Research Design Characteristics of Published Pharmacologic Randomized Clinical Trials for Irritable Bowel Syndrome and Chronic Pelvic Pain Conditions: An ACTTION Systematic Review.

Chronic pain conditions occurring in the lower abdomen and pelvis are prevalent, often challenging to manage, and can negatively affect health-related quality of life, including psychosocial and sexual well-being [ 3 , 8 , 25 , 27 , 32 , 37 ]. Very few U.S. Food and Drug Administration (FDA) or European Medicines Agency (EMA) approved treatments currently exist for these conditions. The dearth of treatments arises, in part, from methodological challenges in designing randomized clinical trials (RCTs) to evaluate the efficacy of novel treatments in these conditions. First, pain in the lower abdomen often co-occurs with other symptoms (e.g., frequent urination, diarrhea, constipation) that can affect pain severity [ 24 ]. Additionally, the condition may involve pain flares (i.e., periods with increased symptom severity) and pain may occur only during specific activities (e.g., sexual activity [ 27 , 29 ]). These characteristics make it challenging to design a single primary endpoint that is clinically meaningful to patients and responsive to treatment. Although it is possible to assess multiple outcomes in RCTs, in order to adequately evaluate the overall effect of a treatment in conditions with multiple symptoms, identifying a single primary endpoint or statistically adjusting for multiple primary analyses is necessary to minimize false positive conclusions in clinical trials. Unlike in other chronic pain conditions that have typically used the 0 – 10 numeric rating scale of pain intensity for the primary endpoint, RCTs investigating chronic pain conditions occurring in the lower abdomen and pelvis may require more appropriately targeted endpoints (e.g., a summary of pain intensity over multiple weeks to capture fluctuating symptoms or an endpoint that incorporates improvements in pain and another symptom). A second challenge of conducting RCTs in lower abdominal chronic pain conditions is defining the study entry criteria. In addition to established diagnostic criteria (e.g., [ 11 , 15 , 18 , 19 ]), study entry criteria must address how other potential causes of pain in the region (e.g., infection, cancer) will be excluded and whether medications that could affect the symptoms of interest will be permitted in the trial. Additionally, IBS and pelvic pain conditions often present concurrently or with other chronic pain conditions [ 4 , 7 , 17 , 23 , 24 , 33 ] which could impede participants’ ability to accurately report pain associated with the condition being studied [ 31 ]. Decisions regarding these entry criteria could affect the assay sensitivity (i.e., the ability of a RCT to detect a treatment effect if one truly exists), generalizability, and feasibility of the trial. To our knowledge, with the exception of 1 review published over 15 years ago on IBS trials [ 2 ], no systematic reviews have summarized the entry criteria and outcome measures used in RCTs for IBS and chronic pelvic pain conditions. Consequently, the current article presents a systematic review of RCTs of pharmacologic treatments for IBS and 3 common chronic pelvic pain conditions (i.e., chronic prostatitis / chronic pelvic pain syndrome (CP/CPPS), (interstitial cystitis / bladder pain syndrome (IC/BPS), and vulvodynia) [ 5 , 6 , 20 , 26 ] that can have a substantial adverse effect health-related quality of life [ 3 , 8 , 25 , 27 ]. This systematic review had 3 goals: (1) to summarize the entry criteria in these RCTs, (2) to summarize the primary endpoints in these RCTs, and (3) to evaluate the clarity of reporting of methodological details that are essential for interpretation of the results of RCTs. This summary of entry criteria, endpoints, and reporting clarity can be used to inform future research directed toward optimizing the design of RCTs for IBS and these chronic pelvic pain conditions and identify areas for improvement in publications of future trials to maximize the interpretation of results by clinicians, policy makers, and researchers.

