Adverse event reporting in nonpharmacologic, noninterventional pain clinical trials: ACTTION systematic review

Ò

PAIN 155 (2014) 2253–2262

www.elsevier.com/locate/pain

Adverse event reporting in nonpharmacologic, noninterventional pain clinical trials: ACTTION systematic review Matthew Hunsinger a,⇑, Shannon M. Smith b, Daniel Rothstein b, Andrew McKeown b, Melissa Parkhurst c, Sharon Hertz d, Nathaniel P. Katz e, Allison H. Lin d, Michael P. McDermott f, Bob A. Rappaport d, Dennis C. Turk g, Robert H. Dworkin b a

School of Professional Psychology, Pacific University, Hillsboro, OR, USA Department of Anesthesiology, University of Rochester School of Medicine and Dentistry, Rochester, NY, USA Department of Psychiatry, University of Rochester School of Medicine and Dentistry, Rochester, NY, USA d United States Food and Drug Administration, Silver Spring, MD, USA e Analgesic Solutions, Natick, MA, and Tufts University, Boston, MA, USA f Department of Biostatistics and Computational Biology, University of Rochester School of Medicine and Dentistry, Rochester, NY, USA g Department of Anesthesiology and Pain Medicine, University of Washington, Seattle, WA, USA b c

Sponsorships or competing interests that may be relevant to content are disclosed at the end of this article.

a r t i c l e

i n f o

Article history: Received 16 April 2014 Received in revised form 30 July 2014 Accepted 6 August 2014

Keywords: Harms reporting Safety reporting Adverse events Complementary and alternative medical treatments

a b s t r a c t Assessment of treatment safety is 1 of the primary goals of clinical trials. Organizations and working groups have created reporting guidelines for adverse events (AEs). Previous research examining AE reporting for pharmacologic clinical trials of analgesics in major pain journals found many reporting inadequacies, suggesting that analgesic trials are not adhering to existing AE reporting guidelines. The present systematic review documented AE reporting in 3 main pain journals for nonpharmacologic, noninterventional (NP/NI) trials examining pain treatments. To broaden our pool of nonpharmacologic trials, we also included trials examining acupuncture, leech therapy, and noninvasive stimulation techniques (eg, transcutaneous electrical nerve stimulation). We documented AE reporting at 2 levels of specificity using coding manuals based on the Consolidated Standards of Reporting Trials (CONSORT) harms reporting standards and Analgesic, Anesthetic, and Addiction Clinical Trial Translations, Innovations, Opportunities, and Networks (ACTTION) AE reporting checklist. We identified a number of inadequacies in AE reporting across the 3 journals. For example, using the ACTTION coding manual, we found that less than one-half of the trials reported specific AE assessment methods; approximately one-third of the trials reported withdrawals due to AEs for each study arm; and about one-fourth of the trials reported all specific AEs. We also examined differences in AE reporting across several trial characteristics, finding that AE reporting was generally more detailed in trials with patients versus those using healthy volunteers undergoing experimentally evoked pain. These results suggest that investigators conducting and reporting NP/NI clinical trials are not adequately describing the assessment and occurrence of AEs. Ó 2014 International Association for the Study of Pain. Published by Elsevier B.V. All rights reserved.

1. Introduction One of the primary goals of a clinical trial is to assess treatment safety, which is needed alongside efficacy information for a trial to be clinically informative [8,12,14,15,17,18,20,26,31]. A major way to assess safety is through the reporting of adverse events (AEs) and serious adverse events (SAEs)—also referred to as harms—that ⇑ Corresponding author. Address: Pacific University, School of Professional Psychology, 190 SE 8th Ave. Hillsboro, OR 97123, USA. Tel.: +1 503 352 3612. E-mail address: matthewh@pacificu.edu (M. Hunsinger).

arise during trials. Regulatory agencies and clinicians rely on this information to understand the possible risks of a treatment decision. Incomplete AE reporting can lead to the underestimation of risk [2,3,9,11,19,20,27,28,32], potentially compromising regulatory approval and informed clinical use of treatments. Professional groups have created standards for AE reporting, resulting in the International Conference on Harmonization (ICH) [13] and Good Clinical Practice (GCP) [10] standards. GCP requires that AEs and SAEs be recorded during a trial and reported to the trial sponsor, all investigators involved, and any appropriate regulatory authorities. These standards do not apply to AE reporting in

http://dx.doi.org/10.1016/j.pain.2014.08.004 0304-3959/Ó 2014 International Association for the Study of Pain. Published by Elsevier B.V. All rights reserved.

2254

Ò

M. Hunsinger et al. / PAIN 155 (2014) 2253–2262

the medical literature, creating the potential for reporting omissions in publications. To ameliorate this potential, the Consolidated Standards of Reporting Trials (CONSORT) harms reporting standards were developed in 2004 to standardize AE reporting by creating a checklist of essential AE information for trial publication [16]. The adequacy of AE reporting in clinical trials for pain treatments has been examined for pharmacologic and interventional (ie, invasive) [7] treatments. Smith et al. [29,30] reviewed AE reporting practices across 3 pain journals and concluded that clinical trials of pharmacologic treatments did not adequately report AE information. Cornelius et al. [5] reviewed AE reporting for trials examining the analgesic efficacy of antidepressant and antiepileptic medications for neuropathic pain, also finding numerous reporting inadequacies. The results of Smith et al. [29,30] and Cornelius et al. [5] suggest that, despite the availability of reporting guidelines, AE reporting practices in published clinical trials of pharmacologic pain treatments do not provide adequate harms information. Inadequacies in harms reporting, however, are not limited to investigations of analgesic interventions [25]. There has been limited investigation of the quality of AE reporting for nonpharmacologic, noninterventional (NP/NI) pain treatments (eg, psychotherapy); that is, treatments that do not involve ingesting or absorbing a chemical or substance and do not involve penetrating the skin. In research on acupuncture, a treatment that involves skin penetration but is typically grouped with NP/NI treatments, Lu et al. [22] examined the degree to which trials examining the efficacy of acupuncture for cancer pain complied with the full CONSORT statement for harms reporting. They found that less than one-half of the trials examined met CONSORT guidelines for harms reporting, suggesting that AE reporting for clinical trials examining acupuncture needs improvement. The objective of the present systematic review was to assess the adequacy of AE reporting in clinical trials of a comprehensive group of NP/NI treatments in the same 3 major pain journals that Smith et al. [29,30] examined. We assessed the degree to which AE reporting practices were consistent with 2 guidelines for harms reporting (described below). In addition, we examined potential relationships between trial characteristics and AE-reporting practices. 2. Methods

