplementary files. Governments and academia should encourage large-scale population-based observational studies in www.thelancet.com/oncology Vol 17 May 2016

Policy Review

Search strategy and selection criteria We searched PubMed with the terms “adverse event”, “symptomatic toxicity”, “patient-reported”, “reporting”, “updated”, “final report”, “eligibility criteria”, “clinical trial participants”, “real-world”, and “cancer” from Jan 1, 2000, to Nov 15, 2015. We reviewed only papers published in English. We cross-referenced all papers to identify further relevant studies. The final reference list was generated on the basis of originality and relevance to the broad scope of this Policy Review.

which patterns of care, toxicity, and the effectiveness of treatment in routine practice can be assessed.70 Such studies are enabled by modern bioinformatics, which enables linkage between different databases, such as cancer registries and hospital records. If outcomes identified in such large-scale studies are inferior to those reported in a randomised controlled trial, patient selection and delivery of treatment should be considered carefully.71 By contrast, in some large-scale observational studies the results from pivotal randomised controlled trials have been reproduced, which is reassuring.72 Observational studies are also important because they enable detection of some rare late and serious toxicities.73 Increased transparency of the information of pharmaceutical companies related to toxicity data should be demanded by the oncology community. Since pharmaceutical companies might delay reporting of serious and fatal toxicities to regulatory agencies, direct submission of spontaneous reports to regulatory agencies rather than passive submissions via industry should be mandated.74 We suggest that independent academic clinicians oversee safety and HRQoL data in industry-sponsored trials, and such data should be publicly available. Documents published by regulatory agencies should disclose all relevant information necessary to adequately assess drug safety.

Conclusion Under-reporting of harm can result in a false perception of the benefit–risk ratio of anticancer drugs. The conduct, analysis, and reporting of clinical trial results should be improved. Some strategies to diminish underdetection and under-reporting of harm have been proposed, and their widespread use and rigorous guidelines for detection and reporting of harm are needed. Increased transparency of safety data in registration trials is required. Journal editors and regulators should demand higher standards for reporting of harm and HRQoL data from randomised controlled trials. Effective pharmacovigilance programmes and large-scale observational studies are essential to assess application of new therapies in general oncology practice and can address the problem of underdetection and under-reporting of harm in clinical trials. www.thelancet.com/oncology Vol 17 May 2016

Contributors BS came up with the concept, did the literature search, collected and interpreted data, drafted the report, and created the tables and figure. AJT and FEVB came up with the concept, interpreted data, and revised the text and tables. AO wrote and reviewed the report. EA interpreted data and revised the text and tables. IFT came up with the concept, interpreted data, revised the text and tables, and sourced funding. All authors approved the final version of the report. Declaration of interests BS has received consultancy fees from Sanofi, Astellas, and Janssen. AJT’s institution received consultancy fees from Sanofi, Astellas, and BMS (no personal compensation). IFT has received consultancy fees from Janssen and Ipsen Pharma. FEVB, AO, and EA declare no competing interests. References 1 Basch E. The missing voice of patients in drug-safety reporting. N Engl J Med 2010; 362: 865–69. 2 US Department of Health and Human Services. National Institutes of Health. National Cancer Institute: Common Terminology Criteria for reporting of Adverse Events (CTCAE). Version 4. May 28, 2009. http://ctep.cancer.gov/protocolDevelopment/electronic_ applications/ctc.htm#ctc_40 (accessed April 4, 2016). 3 Trotti A, Colevas AD, Setser A, et al. Patient-reported outcomes and the evolution of adverse event reporting in oncology. J Clin Oncol 2007; 25: 5121–27. 4 Edgerly M, Fojo T. Is there room for improvement in adverse event reporting in the era of targeted therapies? J Natl Cancer Inst 2008; 100: 240–42. 5 Atkinson TM, Li Y, Coffey CW, et al. Reliability of adverse symptom event reporting by clinicians. Qual Life Res 2012; 21: 1159–64. 6 Di Maio M, Gallo C, Leighl NB, et al. Symptomatic toxicities experienced during anticancer treatment: agreement between patient and physician reporting in three randomized trials. J Clin Oncol 2015; 33: 910–15. 7 Novello S, Capelletto E, Cortinovis D, et al. Italian multicenter survey to evaluate the opinion of patients and their reference clinicians on the “tolerance” to targeted therapies already available for non-small cell lung cancer treatment in daily clinical practice. Transl Lung Cancer Res 2014; 3: 173–80. 8 Quinten C, Maringwa J, Gotay CC, et al. Patient self-reports of symptoms and clinician ratings as predictors of overall cancer survival. J Natl Cancer Inst 2011; 103: 1851–58. 9 Basch E, Jia X, Heller G, et al. Adverse symptom event reporting by patients vs clinicians: relationships with clinical outcomes. J Natl Cancer Inst 2009; 101: 1624–32. 10 Pakhomov SV, Jacobsen SJ, Chute CG, Roger VL. Agreement between patient-reported symptoms and their documentation in the medical record. Am J Manag Care 2008; 14: 530–39. 11 Basch E, Iasonos A, McDonough T, et al. Patient versus clinician symptom reporting using the National Cancer Institute Common Terminology Criteria for Adverse Events: results of a questionnaire-based study. Lancet Oncol 2006; 7: 903–09. 12 Fromme EK, Eilers KM, Mori M, Hsieh YC, Beer TM. How accurate is clinician reporting of chemotherapy adverse effects? A comparison with patient-reported symptoms from the Quality-of-Life Questionnaire C30. J Clin Oncol 2004; 22: 3485–90. 13 Basch E, Reeve BB, Mitchell SA, et al. Development of the National Cancer Institute’s patient-reported outcomes version of the Common Terminology Criteria For Adverse Events (PRO-CTCAE). J Natl Cancer Inst 2014; 106: dju244. 14 Weingart SN, Simchowitz B, Shiman L, et al. Clinicians’ assessments of electronic medication safety alerts in ambulatory care. Arch Intern Med 2009; 169: 1627–32. 15 Basch E, Bennett A, Pietanza MC. Use of patient-reported outcomes to improve the predictive accuracy of clinician-reported adverse events. J Natl Cancer Inst 2011; 103: 1808–10. 16 Alberti P, Rossi E, Cornblath DR, et al. Physician-assessed and patient-reported outcome measures in chemotherapy-induced sensory peripheral neurotoxicity: two sides of the same coin. Ann Oncol 2014; 25: 257–64. 17 Brundage M, Bass B, Davidson J, et al. Patterns of reporting health-related quality of life outcomes in randomized clinical trials: implications for clinicians and quality of life researchers. Qual Life Res 2011; 20: 653–64.