Articles reporting RCTs investigating treatments for pain associated with irritable bowel syndrome (IBS), chronic prostatitis / chronic pelvic pain syndrome (CP/CPPS), (interstitial cystitis / bladder pain syndrome (IC/BPS), and vulvodynia, or for “chronic pelvic pain” published between January 1973 and April 2016 were identified in a PubMed search performed by a medical librarian (see Appendix 1 for search strategy). A second PubMed search was conducted to identify articles that studied treatments that are approved for use by the European Medicines Agency (EMA) or U.S. Food and Drug Administration (FDA) specifically for use in the 4 included conditions. We chose to examine multiple pain conditions that affect the abdominal and lower pelvis areas in a single review because they share common features that pose design challenges for defining entry criteria and outcome measures in clinical trials (i.e., the pain symptoms often co-occur with other symptoms or occur only during specific activities). We chose these conditions because they are common [ 5 , 6 , 20 , 26 ], have substantial adverse impacts on health-related quality of life (e.g., sexual dysfunction and psychological distress) [ 3 , 8 , 25 , 27 ], are representative of multiple organ systems (i.e., gastrointestinal, urological, gynecological) and include conditions specific for men and women. In order to cover each condition adequately in a single manuscript, we limited the review to 4 representative abdominal pain conditions. We examined only a single treatment class (i.e., pharmacologic) so that we could focus on a limited number of research design considerations and methodologic issues. The search hits were each screened for inclusion by 2 authors (KBI, JYC, or RAK) independently. Included articles (1) reported double-blind, placebo controlled RCTs that evaluated the efficacy of a treatment for 1 of the 4 specific conditions (i.e., IBS, CP/CPPS, IC/BPS, vulvodynia) or “chronic pelvic pain” not otherwise defined and (2) investigated pharmacologic interventions delivered orally, topically, intravenously, bolus injection, or bladder instillation. RCTs for surgical procedures and non-pharmacological treatments such as electrical stimulation, acupuncture, neuromodulation therapies, nutritional supplements and probiotics, psychotherapy, and physical therapy were excluded. Articles that investigated bacterial prostatitis were also excluded. Bacterial prostatitis was excluded from the study because we did not consider it to be a pain condition as it has a known pathological cause (i.e., recurrent infection) and is treated with antibiotics. Finally, articles that were identified by the search based on the term “chronic pelvic pain” but that included patients with conditions that were not examined in this review (e.g., endometriosis) were excluded. A coding manual ( Appendix 2 ) was developed to extract eligibility criteria, primary outcome measures, endpoints, data analyses, and adjustment for multiplicity in these trials. The manual was pre-tested and modified for content by JSG, KBI, and RAK using trials that did not meet the criteria for inclusion. This was performed in 3 rounds (9 trials total). The final list of the 126 identified articles (127 trials) was randomized and each trial was coded independently by 2 authors (from among JYC, JSG, KBI, RAK, SMS). The pre-testing process was terminated after the third round because no differing interpretations by coders or questions regarding the manual’s clarity occurred and no further improvements to the manual were necessary. The data were reviewed and discrepancies due to oversight were adjusted by KBI. Differences in interpretation were adjudicated by KBI in conjunction with JSG. For the purposes of this analysis we used the following definitions (1) primary outcome measure - any instrument used to measure the outcome of a trial that the authors identified as primary or that was the sole instrument used to assess efficacy in the trial (e.g., 0 – 10 NRS for pain intensity); (2) primary endpoint – how the primary outcome measure(s) was used (e.g., mean pain intensity on a daily 0 – 10 NRS at endpoint; 0 – 10 NRS used to define response in order to identify “treatment responders” who reported a 30% improvement in pain intensity from baseline to endpoint); (3) primary analysis – identification of the primary endpoint, the statistical test used to compare groups, and which groups were compared in the analysis considered primary. Descriptive statistics were used to characterize trial design features and clarity of reporting in the total sample by condition (i.e., chronic pelvic pain, CP/CPPS, IBS, IC/BPS, and vulvodynia) and by treatment type, that is treatments that have a potential analgesic mechanism (e.g., opioids, antidepressants, antiepileptics, NSAIDs) vs. those that target non-pain symptoms (e.g., constipation, diarrhea). For these analyses, the number of trials was used in the denominator (i.e., 127 for the entire sample).