pain as a primary or secondary outcome (see Appendix S1, available online, for all trial eligibility criteria). In a few trials examining psychotherapeutic interventions, blinding was not possible (eg., comparing a behavioral treatment to physical therapy); we included these trials to create an inclusive pool of NP/NI trials. Trials were included if they had at least 1 condition in which an NP/NI pain treatment was used without being combined with a pharmacologic and/or interventional (eg, nerve blocks, implantable pumps, spinal cord stimulators, and surgery) treatment. As defined above, we considered an NP/NI treatment to be any treatment that does not entail the ingestion or absorption of a chemical or substance orally or topically and does not involve penetrating the skin. This resulted in the exclusion not only of purely pharmacologic trials but also of trials that examined the analgesic efficacy of treatments that are not generally considered medications (eg, vitamins). These criteria were selected to capture a pool of articles examining analgesic treatments not yet investigated for AE reporting. To broaden our pool of nonpharmacologic treatments, we included trials that examined noninvasive stimulation techniques (eg, transcutaneous electrical nerve stimulation, transcranial magnetic stimulation), acupuncture, and leech therapy. We included trials examining acupuncture and leech therapy, even though these techniques entail penetration of the skin, because acupuncture and leech therapy are typically considered types of complementary and alternative medicine (CAM) treatment and our eligibility criteria captured many other CAM treatments. Clinical trial publications from 3 major pain journals (European Journal of Pain, Journal of Pain, and Pain) were selected from 2000 to 2003 (epoch 1, before published standards for harms reporting) and from January of 2006 to July of 2012 (epoch 2, after the publication of harms reporting standards). These journals have similar reporting guidelines for authors, including requiring adherence to CONSORT guidelines [16] (http://www.consort statement.org). Using the prespecified trial eligibility criteria listed in Appendix S1, several authors (M.H., D.R., M.P.) manually searched each issue of the 3 journals during the target years through the journal’s home page; every issue was searched by 2 authors. We supplemented our manual search with a search conducted by a University of Rochester medical librarian using PubMed, which resulted in the identification of 5 additional articles. Any articles that ambiguously met eligibility criteria were discussed by the authors to reach a final decision regarding inclusion in the final pool. We identified 140 articles that met eligibility criteria.

2.1. Study design 2.3. AE definition This systematic review had 4 main goals: 1) to assess whether AE reporting in NP/NI analgesic clinical trials improved after the introduction of the revised CONSORT standards in 2004; 2) to assess the comprehensiveness of AE reporting in NP/NI analgesic clinical trials using criteria proposed by CONSORT and the Analgesic, Anesthetic, and Addiction Clinical Trial Translations, Innovations, Opportunities, and Networks (ACTTION); 3) to evaluate the relationships between AE reporting and trial characteristics; and 4) to compare the adequacy of AE reporting in NP/NI analgesic clinical trials versus pharmacologic analgesic clinical trials. The information that we coded based on the CONSORT standards and ACTTION recommendations was not directly compared because these 2 sets of information were viewed as complementing each other. 2.2. Study selection Based on previous AE reporting research conducted by ACTTION [29,30], the present systematic review included trials with adults that were randomized, were single- or double-blind, and measured

We defined an AE as any adverse event, ranging from minor tolerability issues to serious safety concerns. This definition is consistent with the one set forth by the 2004 CONSORT standards for harms reporting [16]. We included questionnaires, hemodynamic measures, and laboratory tests as AE assessments, but did not consider quantitative sensory testing to be an AE assessment unless the authors specifically stated that this method was used to assess AEs. 2.4. Data extraction Among a pool of 3 authors (M.H., D.R., A.M.), 2 authors coded each article. To examine harms reporting before the CONSORT guidelines were updated, we examined articles published during epoch 1 (2000 to 2003). To examine harms reporting after the updated CONSORT standards were published and sufficient time had passed for them to be disseminated, we examined a subset of articles, published from 2008 to 2012, during epoch 2 (2006– 2012). We used the same coding manual used in our previous work

Ò

M. Hunsinger et al. / PAIN 155 (2014) 2253–2262

on harms reporting [30], which distilled the 10 CONSORT harms reporting guidelines covering the assessment, reporting, and discussion of AEs and SAEs into 10 simplified items. Coders practiced coding articles for harms reporting information with the CONSORT manual using 10 articles from the same journals published in 2005 to identify and resolve coding ambiguities (eg, cases in which coders coded the same information differently) and, once proficient using the coding manual, then coded 113 articles with each article coded twice. The remaining 27 articles from the total article pool were published from 2006 to 2007 (between epochs 1 and 2) and therefore were not included in the pool of articles coded using the CONSORT coding manual (see Appendix S2 to view the manual). In line with past studies, we also extracted more in-depth information about AE reporting using a separate coding manual developed based on ACTTION AE reporting recommendations [29]. The ACTTION coding manual captures additional information about AE assessment, occurrence (including AE severity, AE duration and resolution, and SAEs), and withdrawals. We coded articles using the ACTTION coding manual that were published during Epoch 2 (2006–2012) only. To identify and resolve coding ambiguities, the same coders (M.H., D.R., A.M.) practiced on the same 10 articles that we used for the CONSORT coding manual. Once coders became proficient using the ACTTION coding manual, 2 coders coded each article from a subset of the total pool of articles, 114 of the 140 articles, using the ACTTION coding manual. The remaining 26 articles from the total pool were published during epoch 1 (2000–2003) and therefore were not included in the pool of articles coded using the ACTTION coding manual (see Appendix S3 to view the manual). The final pool of articles included 88 articles that were coded using both manuals (Table 1). To examine potential relationships between trial characteristics and reporting practices, we also coded all 140 articles for the following information: 1) study design; 2) participant type (ie, healthy volunteers vs pain patients); 3) pain type; 4) number of blinded treatment groups; 5) treatment type; 6) treatment comparators; 7) treatment session number (ie, 1 treatment session vs multiple treatment sessions); 8) primary outcome; 9) number of randomized participants; and 10) sponsor. In addition, we recorded data on journal and issue year. Various kinds of analyses were conducted when examining the relationships between trial characteristics and AE reporting practices. Fulfillment of individual CONSORT items and AE reporting data based on the ACTTION coding manual were frequency data, requiring nonparametric analyses (ie, Fisher’s exact test); in contrast, CONSORT total scores were

Table 1 Number of trials per year, journal, and coding manual.

2000 2001 2002 2003 2006 2007 2008 2009 2010 2011 2012 Total

European Journal of Pain

Journal of Pain

Pain

Total

CONSORT

ACTTION

0 2 1 1 1 3 6 2 3 2 6 27

0 2 0 1 4 2 2 7 6 4 2 30

5 5 6 3 9 6 13 4 8 13 11 83

5 9 7 5 14 11 21 13 17 19 19 140

5 9 7 5 — — 21 12 16 19 19 113

— — — — 14 11 21 13 17 19 19 114

Abbreviations: ACTTION, Analgesic, Anesthetic, and Addiction Clinical Trial Translations, Innovations, Opportunities, and Networks; CONSORT, Consolidated Standards of Reporting Trials. Fisher’s exact tests used.