e217

Policy Review

18

19

20

21

22 23 24

25 26

27

28

29

30

31

32

33

34

35

36

37 38

39

40

e218

Motzer RJ, Escudier B, McDermott DF, et al. Nivolumab versus everolimus in advanced renal-cell carcinoma. N Engl J Med 2015; 373: 1803–13. Choueiri TK, Escudier B, Powles T, et al. Cabozantinib versus everolimus in advanced renal-cell carcinoma. N Engl J Med 2015; 373: 1814–23. Tannock IF, Fizazi K, Ivanov S, et al. Aflibercept versus placebo in combination with docetaxel and prednisone for treatment of men with metastatic castration-resistant prostate cancer (VENICE): a phase 3, double-blind randomised trial. Lancet Oncol 2013; 14: 760–68. Dwan K, Altman DG, Clarke M, et al. Evidence for the selective reporting of analyses and discrepancies in clinical trials: a systematic review of cohort studies of clinical trials. PLoS Med 2014; 11: e1001666. Williamson PR, Gamble C, Altman DG, Hutton JL. Outcome selection bias in meta-analysis. Stat Methods Med Res 2005; 14: 515–24. Ioannidis JP. Adverse events in randomized trials: neglected, restricted, distorted, and silenced. Arch Intern Med 2009; 169: 1737–39. Ioannidis JP, Evans SJ, Gøtzsche PC, et al. Better reporting of harms in randomized trials: an extension of the CONSORT statement. Ann Intern Med 2004; 141: 781–88. Sivendran S, Latif A, McBride RB, et al. Adverse event reporting in cancer clinical trial publications. J Clin Oncol 2014; 32: 83–89. Péron J, Maillet D, Gan HK, Chen EX, You B. Adherence to CONSORT adverse event reporting guidelines in randomized clinical trials evaluating systemic cancer therapy: a systematic review. J Clin Oncol 2013; 31: 3957–63. Maillet D, Blay JY, You B, Rachdi A, Gan HK, Péron J. The reporting of adverse events in oncology phase III trials: a comparison of the current status versus the expectations of the EORTC members. Ann Oncol 2015; 27: 192–98. Mhaskar RS, Reljic T, Wao H, Kumar A, Miladinovic B, Djulbegovic B. Treatment-related harms: what was planned and what was reported? National Cancer Institute’s Co-operative group phase III randomized controlled trials: a systematic review. J Clin Epidemiol 2014; 67: 354–56. Vera-Badillo FE, Shapiro R, Ocana A, Amir E, Tannock IF. Bias in reporting of end points of efficacy and toxicity in randomized, clinical trials for women with breast cancer. Ann Oncol 2013; 24: 1238–44. Pitrou I, Boutron I, Ahmad N, Ravaud P. Reporting of safety results in published reports of randomized controlled trials. Arch Intern Med 2009; 169: 1756–61. Scharf O, Colevas AD. Adverse event reporting in publications compared with sponsor database for cancer clinical trials. J Clin Oncol 2006; 24: 3933–38. Chan AW, Hróbjartsson A, Haahr MT, Gøtzsche PC, Altman DG. Empirical evidence for selective reporting of outcomes in randomized trials: comparison of protocols to published articles. JAMA 2004; 291: 2457–65. Booth CM, Le Maître A, Ding K, et al. Presentation of nonfinal results of randomized controlled trials at major oncology meetings. J Clin Oncol 2009; 27: 3938–44. Montori VM, Devereaux PJ, Adhikari NK, et al. Randomized trials stopped early for benefit: a systematic review. JAMA 2005; 294: 2203–09. Bassler D, Briel M, Montori VM, et al. Stopping randomized trials early for benefit and estimation of treatment effects: systematic review and meta-regression analysis. JAMA 2010; 303: 1180–87. Elimova E, Seruga B, Jiang H, Tannock IF. Comparison of results between the first and updated reports of phase III clinical trials (RCTs). J Clin Oncol 2014; 32: abstr 6520. Rothwell PM. External validity of randomised controlled trials: “to whom do the results of this trial apply?” Lancet 2005; 365: 82–93. Mitchell AP, Harrison MR, Walker MS, George DJ, Abernethy AP, Hirsch BR. Clinical trial participants with metastatic renal cell carcinoma differ from patients treated in real-world practice. J Oncol Pract 2015; 11: 491–97. Treweek S, Dryden R, McCowan C, Harrow A, Thompson AM. Do participants in adjuvant breast cancer trials reflect the breast cancer patient population? Eur J Cancer 2015; 51: 907–14. Heng DY, Choueiri TK, Rini BI, et al. Outcomes of patients with metastatic renal cell carcinoma that do not meet eligibility criteria for clinical trials. Ann Oncol 2014; 25: 149–54.