A total of 4011 survey items were coded; 514 (13%) discrepancies occurred. Of those 511 discrepancies, 408 (80%) were due to oversight and 106 (21%) were due to differences in interpretation. Of the 750 records identified in our first PubMed search, 628 articles were excluded, leaving 122 articles that met the eligibility criteria ( Figure 1 ). Of the 99 publications found in the second set of PubMed searches, 67 articles were excluded, leaving 32 articles that met the eligibility criteria (1 of which reported the results from 2 RCTs that were coded separately). Of the 32 eligible articles from the second search, 28 were already included in the results of the first PubMed search ( Supplemental Figure 1 ), leaving a total of 126 articles that reported 127 RCTs in our final set (See Appendix 3 for list of included articles). The reasons for exclusion are outlined in Figure 1 and Supplemental Figure 1 . The articles were published between 1973 and 2016, with approximately two-thirds published between 2003 and 2016. The majority of trials evaluated treatments for IBS (n=86, 68%). Comparable numbers of trials investigated treatments for IC/BPS (n=19, 15%) and CP/CPPS (n=16, 13%), with the remaining trials including patients with “chronic pelvic pain” with no specific condition criteria (n=4, 3%) and vulvodynia (n=2, 2%). The most frequently examined treatments were anti-spasmodic agents (n=37, 29%), anti-constipation agents (n=12, 9%), anti-inflammatory medications (n=9, 7%), tricyclic anti-depressants (n=7, 6%), and botulinum toxin (n=7, 6%). Most experimental treatments were delivered orally (n=110, 87%). All but 6 trials used an inactive placebo in the control group (n=121, 95%). Almost two-thirds of the trials reported at least some industry sponsorship (n=80, 63%) ( Table 1 ). The most common inclusion criterion was a minimum pain duration (n=103, 81%) (Note: although 49 of these trials did not specify a minimum pain duration in the entry criteria, they based eligibility on the Rome Criteria for IBS that requires a minimum pain duration of 6 months). The next most common inclusion criteria were fulfilling an established diagnostic criteria (e.g., Rome Criteria for IBS [ 11 ], Friedrich’s criteria for vulvodynia [ 15 ], NIDDK IC/BPS criteria [ 18 ]; n=78, 61%) and reporting a minimum pain intensity (n=53, 42%). The presence of a comorbid condition associated with the abdominal/pelvic area (e.g. urinary tract infection, kidney stone, anal fissure, recent surgery) was the most common exclusion criterion (n=96, 76%). Eighty of these 96 (83%) trials reported some kind of diagnostic imaging (e.g., colonoscopy endoscopy), physical examination (e.g., pelvic or rectal exam), or laboratory testing (e.g., urinalysis or biopsy) to rule out these other conditions ( Table 2 ). Fifty-four trials (43%) excluded participants who had previously undergone abdominal surgery. Eighty-six trials (68%) prohibited the use of 1 or more concomitant drugs that could affect pain. The most frequently prohibited drugs were opioids (n=37, 43%), antispasmodics/anticholinergics (n=27, 31%), “treatments for the condition” or that could affect efficacy evaluation (without further explanation) (n=21, 24%), antidepressants (n=17, 20%), and anti-inflammatory drugs (n=13, 15%) ( Table 2 ). Common eligibility criteria by individual conditions and treatment type are illustrated in Supplemental Tables 1 and 2 . A single primary outcome measure was identified for seventy-eight (61%) trials. For 11, multiple primary outcome measures (9%) were identified. When considering all of the included trials, some of the most common primary outcome measures were applicable only to IBS trials because over half of the trials reviewed investigated IBS. The 2 types of primary outcome measures that were used most frequently were (1) a composite score including pain and non-pain symptoms, and (2) a single question evaluating disease-specific symptom relief (e.g., “IBS symptom relief”), (n=15, 17% for each). The next most common primary outcome measures were a single question evaluating IBS or abdominal pain and discomfort relief (n=12, 13%) and pain intensity (n=11, 12%). Seventy-two (57%) trials identified a primary endpoint. The primary endpoints used most frequently were: (1) response defined as adequate pain relief for a certain percentage of time (n=9, 13%), (2) response defined as adequate IBS symptom relief for a certain percentage of time (n=9, 13%), (3) severity or change from baseline in pain at endpoint (n=8, 11%), (4) severity or change from baseline in pain and non-pain composite at endpoint (n=7, 10%), and (5) response based on adequate symptom relief at endpoint (n=6, 8%). Pain intensity was more frequently used as the primary outcome measure in trials evaluating treatments from drug classes with known analgesic efficacy as compared to trials using other treatments (e.g., anti-constipation agents) (25% vs. 9%, respectively). Primary outcome measures and primary endpoints by pelvic pain condition and treatment type are presented in Supplemental Tables 3 and 4 . Items from diaries that were used by patients to record symptoms related to the condition (e.g., number of defecations or urinations per day or quality of defecation; n=83, 65%) and pain intensity (n=82, 65%) were the most common non-primary outcome measures reported for the included trials. Other non-primary outcome measures included composite measures that included pain and non-pain symptoms (n=43, 34%) ( Table 4 ). At least one primary analysis was identified for 55 trials (43%); 10 of these (18%) identified multiple primary analyses. Eight of these 10 trials (80%) adjusted for multiplicity, with a majority (63%) using a gatekeeping or hierarchical procedure to do so. The other 3 approaches employed were co-primary endpoints that both were required to be statistically significant for the trial to be considered a success (n=1, 13%), the Bonferroni method (n=1, 13%), and a combination of a gatekeeping method and Bonferroni method (n=1, 13%). The length of time that participants were treated with the experimental medication and the duration of follow-up after randomization was reported for all 127 trials. The median duration that participants were treated with experimental medication was 12 weeks (interquartile range [IQR], 4-12). The median number of weeks that participants were followed after randomization was 12 weeks (IQR, 8-16). The number of participants who were screened and randomized, completed the trial, and were included in the analyses is reported in Table 5 . Fifty-eight trials (46%) reported the reasons that patients were not enrolled in the trial (e.g., did not meet eligibility criteria, did not complete required questionnaires, refused to participate). Eight (6%) of the included trials reported the percentage of participants with at least 1 comorbid mental health or pain condition. In these 8 trials, the most frequently reported comorbid conditions were depression (n=5, 63%), anxiety (n=2, 25%), and endometriosis (n=2, 25%). Two of the 8 trials (25%) used a minimum score on a patient self-report or questionnaire or clinician-rated interview (e.g., minimum score on the Hamilton Rating Scale for Depression [ 21 ]) to identify a comorbid condition, while 4 of the trials (50%) did not report their method. One trial used a medical chart review and another used patient-identified comorbid conditions (e.g., patient reports that they are depressed or that a doctor previously diagnosed them as being depressed) to identify comorbid conditions.

As would be expected, the results of this systematic review identified variability in eligibility criteria and outcome measures among trials of IBS and 3 common chronic pelvic pain conditions. However, this variability was also observed within trials of the same conditions. Exclusion of co-morbid conditions that could cause related symptoms in the same part of the body and requirement of a minimum pain duration were included as entry criteria in at least three-quarters of the trials. However, fewer than half of the trials required a minimum pain severity at baseline. For many chronic pain conditions, a minimum baseline pain severity is one of the most common inclusion criteria used in RCTs in order to minimize the chance of a “floor effect” (i.e., inability to detect an improvement in mild symptoms) [ 12 ]. Interestingly, co-morbid pain conditions outside the abdominal and pelvic area (e.g., fibromyalgia, tension-type headache) were infrequently listed as an exclusion criterion and participants’ characteristics in regards to co-morbid pain or mental health conditions were reported for only 6% of trials. Since co-morbid pain and mental health conditions are common and can affect the experience and reporting of pain [ 1 , 31 ], research examining how inclusion or exclusion of these patients affects the assay sensitivity and generalizability of IBS and pelvic pain trials would be valuable to optimize clinical trial design. However, exclusion of certain conditions may be challenging when there are no well-accepted diagnostic criteria. Additionally, considering the overlap among many pelvic pain conditions exclusion of co-occurring conditions will inhibit recruitment and generalizability. It was also interesting that few trials (9%) prohibited hormone treatments (e.g., corticosteroids and androgens), given that these can affect pain in the lower abdominal and pelvis area [ 22 , 28 ]. The most common type of primary outcome measure (i.e., a composite outcome measure incorporating pain and non-pain symptoms) was reported in only 17% of trials, although this percentage was higher in trials of IC/BPS and CP/CPPS (i.e., 53% and 64%, respectively). The high variability in primary outcome measures in the included trials is not surprising given the complicated nature of the symptom presentation in these conditions. Similarly, we identified a considerable amount of variability in how the outcome measures were incorporated into trial endpoints. For example, in IBS trials, outcome measures were used to define symptom severity at the trial endpoint and also to define “responders” based on varying time frames of response (e.g., response at endpoint vs. over a certain percentage of time throughout the trial). The FDA and EMA guidances both recommend endpoints that identify