2255

ratio data, which allowed parametric analyses (ie, t test and regression). 2.5. Statistical analyses To examine the degree to which AE reporting in NP/NI trials was consistent with CONSORT standards, we analyzed the frequencies of particular responses for each CONSORT item as well as a CONSORT harms reporting total score, created by summing the number of items present for each trial without assigning a weight to any item [30]. Therefore, the higher the total score, the more CONSORT harms reporting items the trial fulfilled. For all inferential analyses with the CONSORT total score (with the exception of an analysis comparing AE reporting in this review to AE reporting in past research), we transformed the scores using a log transformation because the distribution of the original total scores was positively skewed. For descriptive statistics, we report medians and interquartile ranges for nontransformed data; however, we used transformed data for all inferential analyses (with the exception of the analysis mentioned above). We conducted several analyses to examine relationships between trial characteristics and fulfillment of CONSORT items. To examine whether AE reporting practices improved after the introduction of the revised CONSORT standards in 2004, Fisher’s exact tests were used to compare the percentage of trials that fulfilled each CONSORT item across the 2 epochs. To examine whether CONSORT items were fulfilled at different frequencies as a function of trial characteristics, we conducted 2-sample t tests to examine potential differences in mean transformed CONSORT total scores between groups defined by trial characteristics (ie, epoch, study design, participant type, pain type, number of double-blind treatment groups/periods, treatment session number, pain outcome, treatment type, and number of randomized participants/patients). In addition, we performed a multiple linear regression analysis on the transformed CONSORT total score with all of the trial characteristics listed above as independent variables. Finally, we used a Wilcoxon rank sum test to compare nontransformed CONSORT total scores in our data set to nontransformed CONSORT total scores in a data set used for past research examining AE reporting in trials examining pharmacologic analgesic treatments [30]. Unless otherwise specified, statistical tests were performed using a 5% significance level (2-tailed). We also examined the degree to which AE reporting in NP/NI trials was consistent with ACTTION recommendations. That is, we also examined the number of trials that reported specific kinds of AE information, in contrast to analyses with CONSORT items and total scores, which capture AE reporting at a less fine-grained level of detail. We calculated the frequency of trials fulfilling each item for all trials. We also combined responses pertaining to methods of AE assessment such that trials were coded as ‘‘all assessment methods reported’’ when the authors reported specific AE assessment methods; as ‘‘some assessment methods reported’’ when the authors reported at least 1 specific AE assessment method along with an ambiguous reference to an assessment method (eg, ‘‘safety was monitored by the nursing staff’’); and as ‘‘not reported’’ when the authors described no AE assessment procedures [29]. We coded SAE information the same way. We were not able to calculate a total score for data derived from the ACTTION coding manual and enter all trial characteristics into a single analysis, as we did with CONSORT total scores in the regression analysis. To examine potential trial characteristic group differences in the fulfillment of ACTTION recommendations, we focused on 2 key study factors, namely, differences between the 2 NP/NI approaches that occurred with the greatest frequency in the trials that we examined (ie, psychotherapy versus noninvasive stimulation techniques) and differences between participant types (pain patients vs

2256

Ò

M. Hunsinger et al. / PAIN 155 (2014) 2253–2262

healthy volunteers) using Fisher’s exact tests. For all Fisher’s exact tests examining differences between treatment types and participant types, we focused on the number of trials that provided all AE and SAE information for that category. 3. Results 3.1. Coder discrepancies Coders demonstrated a 93% level of agreement overall (199 discrepancies among 2938 items coded) when coding articles using the CONSORT guidelines. In all, 31% of the disagreements were due to oversights (61 oversights out of 199 discrepancies) and 69% were due to differing interpretation (138 differing interpretations out of 199 discrepancies). When we used ACTTION recommendations to code trials, coders had a 93% level of agreement overall (603 discrepancies among 8208 items coded). Approximately 14% of the disagreements were due to oversights (87 oversights among 603 discrepancies), and 86% were due to differing interpretation (516 differing interpretations among 603 discrepancies). Discrepancies for both sets of recommendations were resolved through discussion among authors. 3.2. Trial characteristics Trial characteristics for the articles coded with the CONSORT and ACTTION manuals are presented in Table 2 (frequencies and percentages across trial characteristics for data derived from both coding manuals) and Table S2 (medians, interquartile ranges, means, and standard deviations across trial characteristics for CONSORT total scores), available online. We coded 2 overlapping data sets that were very similar to each other (88 of the trials were in both data sets; Table 1) with the CONSORT (n = 113) and ACTTION (n = 114) manuals and did not compare these data sets to each other. Although our trial pool included trials for every year and journal that we examined, there were more trials from later years, and most of the trials were published in Pain. The majority of the trials used a parallel groups research design (CONSORT: 88%; ACTTION: 90%), enrolled pain patients versus healthy volunteers participating in experimentally evoked pain (CONSORT, 77%; ACTION, 78%), enrolled patients with chronic pain (CONSORT, 62%; ACTTION, 64%), had 2 treatment groups or periods (CONSORT, 62%; ACTTION, 61%), compared the experimental intervention to another NP/NI treatment (CONSORT, 51%; ACTTION, 55%), administered the intervention across multiple sessions (CONSORT, 66%; ACTTION, 66%), and examined pain as the primary outcome (CONSORT, 89%; ACTTION, 90%). The treatments that were studied the most frequently were acupuncture (CONSORT, 14%; ACTTION: 14%), psychotherapy (CONSORT, 22%; ACTTION, 18%), and noninvasive stimulation techniques (CONSORT, 27%; ACTTION, 25%). 3.3. Adherence to CONSORT harms reporting standards 3.3.1. CONSORT harms reporting items Only 1 CONSORT harms reporting item was fulfilled in more than 50% of the articles (ie, ‘‘Describe for each arm the participant withdrawals that are due to harms and their experiences with the allocated treatment’’; Table S1). There was 1 item that was not fulfilled in any of the trials (ie, ‘‘Describe any subgroup analyses and exploratory analyses for harms’’) and 1 item that was fulfilled in only 1 article (ie, ‘‘Describe plans for presenting and analyzing information on harms (including coding, handling of recurrent events, specification of timing issues, handling of continuous measures, and any statistical analyses)’’). Although the frequencies of fulfillment of some items were slightly higher in epoch 2 than in

epoch 1, these numerical differences were not statistically significant (Table S1). 3.3.2. CONSORT harms reporting total scores across trial characteristics We conducted 2 sets of analyses to examine whether mean transformed CONSORT harms reporting total scores differed according to trial characteristics. First, we examined these potential differences using 2-sample t tests (Table S2). These tests revealed that mean transformed CONSORT harms reporting total scores differed significantly across the levels of 2 trial characteristics, namely, outcome type and treatment type. Trials using pain and analgesic sparing as co-primary outcomes (median = 3.0; interquartile range [IQR] = 2.0–3.5 for untransformed scores) had higher mean transformed CONSORT harms reporting total scores compared to trials that used pain as the only primary outcome (median = 1.0; IQR = 0.0–3.0, P = .008). We also found that mean transformed CONSORT harms reporting total scores were higher for trials examining acupuncture (median = 3.0; IQR = 2.25–4.0) and noninvasive stimulation techniques (median = 1.0; IQR = 1.0– 4.0) than for trials examining psychotherapy (median = 1.0; IQR = 0.0–1.0), P < .05. The multiple regression analysis including all trial characteristics revealed that the only significant differences were among treatment types. Using the Holm technique for adjusting for multiple comparisons, the regression revealed that the differences found in the t tests between acupuncture, noninvasive stimulation techniques, and psychotherapy remained significant. 3.3.3. CONSORT harms reporting for NP/NI versus pharmacologic treatments NP/NI trials of pain treatments had significantly lower CONSORT harms reporting total scores (median = 1.0; IQR = 1.0–3.0) than those for pharmacologic trial data reported by Smith et al. [29] (median = 7.0; IQR = 5.0–8.0), P < .001. This pattern was found even when comparing Smith et al.’s [29] CONSORT harms reporting total scores for pharmacologic trials to those for only acupuncture trials in the NP/NI sample (median = 3.0; IQR = 2.25–4.0, P < .001). 3.4. Consistency with ACTTION AE reporting recommendations 3.4.1. AE assessment methods Less than one-half of the trials reported specific methods for AE assessment (n = 46; 40%) and 2 trials (2%) collected AE data without specifying the method (Table 3). The AE assessment methods reported most often were spontaneous patient/participant report (n = 10; 21%), patient/participant responses to specific AE prompts (n = 9; 19%), and other miscellaneous methods (n = 6; 13%); only 1 trial (2%) used a physical examination and 1 (2%) used vital signs, laboratory procedures, diagnostic tests, or a radiological examination. Only 19 trials (40%) reported the timing of AE data collection. No trials reported using the Coding Symbols for a Thesaurus of Adverse Reaction Terms (COSTART), Medical Dictionary for Regulatory Activities (MedDRA), or World Health Organization (WHO) Adverse Drug Reaction Terminology to code AE data; 3 trials (6.2%) reported the use of other systems. No trials described plans to present descriptive information based on their AE assessment. 3.4.2. Reporting of participant withdrawals The sample of published trials more thoroughly reported the frequency of withdrawals compared to AE reporting (Table S3). More than one-half of the trials (n = 77; 68%) reported withdrawals for each treatment group. These trials were also more thorough in reporting the reasons for withdrawals (compared to AE reporting), with reporting of the specific reasons for withdrawals in 65 trials (65%). In 49 trials (49%), the initiator of all study withdrawals