41

42

43

44

45

46

47

48

49

50

51

52

53 54

55 56 57

58 59

60

61 62

63

Templeton AJ, Vera-Badillo FE, Wang L, et al. Translating clinical trials to clinical practice: outcomes of men with metastatic castration resistant prostate cancer treated with docetaxel and prednisone in and out of clinical trials. Ann Oncol 2013; 24: 2972–27. Mol L, Koopman M, van Gils CW, Ottevanger PB, Punt CJ. Comparison of treatment outcome in metastatic colorectal cancer patients included in a clinical trial versus daily practice in The Netherlands. Acta Oncol 2013; 52: 950–55. Sorbye H, Pfeiffer P, Cavalli-Björkman N, et al. Clinical trial enrollment, patient characteristics, and survival differences in prospectively registered metastatic colorectal cancer patients. Cancer 2009; 115: 4679–87. AlHashem HY, Al-Mubarak M, Vera-Badillo FE, et al. Impact of geographic region on benefit of approved anticancer drugs evaluated in international phase III clinical trials. Clin Oncol (R Coll Radiol) 2015; published online Nov 2. DOI:10.1016/j.clon.2015.09.010. Bracarda S, Gernone A, Gasparro D, et al. Real-world cabazitaxel safety: the Italian early-access program in metastatic castration-resistant prostate cancer. Future Oncol 2014; 10: 975–83. Ascierto PA, Simeone E, Sileni VC, et al. Clinical experience with ipilimumab 3 mg/kg: real-world efficacy and safety data from an expanded access programme cohort. J Transl Med 2014; 12: 116. Berlin JA, Glasser SC, Ellenberg SS, et al. Adverse event detection in drug development: recommendations and obligations beyond phase 3. Am J Public Health 2008; 98: 1366–71. Institute of Medicine CoQoHCiA. Crossing the quality chasm: a new health system for the 21st century. Washington, DC: National Academy Press, 2001. Edwards BJ, Gounder M, McKoy JM, et al. Pharmacovigilance and reporting oversight in US FDA fast-track process: bisphosphonates and osteonecrosis of the jaw. Lancet Oncol 2008; 9: 1166–72. Seruga B, Sterling L, Wang L, Tannock IF. Reporting of serious adverse drug reactions of targeted anticancer agents in pivotal phase III clinical trials. J Clin Oncol 2011; 29: 174–85. FDA. FDA Approved Drug Products. http://www.accessdata.fda. gov/drugsatfda_docs/label/2006/125084s046LBL.pdf (accessed Dec 13, 2015). Psaty BM, Kronmal RA. Reporting mortality findings in trials of rofecoxib for Alzheimer disease or cognitive impairment: a case study based on documents from rofecoxib litigation. JAMA 2008; 299: 1813–17. Lurie P, Wolfe SM. Misleading data analyses in salmeterol (SMART) study. Lancet 2005; 366: 1261–62. Habibi R, Lexchin J. Quality and quantity of information in summary basis of decision documents issued by health Canada. PLoS One 2014; 9: e92038. Hazell L, Shakir SA. Under-reporting of adverse drug reactions: a systematic review. Drug Saf 2006; 29: 385–96. Carpenter D, Zucker EJ, Avorn J. Drug-review deadlines and safety problems. N Engl J Med 2008; 358: 1354–61. Hazell L, Cornelius V, Hannaford P, et al. How do patients contribute to signal detection? A retrospective analysis of spontaneous reporting of adverse drug reactions in the UK’s Yellow Card Scheme. Drug Saf 2013; 36: 199–206. Margraff F, Bertram D. Adverse drug reaction reporting by patients: an overview of fifty countries. Drug Saf 2014; 37: 409–19. Fu AZ, Graves KD, Jensen RE, Marshall JL, Formoso M, Potosky AL. Patient preference and decision-making for initiating metastatic colorectal cancer medical treatment. J Cancer Res Clin Oncol 2015; 142: 699–706. Pignatti F, Jonsson B, Blumenthal G, Justice R. Assessment of benefits and risks in development of targeted therapies for cancer—the view of regulatory authorities. Mol Oncol 2015; 9: 1034–41. Lexchin J. Why are there deadly drugs? BMC Med 2015; 13: 27. Shih YC, Halpern MT. Economic evaluations of medical care interventions for cancer patients: how, why, and what does it mean? CA Cancer J Clin 2008; 58: 231–44. Dueck AC, Mendoza TR, Mitchell SA, et al. Validity and reliability of the US National Cancer Institute’s Patient-Reported Outcomes Version of the Common Terminology Criteria for Adverse Events (PRO-CTCAE). JAMA Oncol 2015; 1: 1051–59.