responders based on improvement in both pain and bowel function over a certain percentage of time throughout the study [ 13 , 36 ]. Only 8% of IBS trials reported this type of endpoint; however, this is not surprising considering the guidances were published in 2012 and 2014 and only 6% of the IBS trials were published after 2014. Furthermore, the FDA guidance suggests that improvement in pain alone could be considered a primary endpoint in trials of IBS as long as other IBS symptoms are not worsened by the experimental treatment. Within the trials that identified a primary outcome measure, pain intensity was primary in only 25% of the IBS trials that evaluated treatments in drug classes with known analgesic efficacy. Use of a primary outcome measure that evaluates pain alone would be expected to improve assay sensitivity in trials designed to evaluate efficacy of an analgesic that lacks a mechanism that would be expected to improve other symptoms associated with the condition. Both the FDA and EMA guidances on IBS acknowledge that their recommended outcome measures for IBS have yet to be validated [ 13 , 36 ]. Regulatory guidances are not available for the other conditions included in this review. Given the regulatory guidance and the high level of variability in the primary endpoints that we identified in this review, it is clear that significant work remains to validate trial endpoints for specific chronic pelvic pain conditions chronic pelvic pain syndrome. Regardless of whether a consensus is available on the optimized primary outcome measures, endpoints, and analyses to use for trials in these conditions, it is important that RCT investigators pre-specify a single primary analysis or adjustment for multiplicity in order to decrease the chance of a false positive conclusion [ 16 , 34 ]. A primary analysis or adjustment for multiplicity in the primary analysis was identified for only approximately 40% of the trials reviewed. This result identifies a major area for improvement in the design and reporting of IBS and pelvic pain trials. The included trials utilized several strategies to address multiplicity, either by incorporating multiple symptoms into a single primary endpoint or using statistical strategies. These strategies included (1) a responder analysis that required a predetermined degree of improvement in multiple symptoms, (2) a composite outcome measure that incorporates assessment of multiple symptoms in a single composite score, (3) a hierarchical gatekeeping approach that pre-specifies the order of analysis of individual outcome measures in which the treatment effect of each subsequent symptom is only tested if the previous analysis yields a statistically significant result [ 9 , 10 ], and (4) a Bonferroni correction, which involves adjusting the significance level of each analysis so that the overall false positive rate is protected, typically to 5% [ 30 , 35 ]. All of these methods to evaluate multiple symptoms without inflating the false positive rate in a RCT provide information regarding the magnitude of change in each of the symptoms with the exception of the composite score. Changes in a composite score can be driven by a single symptom, and therefore it can be difficult to interpret the meaning of a change in composite score. If a composite score is used, sensitivity analyses that investigate the treatment effects on the individual components of the composite are recommended. This fact should be taken into consideration when designing new endpoints for IBS and chronic pelvic pain trials in order to maximize the interpretability of the potential clinical impact of treatments identified in RCTs. Several limitations of this review should be acknowledged. First, we only evaluated published articles reporting on RCTs in IBS, CP/CPPS, IC/BPS, vulvodynia and “chronic pelvic pain” and not those that specifically recruited patients with other chronic or acute abdominal and pelvic pain conditions (e.g., endometriosis; dysmenorrhea). Second, although important methodological details of RCTs should be reported in publications, it is possible that investigators did not report all eligibility criteria and primary outcome measures used in the trials and that our results may reflect, at least in part, inadequacies in reporting rather than deficiencies in clinical trial design. In conclusion, our review identified a lack of consensus in the literature on key design features of clinical trials for IBS, CP/CPPS, IC/BPS, vulvodynia, and chronic pelvic pain. Research is necessary to identify research design features and methods that can maximize assay sensitivity and clinical meaningfulness of endpoints in these trials (e.g., secondary analyses that compare the effect sizes associated with different endpoints in existing trial data; qualitative interviews with patients regarding the impact of different symptom patterns). Furthermore, more consistent reporting of design characteristics that are critical to the interpretation of the trial results and for meta-analyses of results across studies is important to improve the dissemination and incorporation of results into clinical practice and management guidelines.