Ò

2257

M. Hunsinger et al. / PAIN 155 (2014) 2253–2262 Table 2 Trial characteristics.

Table 2 (continued) Characteristic

Characteristic

CONSORT No. of trials (n = 113)

ACTTION No. of trials (n = 114)

Study design Parallel Crossover

99 (87.6%) 14 (12.4%)

102 (89.5%) 12 (10.5%)

Participant type Healthy volunteers Pain patients

26 (23.0%) 87 (77.0%)

24 (21.1%) 89 (78.1%)

Pain type Acute Chronic Recurrent pain Experimentally evoked pain Other

12 (10.6%) 70 (61.9%) 2 (1.8%) 25 (22.1%) 4 (3.6%)

10 (8.8%) 73 (64%) 3 (2.6%) 25 (21.9%) 3 (2.7%)

No. of double-blind treatment groups/periods 2 70 (61.9%) 3 28 (24.8%) 4 11 (9.7%) 5 1 (0.9%) 6 1 (0.9%) 8 2 (1.8%)

69 (60.5%) 30 (26.3%) 13 (11.4%) 1 (0.9%) 0 (0%) 1 (0.9%)

Experimental treatment Acupuncture Attention-regulation techniques Audiovisual techniques Cupping therapy Educational Exercise Leech therapy Massage Mixed Other Psychotherapy Physical therapy/occupational therapy/ physiotherapy Qi Gong Spiritual healing Simulation techniques Yoga

16 (14.2%) 8 (7.1%) 3 (2.7%) 1 (0.9%) 5 (4.4%) 3 (2.7%) 1 (0.9%) 2 (1.8%) 12 (10.6%) 2 (1.8%) 25 (22.1%) 1 (0.9%)

16 (14%) 8 (7%) 2 (1.8%) 1 (0.9%) 5 (4.4%) 5 (4.4%) 1 (0.9%) 2 (1.8%) 14 (12.3%) 5 (4.4%) 21 (18.4%) 1 (0.9%)

1 (0.9%) 1 (0.9%) 30 (26.5%) 2 (1.8%)

1 (0.9%) 0 (0%) 28 (24.6%) 2 (1.8%)

58 (51.3%)

63 (55.3%)

45 (39.8%)

42 (36.8%)

20 (17.7%) 10 (8.8%) 8 (7.1%)

21 (18.4%) 12 (10.5%) 8 (7%)

9 (8.0%)

0 (0%)

3 (2.7%) 2 (1.8%)

0 (0%) 8 (7%)

No. of treatment sessions Single treatment session Multiple treatment sessions

38 (33.6%) 75 (66.4%)

39 (34.2%) 75 (65.8%)

Primary or co-primary outcomes Pain Pain and opioid/analgesic sparing

100 (88.5%) 13 (11.5%)

103 (90.4%) 11 (9.6%)

18–1256 77.0; 42.00– 120.50

17–9990 83.0; 47.5– 156.50

2 (1.8%) 18 (15.9%) 27 (12.7%) 31 (27.4%)

4 (3.5%) 17 (14.9%) 33 (28.9%) 26 (22.8%)

Comparator Other experimental nonpharmacologic/ noninterventional treatment Control nonpharmacologic/ noninterventional condition Standard of care Wait-list control Active interventional or noninterventional treatment (ie, analgesic delivered orally or intravenously) Control condition without experimental treatment (eg, control condition without virtual reality glasses) Control condition procedure (eg, surgery) Mixed treatments

No. of subjects randomized Range Median; interquartile range Sponsor Industry Institution/university Government agency Professional organization/foundation

Multiple sources Not reported

CONSORT No. of trials (n = 113)

ACTTION No. of trials (n = 114)

15 (13.3%) 19 (16.8%)

12 (10.5%) 22 (19.3%)

Abbreviations: ACTTION, Analgesic, Anesthetic, and Addiction Clinical Trial Translations, Innovations, Opportunities, and Networks; CONSORT, Consolidated Standards of Reporting Trials.

was clear; in 11 trials (11%), it was clear for only some of the withdrawals reported. Relative to the reporting of all withdrawals, less information was provided regarding withdrawals because of AEs (Table 4). Only 39 (34%) of the trials reported AE-related withdrawals separately for each treatment group. Moreover, only 38 trials (54%) reported all specific AEs associated with withdrawals. When the types of AEs associated with withdrawals were reported, 39 trials (56%) reported them for each treatment group. Almost all of the trials (n = 113; 99%) did not report AE withdrawals by severity, and only 10 trials (14%) reported withdrawals because of SAEs. When reporting the AEs for participants who withdrew because of AEs, only 36 trials (51%) reported all of the AEs that caused withdrawal; 1 trial reported some of the AEs (1%); and 2 trials (3%) reported no AEs. 3.4.3. Reporting of specific AEs and SAEs Reporting of specific AEs and SAEs was generally inadequate (Table 5 and Table S4). Only 26 trials (23%) reported specific AEs. When reporting specific AEs, only 23 trials (56%) reported the number or percentage of participants experiencing specific AEs in each treatment group; 1 trial (2%) reported the mean severity of each AE per group; 2 trials (5%) listed the AEs across all groups. One trial (2%) identified specific AEs as treatment emergent and 14 trials (34%) identified specific AEs as treatment related. However, none of the trials specified the methods that investigators used to identify AEs as treatment emergent or treatment related. Moreover, no trials distinguished expected from unexpected for specific AEs. When reporting information about AE severity, only 5 trials (12%) did so for all specific AEs. Only 8 trials (19%) reported the number or percentage of mild, moderate, or severe AEs. About the same number of trials reported the number or percentage of participants experiencing moderate or severe AEs, with 7 trials (17%) reporting this information for each group. Only 1 trial (1%) reported any SAE information, reporting the number or percentage of participants experiencing specific SAEs for each treatment group (Table S4). 3.4.4. ACTTION recommendations and trial characteristics (treatment type and research participant type) In many cases, trials examining noninvasive stimulation techniques provided more detailed AE information than trials examining psychotherapy. Compared to trials examining psychotherapy, trials examining noninvasive stimulation techniques more often reported specific AE assessment methods (53% vs 14%), reported withdrawals separately for each arm (85.7% vs 50%), and reported specific AEs (21% vs 10%). Descriptive and inferential information is provided in Tables 3–5 and in Tables S3 and S4. When comparing AE reporting across research participant type, we found that trials using patients generally reported more detailed AE information than those using healthy volunteers. Compared to trials using healthy volunteers, trials using patients more often reported specific AE assessment methods (47% vs 17%), reported withdrawals separately for each treatment group (78%