www.thelancet.com/oncology Vol 17 May 2016

Policy Review

64

65

66

67

68

69

Basch E, Deal AM, Kris MG, et al. Symptom monitoring with patient-reported outcomes during routine cancer treatment: a randomized controlled trial. J Clin Oncol 2015; published online Dec 7. DOI:10.1200/JCO.2015.63.0830. Morgans AK, van Bommel AC, Stowell C, et al. Development of a standardized set of patient-centered outcomes for advanced prostate cancer: an international effort for a unified approach. Eur Urol 2015; 68: 891–98. Basch E. Systematic collection of patient-reported adverse drug reactions: a path to patient-centred pharmacovigilance. Drug Saf 2013; 36: 277–78. Ocana A, Ethier JL, Díez-González L, et al. Influence of companion diagnostics on efficacy and safety of targeted anti-cancer drugs: systematic review and meta-analyses. Oncotarget 2015; 637: 39538–49. Amir E, Seruga B, Martinez-Lopez J, et al. Oncogenic targets, magnitude of benefit, and market pricing of antineoplastic drugs. J Clin Oncol 2011; 29: 2543–49. Calvert M, Blazeby J, Revicki D, Moher D, Brundage M. Reporting quality of life in clinical trials: a CONSORT extension. Lancet 2011; 378: 1684–85.

www.thelancet.com/oncology Vol 17 May 2016

70

71

72

73 74

Booth CM, Tannock IF. Randomised controlled trials and population-based observational research: partners in the evolution of medical evidence. Br J Cancer 2014; 110: 551–55. Juurlink DN, Mamdani MM, Lee DS, et al. Rates of hyperkalemia after publication of the Randomized Aldactone Evaluation Study. N Engl J Med 2004; 351: 543–51. Booth CM, Shepherd FA, Peng Y, et al. Adoption of adjuvant chemotherapy for non-small-cell lung cancer: a population-based outcomes study. J Clin Oncol 2010; 28: 3472–78. Fosså SD, Gilbert E, Dores GM, et al. Noncancer causes of death in survivors of testicular cancer. J Natl Cancer Inst 2007; 99: 533–44. Ma P, Marinovic I, Karaca-Mandic P. Drug manufacturers’ delayed disclosure of serious and unexpected adverse events to the US Food and Drug Administration. JAMA Intern Med 2015; 175: 1565–66.

e219