By definition (Evers et al., 2018), placebo responses refer to the (averaged) symptom response of a group of patients receiving a placebo during a randomized, placebo-controlled trial, e.g., of drugs. In contrast, the placebo effects refer to the individual change in symptoms that can be attributed to a known mechanism eliciting this change. Therefore, the placebo response in RCT includes factors that may not be driven by placebo effects but by the natural course of the disease or by methodological biases, e.g., repeated measures eliciting the “regression-to-the-mean” effect. The placebo effect in individual patients (as well as in many patients) is thought to be due to two major mechanisms: expectations that treatment will result in improved or disappearing symptoms because this is the reason for the treatment; learning, based on previous experience with the same or different symptoms, treatments, or medical settings. In experimental medicine, both factors can be separated to investigate their contribution, whereas in clinical medicine, they are usually may be compounded but their relative contribution may vary between individual patients. In theory, however, learning and expectations are not completely independent drivers of the placebo response. And finally, the nocebo responses or nocebo effects are thought to be governed by similar mechanisms (Elsenbruch and Enck, 2015; Enck and Klosterhalfen, 2019; Enck and Klosterhalfen, 2020)

Table 1 summarizes the pooled placebo response rates in CD and UC according to the meta-analyses performed and based on the clinical conditions and the respective endpoints of the studies.

As can be seen both across the meta-analyses and within some of them [e.g., (Gallahan et al., 2010)], the placebo response rates vary between 5 and 50% in CD and between 10 and 35% in UC, with respect to clinical improvement and remission. As expected, remission rates are lower than improvement rates, and the higher the standard of criteria (clinical vs. endoscopic), the lower the placebo response. Remission maintenance rates can also vary between 15 and 50%, following similar rules. Postsurgery placebo remission and recurrence rates are as high as 50%, while placebo response rates for fistula closures are lower, below 20%. Only one study has focused on pouchitis, but the reported 24% placebo clinical improvement response is well within the range of the effects in CD and UC.

Some of the variances between studies may be due to different sensitivities and specificities of the various IBD activity scores used or differences in the cut-off or response criteria, e.g., CDAI-based definition of remission. How much of these placebo responses can be attributed to a true placebo effect or to the natural course of the disease cannot be judged from these meta-analyses. We will discuss this further below.

While there is no consistent factor that has been reported in all or most meta-analyses, a few factors are listed more than once. They can be separated into disease, design, and other factors.

Disease factors: As with many other and noninflammatory diseases, lower disease activity at the beginning of the study was associated with higher placebo responses (Su et al., 2004; Jairath et al., 2016; Jairath et al., 2017); the same mechanism may be seen when “prior surgery, concomitant small bowel and colonic disease, fistulizing phenotype, or prior immunomodulator therapy” (Pascua et al., 2008), “prior exposure to TNF antagonists and increased concentrations of CRP at baseline” (Duijvestein et al., 2020), disease duration, and “ being non-naïve to anti-TNF’s"” (Macaluso et al., 2019) were found to be associated with lower placebo responses, while—not surprising—higher “placebo remission was associated with concomitant corticosteroids” (Ma et al., 2018; Macaluso et al., 2019).

Design factors: The study duration in CD (Su et al., 2004; Gallahan et al., 2010; Ford et al., 2014; Jairath et al., 2017) and UC (Garud et al., 2008; Jairath et al., 2016) is among the factors found to affect the placebo response, but data are less consistent in UC. Another factor associated with higher placebo responses is the number of study visits during the trials (Ilnyckyj et al., 1997; Su et al., 2004; Jairath et al., 2016) (Figure 2).

A factor that has frequently been discussed in areas outside IBD, e.g., in depression and other neurological and psychiatric diseases [e.g., (Enck, 2016)], was explicitly found not to be of relevance in IBD, i.e., the increase of placebo response rates in RCTs over time, with more recent studies exhibiting higher rates (Gallahan et al., 2010; Ma et al., 2018). The number of study centers was found to be effective as well, with more centers being associated with lower placebo responses (Jairath et al., 2017). One meta-analysis of UC trials found the country to determine the placebo response, with higher values in European countries (pooled improvement rate 36.4%) than those in the United States (24.9%) (Garud et al., 2008).

Other factors: It is not surprising that studies providing a new and expensive class of drugs such as biologics produce higher placebo response rates (Jairath et al., 2017). They also found the route of administration to be predictive of the placebo response—this is in line with data from the general literature: the more invasive procedures produce higher responses and injections are more effective than pills.