2258

Ò

M. Hunsinger et al. / PAIN 155 (2014) 2253–2262

Table 3 Trials reporting adverse event (AE) methodology. Treatment type

Research participant type

All trials (n = 114)

Acupuncture (n = 16)

Psychotherapy (n = 21)

Noninvasive stimulation techniques (n = 28)

Patients (n = 90)

46 (40%) 2 (2%) 66 (58%)

12 (75%) 0 (0%) 4 (25%)

P = .001a 3a (14%) 0 (0%) 18 (86%)

16a (57%) 0 (0%) 12 (43%)

P = .001b 42b (47%) 1 (1%) 47 (52%)

4b (17%) 1 (4%) 19 (79%)

10 (21%)

1 (8%)

0 (0%)

0 (0%)

10 (23%)

0 (0%)

9 (19%)

1 (8%)

1 (33%)

4 (25%)

8 (19%)

1 (20%)

1 (2%) 1 (2%)

0 (0%) 0 (0%)

0 (0%) 0 (0%)

1 (6%) 0 (0%)

1 (2%) 1 (2%)

0 (0%) 0 (0%)

6 (13%) 29 (60%)

2 (17%) 9 (75%)

0 (0%) 2 (66%)

1 (6%) 9 (56%)

6 (14%) 28 (65%)

0 (0%) 1 (20%)

Specify timing of AE data collection

19 (40%)

5 (42%)

0 (0%)

8 (50%)

18 (42%)

1 (20%)

Code AE data COSTART, MedDRA, or WHO-ART Another standardized system Researcher-defined terms No/not specified Describe scale used to rate AE severity

0 (0%) 3 (6%) 0 (0%) 45 (94%) 2 (4.1%)

0 (0%) 0 (0%) 0 (0%) 12 (100%) 0 (0%)

0 1 0 2 0

0 (0%) 2 (13%) 0 (0%) 14 (88%) 1 (6%)

0 (0%) 3 (7%) 0 (0%) 40 (93%) 2 (5%)

0 0 0 5 0

0 (0%) 2 (4%)

0 (0%) 1 (8%)

P = 1.0a 0a (0%) 1 (33%)

0a (0%) 0 (0%)

P = 1.0b 0b (0%) 2 (5%)

0b (0%) 0 (0%)

1 (2%) 45 (94%)

1 (8%) 10 (83%)

0 (0%) 2 (66%)

0 (0%) 16 (100%)

1 (2%) 40 (93%)

0 (0%) 5 (100%)

Report AE assessment methods Yes (reported specific methods) No (AEs collected, method not specified) No AE information reported AE assessment methods (trials can use more than 1 method) Participants’ spontaneous report or responses to open-ended questioning Participants’ responses to specific AE prompts or a scale of drug-associated AEs Physical examination Vital signs, laboratory procedures, diagnostic tests, radiological examination Other Not specified

Statistical plans presented for AE data Descriptive statistics to be presented Statistical comparisons between arms to be presented Other No statistical plans presented

(0%) (33%) (0%) (66%) (0%)

Healthy volunteers (n = 24)

(0%) (0%) (0%) (100%) (0%)

Abbreviations: COSTART, Coding Symbols for a Thesaurus of Adverse Reaction Terms; MedDRA, Medical Dictionary for Regulatory Activities; WHO-ART, World Health Organization Adverse Drug Reaction Terminology. Notes: Denominators for proportions in every section after ‘‘Report AE Assessment Method’’ and except for the ‘‘No AE information reported’’ rows are the sum of the first and second rows of the ‘‘Report AE Assessment Method’’ section. Superscript letters and italics denote the pair of proportions that were compared using Fisher’s exact tests.

vs 29%), reported the reasons for withdrawals (88% vs 88%), reported the number or percentage of withdrawals due to AEs separately for each group (67% vs 22%), reported all of the specific AEs causing withdrawals (65% vs 22%), and reported specific AEs (26% vs 13%). Descriptive and inferential information is listed in Tables 3–5 and in Tables 3 and 4. 4. Discussion Randomized clinical trials can best inform clinicians, patients, regulators, and payers about treatment safety only if comprehensive information about AE assessment and reporting is provided [8,12,14,15,17,18,20,26,31]. Past research on AE reporting practices in trials of pharmacologic treatments published in pain journals identified many reporting inadequacies [5,29,30]. The implication of these findings is that a considerable number of trials in the pain literature do not report sufficient information about AEs for readers to assess fully the safety of analgesic treatments. Given the limited focus of prior research on AE reporting for NP/NI and related nonmainstream pain treatments [22], we conducted a systematic review of harms reporting practices for NP/NI treatments in 3 major pain journals. We found many of the same kinds of reporting inadequacies that Smith et al. [29,30] identified for pharmacologic trials. Across the entire sample of NP/NI trials, approximately one-half did not mention the collection of AE information. Even when trials did report collecting AE data, reporting on the details of assessment was poor. One-fourth of the trials did not specify the methods of AE assessment, and only 17% reported the timing of AE assessment. In addition, the trials generally did a poor

job of reporting specific AEs; only 23% of the trials reported all specific AEs. Reporting was also poor for SAEs; all but 1 of the trials provided no information on the occurrence of SAEs. Our results indicate that investigators conducting and reporting NP/NI clinical trials are not describing the assessment and occurrence of AEs adequately. Not doing so creates ambiguity for the reader as to whether these events occurred without being reported or did not occur. The trials that we examined generally reported participant withdrawals more thoroughly than AEs; 68% reported the number of withdrawals in each group, and of these trials, 65% reported the reasons for withdrawals. Interestingly, the adequacy of reporting withdrawals overall was better than for reporting withdrawals due to AEs—only 34% of the trials reported the number of AE withdrawals in each treatment group. It is important to emphasize that all withdrawals should be comprehensively described. Given the results of this review and past research on AE reporting [29,30], the ubiquity of AE reporting inadequacies in published pain clinical trials points to a strong need to improve reporting practices as well as adherence to and consistency with established reporting recommendations. 4.1. Associations between AE reporting and time, trial characteristics, and treatment type We found no reporting differences across time for individual CONSORT harms reporting items and the CONSORT harms reporting total score; however, it is important to keep in mind that it is unclear whether this finding resulted from equivalent reporting over time or from the small number of trials in the first epoch.

Ò

2259

M. Hunsinger et al. / PAIN 155 (2014) 2253–2262 Table 4 Trials reporting participant withdrawal due to adverse events (AEs). Treatment type

Research participant type

All trials (n = 114)

Acupuncture (n = 16)

Psychotherapy (n = 21)

Noninvasive stimulation techniques (n = 28)

Patients (n = 90)

39 (34%) 0 (0%) 0 (0%) 31 (27%) 27 (24%)

4 (25%) 0 (0%) 0 (0%) 5 (31%) 5 (31%)

P = .41a 7a (33%) 0 (0%) 0 (0%) 2 (10%) 9 (43%)

8a (29%) 0 (0%) 0 (0%) 16 (57%) 1 (4%)

P = .001b 35b (39%) 0 (0%) 0 (0%) 17 (19%) 26 (29%)

4b (17%) 0 (0%) 0 (0%) 14 (58%) 1 (4%)

38 (54%) 1 (1%) 3 (4%)

4 (44%) 0 (0%) 0 (0%)

P = .007c 7c (77.7%) 0 (0%) 1 (11%)

8c (33%) 0 (0%) 0 (0%)

P < .001d 34d (65%) 1 (2%) 3 (6%)

4d (22%) 0 (0%) 0 (0%)

Selected AEs causing withdrawal Not specified

1 (1%)

0 (0%)

0 (0%)

0 (0%)

1 (1%)

0 (0%)

How report types of AEs causing withdrawal Separately for each arm Separately for a selection of arms Globally for all arms

39 (56%) 0 (0%) 0 (0%)

4 (44%) 0 (0%) 0 (0%)

7 (78%) 0 (0%) 0 (0%)

8 (33%) 0 (0%) 0 (0%)

35 (67%) 0 (0%) 0 (0%)

4 (22%) 0 (0%) 0 (0%)

Report no. or % of withdrawals due to AEs Separately for each arm Separately for a selection of arms Globally for all arms No withdrawals due to AEs occurred Withdrawals due to AEs not reported Report specific AEs causing withdrawal Yes, for all AE withdrawals Yes, for selected AE withdrawals Specific AEs causing withdrawal not reported (not specified)

Healthy volunteers (n = 24)

Report AE withdrawals by severity

1 (1%)

0 (0%)

0 (0%)

1 (4%)

0 (0%)

1 (6%)

Report AE withdrawals due to SAEs

10 (14%)

0 (0%)

1 (11%)

2 (8%)

10 (19%)

0 (0%)

For each patient/participant withdrawing due to AEs, were all AEs causing withdrawal reported Yes, for all AE withdrawals Yes, for some AE withdrawals No

36 (51%) 1 (1%) 2 (3%)

3 (33%) 0 (0%) 1 (11%)

7 (78%) 0 (0%) 0 (0%)

8 (33%) 0 (0%) 0 (0%)

32 (62%) 1 (2%) 2 (4%)

4 (22%) 0 (0%) 0 (0%)

Note: Denominators for proportions in every section after ‘‘Report no. or % of withdrawals due to AEs’’ and except for the ‘‘No withdrawals due to AEs occurred’’ and ‘‘Withdrawals due to AEs not reported’’ rows are the sum of the first 4 rows of the ‘‘Report no. or % of withdrawals due to AEs’’ section. Superscript letters and italics denote the pair of proportions that were compared using Fisher’s exact tests.

An absence of change in reporting practices would suggest that the dissemination of the CONSORT harms reporting standards did not have an impact on AE reporting for NP/NI trials in the pain journals that we examined. When we examined CONSORT harms reporting total scores as a function of the other trial characteristics, there were no significant differences after controlling for all other trial characteristics (except for treatment type, which we discuss below), which suggests that reporting inadequacies were present across a wide range of trials versus being present in trials with particular characteristics. Using the ACTTION coding manual, we also found that AE reporting was generally more detailed for trials enrolling patients than for trials enrolling healthy volunteers. This may be explained by a greater proportion of these trials having pharmacologic control conditions, a type of treatment for which assessing and reporting AEs is more common. This finding, however, was not reflected in the CONSORT harms reporting total scores. This is likely due to differences in the specificity of our coding for the CONSORT versus ACTTION recommendations. The ACTTION recommendations focus on detailed information about AE reporting and consequently may be better able to detect differences in reporting across variables of interest in contrast to the more general CONSORT harms reporting guidelines. Investigating AE reporting practices for NP/NI treatments is important for at least 2 reasons. First, it is possible that investigators publishing research on NP/NI treatments perceive them as having fewer risks than pharmacologic treatments (eg, investigators may believe that NP/NI treatments are intrinsically safer than pharmacological and interventional treatments) and are therefore less motivated to record and report AE information comprehensively. NP/NI treatments are heterogeneous and include many qualitatively different treatment approaches (eg, psychotherapy, hydrotherapy). Individuals investigating these treatments likely differ from those investigating pharmacological treatments in their

perceptions of the risks of these treatments, perhaps believing that the treatment has no risks. If no risk is perceived, an investigator may decide not to collect or report AE information. Consistent with the possibility that there are differences in perceptions of risk, we found that AE reporting for NP/NI pain treatments was poorer than that for pharmacologic pain treatments published in the same journals for an overlapping span of years [30]. All treatments have the potential for associated risks, even if NP/NI treatments tend to be less risky than pharmacologic and interventional treatments. Thorough AE reporting is, therefore, vital for any clinical trial of a pain treatment. A lack of AE recording and reporting creates ambiguity regarding whether there are risks associated with a particular NP/NI treatment, whereas recording AEs and reporting the absence of such events would support the conclusion that a particular NP/ NI treatment has few or no risks. Second, the regulatory controls that exist for determining and disseminating the risks of pharmacologic pain treatments do not exist for many NP/NI treatments. Because many of the treatment approaches that fall into this category are not under the purview of regulatory agencies, adequate AE reporting in peer-reviewed journal articles is especially important for identifying and disseminating risks associated with NP/NI pain treatments. We also found differences in AE reporting across the 3 treatments occurring most frequently in our sample of trials—acupuncture, psychotherapy, and noninvasive stimulation techniques— with more detailed reporting associated with acupuncture and noninvasive stimulation techniques compared to reporting associated with psychotherapy. These differences in reporting remained even when we controlled for the other trial characteristics that we examined. Our methodology did not allow us to assess the reasons why there were reporting differences across these 3 treatments. Just as with the difference in AE reporting between pharmacologic and NP/NI treatments, it is possible that AE reporting differences

Ò

2260

M. Hunsinger et al. / PAIN 155 (2014) 2253–2262

Table 5 Trials reporting specific adverse events (AEs). Treatment type

Report specific AEs Yes, for all specific AEs Yes, for selected specific AEs No AEs occurred Specific AEs not reported Selected specific AEs Not specified How report specific AEs No. or% of participants experiencing each event per arm No. or % of participants experiencing each event across arms Mean severity of each event per arm Mean severity of each event across arms List of events per arm List of events across arms

Research participant type

All trials (n = 114)

Acupuncture (n = 16)

Psychotherapy (n = 21)

Noninvasive stimulation techniques (n = 28)

Patients (n = 90)

26 (23%) 2 (2%) 13 (11%) 73 (64%)

8 1 2 5

P = .02a 2a (10%) 0 (0%) 1 (5%) 18 (86%)

6a (21%) 1 (4%) 7 (25%) 14 (50%)

P = .03b 23b (26%) 2 (2%) 11 (12%) 54 (59%)

3b (13%) 0 (0%) 2 (8%) 19 (79%)

2 (5%)

1 (9%)

0 (0%)

0 (0%)

2 (6%)

0 (0%)

23 (56%)

7 (64%)

2 (67%)

5 (36%)

20 (56%)

3 (60%)

2 (5%)

1 (9%)

0 (0%)

1 (7%)

2 (6%)

0 (0%)

1 0 0 2

0 0 0 1

0 0 0 0

1 0 0 0

1 0 0 2

0 0 0 0

(2%) (0%) (0%) (5%)

(50%) (6%) (13%) (31%)

(0%) (0%) (0%) (9%)

(0%) (0%) (0%) (0%)

(7%) (0%) (0%) (0%)

(3%) (0%) (0%) (6%)

Healthy volunteers (n = 24)

(0%) (0%) (0%) (0%)

Statistically compare specific AEs between arms Yes No

4 (13%) 24 (59%)

1 (9%) 8 (73%)

P = .67c 0c (0%) 2 (67%)

2c (14%) 5 (36%)

P = .58d 4d (11%) 21 (58%)

0d (0%) 3 (60%)

AE denominators per arm reported or clearly inferable Yes No

28 (68%) 0 (0%)

9 (82%) 0 (0%)

P = .12e 2e (67%) 0 (0%)

7e (50%) 0 (0%)

P = .64f 25f (69%) 0 (0%)

3f (60%) 0 (0%)

Report method(s) to classify specific AEs as treatment emergent Yes, method specified Yes, method not specified AEs not coded as treatment emergent

0 (0%) 1 (2%) 27 (66%)

0 (0%) 1 (55%) 8 (73%)

0 (0%) 0 (0%) 2 (67%)

0 (0%) 0 (0%) 7 (50%)

0 (0%) 1 (3%) 24 (67%)

0 (0%) 0 (0%) 3 (60%)

0 (0%) 14 (34%) 14 (34%)

0 (0%) 6 (55%) 3 (19%)

0 (0%) 1 (33%) 1 (33%)

0 (0%) 3 (21%) 4 (29%)

0 (0%) 14 (39%) 11 (12%)

0 (0%) 0 (0%) 3 (60%)

Report method(s) to classify specific AEs as treatmentrelated adverse reactions Yes, method specified Yes, method not specified AEs not coded as treatment-related adverse reactions Specific AE severity

P = .82g

P = .52h

Report any information about specific AE severity (mild, moderate, severe) Yes, for all specific AEs Yes, for selected specific AEs Yes, a general statement about overall AEs

5 (12%) 1 (2%) 2 (5%)

2 (18%) 0 (0%) 1 (9%)

0g (0%) 0 (0%) 0 (0%)

1g (7%) 1 (7%) 1 (7%)

5h (14%) 1 (3%) 2 (6%)

0h (0%) 0 (0%) 0 (0%)

Selected specific AEs reported by severity Severe or moderate and severe

1 (2%)

0 (0%)

0 (0%)

1 (7%)

0 (0%)

1 (20%)

Report no. or % of mild, moderate, severe AEs Separately for each arm Separately for a selection of arms Globally for all arms No. or% of mild, moderate, severe AEs not reported

7 0 1 0

(17%) (0%) (2%) (0%)

2 0 0 0

(18%) (0%) (0%) (0%)

0 0 0 0

(0%) (0%) (0%) (0%)

1 0 1 0

(7%) (0%) (7%) (0%)

7 0 1 0

(19%) (0%) (3%) (0%)

0 0 0 0

(0%) (0%) (0%) (0%)

7 0 0 1 0

(17%) (0%) (0%) (2%) (0%)

0 0 0 0 0

(0%) (0%) (0%) (0%) (0%)

0 0 0 0 0

(0%) (0%) (0%) (0%) (0%)

1 0 0 0 0

(7%) (0%) (0%) (0%) (0%)

7 0 0 1 0

(19%) (0%) (0%) (3%) (0%)

0 0 0 0 0

(0%) (0%) (0%) (0%) (0%)

Report no. or % of participants experiencing severe or moderate and severe AEs Separately for each arm Separately for a selection of arms Globally for all arms None occurred No. or% of participants experiencing severe, moderate, and severe AEs not reported Statistically compare AE severity between arms

1 (2%)

0 (0%)

0 (0%)

1 (7%)

1 (3%)

0 (0%)

Note: Denominators for proportions in every section after ‘‘Report Specific AEs’’ and except for the ‘‘No AEs occurred’’ and ‘‘Specific AEs not reported’’ rows are the sum of the first 3 rows of the ‘‘Report Specific AEs’’ section. Superscript letters and italics denote the pair of proportions that were compared using Fisher’s exact tests.

within NP/NI treatments are also due to differences in perceptions of risk. This possibility is consistent with criticism of the lack of attention paid to the risks associated with psychotherapy [1,4,6,21,24]. This criticism appears to be warranted, given the evidence of deterioration in clients undergoing psychotherapeutic treatment for pain [23].

4.2. Study limitations There are several limitations of our systematic review. One is that we examined only what was described in the publications. When a publication describing a clinical trial did not report AE information, it is possible that these events were assessed but were

Ò

M. Hunsinger et al. / PAIN 155 (2014) 2253–2262

not reported because they were deemed to be minor or did not occur. Therefore, we cannot make definitive statements about investigator failure to assess AE information. Nevertheless, our methodology allows us to evaluate directly the comprehensiveness of the information to which readers have access. We believe that it is the responsibility of researchers to explicitly report that AEs did not occur versus expecting the reader to assume that no explicit statement regarding AEs equates with their absence in the trial. A second limitation of our review involves whether our results can be generalized to trials investigating all NP/NI pain treatments. We examined AE reporting practices for the NP/NI treatments investigated in 3 pain journals, which is not necessarily an exhaustive list of all NP/NI pain treatments that have been investigated. Future research will need to investigate AE reporting for NP/NI pain treatments beyond those that we examined. However, given the AE reporting inadequacies found by Smith et al. [29,30], Cornelius et al. [5], and the present systematic review, it is likely that AE reporting in other journals and for other NP/NI treatment modalities will also be inadequate. 4.3. Conclusions The results of our systematic review suggest that, just as with the pharmacologic clinical trials, AE reporting needs to improve in published trials of pain treatments that are neither pharmacologic nor interventional. Given the availability of AE reporting recommendations, we strongly encourage researchers investigating NP/NI pain treatments to keep these standards in mind when designing clinical trials and to follow them when reporting trial outcomes. Doing so will be critical in improving AE reporting in the pain literature and, ultimately, informing all parties interested in pain treatment and enhancing patient safety. We contend that every pain treatment has potential risks associated with it, and safety cannot be assumed; therefore, comprehensive and systematic assessment and reporting of AEs is essential in all clinical trials. Although the overall risks of NP/NI treatments may be less than the overall risks of pharmacologic and interventional treatments, treatments such as acupuncture and psychotherapy can have serious risks, for example, infection and increased suicidal ideation, respectively. Conflict of interest statement The authors declare that there are no conflicts of interest with regard to this work. Acknowledgements The authors thank Michele Shipley, MLS (University of Rochester Medical Center, Miner Library) for help with finding articles for coding. Financial support for this project was provided by the ACTTION public–private partnership with the FDA, which has received research grants or other revenue from the FDA, various pharmaceutical companies, and other sources. The views expressed in this article are those of the authors and no official endorsement by the FDA or the pharmaceutical companies that have provided unrestricted grants to support the activities of ACTTION should be inferred. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.pain.2014.08.004.

2261

References [1] Barlow DH. Negative effects from psychological treatments: a perspective. Am Psychol 2010;65:13–20. [2] Basch E. The missing voice of patients in drug-safety reporting. N Engl J Med 2010;362:865–9. [3] Bent S, Padula A, Avins AL. Brief communication: better ways to question patients about adverse events. Ann Intern Med 2006;144:257–61. [4] Castonguay LG, Boswell JF, Constantino MJ, Goldfried MR, Hill CE. Training implications of harmful effects of psychological treatments. Am Psychol 2010;65:34–49. [5] Cornelius VR, Sauzet O, Williams JE, Ayis S, Farquhar-Smith P, Ross JR, Branford RA, Peacock JL. Adverse event reporting in randomized controlled trials of neuropathic pain: considerations for future practice. PAINÒ 2013;154:213–20. [6] Dimidjian S, Hollon SD. How would we know if psychotherapy were harmful? Am Psychol 2010;65:21–33. [7] Dworkin RH, O’Connor AB, Kent J, Mackey SC, Raja SN, Stacey BR, Levy RM, Backonja M, Baron R, Harke H, Loeser JD, Treede R, Turk DC, Wells CD. Interventional management of neuropathic pain: NeuPSIG recommendations. PAINÒ 2013;154:2249–61. [8] Dworkin RH, Turk DC, Farrar JT, Haythornwaite JA, Jensen MP, Katz NP, Kerns RD, Stucki G, Allen RR, Bellamy N, Carr DB, Chandler J, Cowan P, Dionne R, Galer BS, Hertz S, Jadad AR, Kramer LD, Manning DC, Martin S, McCormick CG, McDermott MP, McGrath P, Quessy S, Rappaport BA, Robbins W, Robinson JP, Rothman M, Royal MA, Simon L, Stauffer JW, Stein W, Tollett J, Wernicke J, Witter J. Core outcome measures for chronic pain clinical trials: IMMPACT recommendations. PAINÒ 2005;113:9–19. [9] Edwards JE, McQuay HJ, Moore A. Reporting of adverse effects in clinical trials should be improved: lessons from acute postoperative pain. J Pain Symptom Manage 1999;18:427–37. [10] Good Clinical Practice. Available at: [accessed November 15, 2013]. [11] Hammad TA, Pinheiro SP, Neyarapally GA. Secondary use of randomized controlled trials to evaluate drug safety: a review of methodological considerations. Clin Trials 2011;8:559–70. [12] Hauser W, Bartram C, Bartram-Wunn E, Tolle T. Adverse events attributable to nocebo in randomized controlled drug trials in fibromyalgia syndrome and painful diabetic peripheral neuropathy: systematic review. Clin J Pain 2012;28:437–51. [13] International Conference on Harmonization of Technical Requirements for Registration of Pharmaceuticals for Human Use, ICH Harmonized Tripartite Guideline. Structure and content of clinical study reports E3, Available at: ; 1995. [accessed 1.03.2014]. [14] Ioannidis JPA. Adverse events in randomized trials: neglected, restricted, distorted, and silenced. Arch Intern Med 2009;169:1737–9. [15] Ioannidis JPA, Contopoulos-Ioannidis DG. Reporting of safety data from randomized trials. Lancet 1998;352:1752–3. [16] Ioannidis JPA, Evans SJ, Gotzsche PC, O’Neill RT, Altman DG, Schulz K, Moher D. Better reporting of harms in randomized trials: an extension of the CONSORT statement. Ann Intern Med 2004;141:781–8. [17] Ioannidis JPA, Lau J. Improving safety reporting from randomized trials. Drug Saf 2002;25:77–84. [18] Ioannidis JPA, Lau J. Completeness of safety reporting in randomized trials: an evaluation of 7 medical areas. JAMA 2001;285:437–43. [19] Ioannidis JPA, Mulrow CD, Goodman SN. Adverse effects: the more you search, the more you find. Ann Intern Med 2006;144:298–300. [20] Katz NP. The measurement of symptoms of side effects in clinical trials of chronic pain. Contemp Clin Trials 2012;33:903–11. [21] Liliendeld SO. Psychological treatments that cause harm. Perspect Psychol Sci 2007;2:53–70. [22] Lu L, Liao M, Zeng J, He J. Quality of reporting and its correlates among randomized controlled trials on acupuncture for cancer pain: application of the CONSORT. Expert Rev Anticancer Ther 2013;13:489–98. [23] Morley S, Williams A, Eccleston C. Examining the evidence about psychological treatments for chronic pain: time for a paradigm shift? PAINÒ 2013;154:1929–31. [24] Morley S, Williams A, Hussain S. Estimating the clinical effectiveness of cognitive behavioural therapy in the clinic: evaluation of a CBT informed pain management programme. PAINÒ 2008;137:670–80. [25] Papanikolaou PN, Christidi GD, Ioannidis JPA. Comparison of evidence on harms of medical interventions in randomized and nonrandomized studies. Can Med Assoc J 2006;174:635–41. [26] Rief W, Barsky AJ, Glombiewski JA, Nestoriuc Y, Glaesmer H, Braehler E. Assessing general side effects in clinical trials: reference data from the general population. Pharmacoepidemiol Drug Saf 2011;20:405–15. [27] Reif W, Avorn J, Barsky AJ. Medication-attributed adverse effects in placebo groups. Arch Intern Med 2006;199:155–60. [28] Sheftell FD, Feleppa M, Tepper SJ, Rapoport AM, Ciannella L, Bigal ME. Assessment of adverse events associated with triptans—methods of assessment influence the results. Headache 2004;44:978–82. [29] Smith SM, Wang AT, Katz NP, McDermott MP, Burke LB, Coplan P, Gilron I, Hertz SH, Lin AH, Rappaport BA, Rowbotham MC, Sampaio C, Sweeney M, Turk DC, Dworkin RH. Adverse event assessment, analysis, and reporting in recent

2262

Ò

M. Hunsinger et al. / PAIN 155 (2014) 2253–2262

published analgesic clinical trials: ACTTION systematic review and recommendations. PAINÒ 2013;154:997–1008. [30] Smith SM, Chang RD, Pereira A, Shah N, Gilron I, Katz NP, Lin AH, McDermott MP, Rappaport BA, Rowbotham MC, Sampaio C, Turk DC, Dworkin RH. Adherence to CONSORT harms-reporting recommendations in publications of recent analgesic clinical trials: an ACTTION systematic review. PAINÒ 2012;153:2415–21. [31] Turk DC, Dworkin RH, Allen RR, Bellamy N, Brandenburg N, Carr DB, Cleeland C, Dionne R, Farrar JT, Galer BS, Hewitt DJ, Jadad A, Katz NP, Kramer LD,

Manning DC, McCormic CG, McDermott M, McGrath P, Quessy S, Rappaport BA, Robinson JP, Royal MA, Simon L, Stauffer JW, Stein W, Tollett J, Witter J. Core outcome domains for chronic pain clinical trials: IMMPACT recommendations. PAINÒ 2003;106:337–45. [32] Weingart SN, Gandhi TK, Seger AC, Seger DL, Borus J, Burdick E, Leape LL, Bates DW. Patient-reported medication symptoms in primary care. Arch Intern Med 2005;165:234–40.