Increased placebo response can also be expected when instead of biomarkers and clinical disease signs, subjective, patient-reported outcomes (PRO) are used to assess the clinical efficacy of an intervention—this was shown in a meta-analysis of 16 RCTs in IBD (11 UC and 15 CD) in which IBS-specific (IBDQL) and generic (SF36) health-related quality of life was measured as the primary outcome (Estevinho et al., 2018): while overall drug-over-placebo benefit was maintained, the placebo response rates were high with 41% for clinical improvement and 31% for remission. This effect is as well established in the literature, but for many diseases, appropriate biomarkers are unfortunately missing. However, as has been shown recently that for IBD, fecal calprotectin may serve as such an appropriate biomarker (Bertani et al., 2020), possibly able to reduce (or limit) placebo responses, specifically if coupled with PRO. The important role of disease biomarkers is also evident in Figure 2 (above), where histological and endoscopic findings produce overall lower placebo response rates than clinical assessments, irrespective of the number of study visits.

To distinguish and parcel out true placebo effects from the placebo response as a summary of different mechanisms in RCT (Figure 1), it would be necessary to conduct a three-arm trial in which an equal fraction of patients would receive the active drug, the placebo, and no treatment, and this should be fully randomized and blinded—which of course is impossible for the “no-treatment” control, and ethically questionable as well. One way around this problem has been to install “waiting list controls” where patients are kept on a waiting list for a specific period of time before receiving treatment or being randomized into the RCT. In this case, the trial would be applied in clinical conditions of minor severity such as depression, nausea, and functional bowel disorders or with therapies other than drugs, e.g., with psychotherapy. Meta-analyses have shown that about 50% of the placebo response in RCT may be attributed to the spontaneous course of the disease—whether this holds true for IBD is an open question. When placebo arms of RCTs were used to estimate the natural course of IBD patients in remission (Meyers and Janowitz, 1989), about 50% of patients remained in remission for at least 1 year. Whether this allows concluding that a “true” placebo effect in IBS should be beyond 50% in remission trials remains to be known.

However, there should be better ways to test this hypothesis, at least in maintenance trials, e.g., with the cohort multiple randomized controlled trial (cmRCT) (Relton et al., 2010) or the so-called Zelen design (Zelen, 1979). Such a trial would follow a stacked approach with double recruitment: first for an observation-only study, a large cohort of IBD patients in remission are recruited who are not taking any remission drug; subsequently, a subset of them is recruited for an interventional, placebo-controlled trial of a maintenance trial. This, to the best of our knowledge, has never been tried in IBD (Figure 3).

The ethical limitations of such a trial are at hand: leaving a group of patients with UC in remission untreated for some time can only be accomplished when narrow monitoring is warranted (Colombel et al., 2017), with patients who have minor symptoms during remission, and who, without the interference of the treating physician, would abstain from medication as long as possible anyway. Once remission is terminated, all patients need to receive appropriate standard treatment as soon as possible.

During the discussion of the placebo responses in IBD, as above, it occurred to us that while the placebo response may be somewhat higher in IBS (across around 100 RCTs approximately 40% (Ford and Moayyedi, 2010)) than in IBD, the moderators and mediators were remarkably similar. As we have shown in recent systematic reviews (Enck and Klosterhalfen, 2020), these factors may be grouped in both conditions into disease factors (severity, duration, and pre- and concomitant treatments), design factors (duration of treatment, number of visits, number of centers, and number of study arms), individual conditions (age, sex, race, and proxies), and environmental factors (setting, therapists, healthcare plans, nationality, etc.). Clearly, many have been found to be in effect also in IBD, but whether they can be ruled out as relevant or irrelevant also depends on the focus these factors given in RCT and whether such determinants are part of the study reporting and publishing routine. These arguments have also been summarized by other studies (Sands, 2009). Probably the most neglected area in IBD research is the influence of the treating physician and his/her empathy and management skills in the therapy procedure—but that has probably surfaced in the IBS world only because of the lack of effective drug therapies.

To the best of our knowledge, only one study (Schmid et al., 2015) has compared central processing of placebo analgesia using an established experimental model in both patients with IBS and IBD (UC), compared to healthy control volunteers, and found that while IBS patients lacked downregulation of brain activity in relevant areas of the pain matrix during placebo application, no difference was observed in UC patients and healthy controls in terms of their ability of proper pain control, indicating normal placebo responses in UC.

To the best of our knowledge, there is only one meta-analysis (Ma et al., 2019) that has focused on AE reporting in the placebo arm of RCT, much in contrast to many such analyses in other areas of medicine, especially in the pain literature. This meta-analysis included 124 CD and 71 UC RCTs, so most of the trials that have also been included are the above placebo regression studies. The authors reported a pooled AE reporting of 70.6% in CD and 54.4% in UC and noted no differences in comparison to the respective active arms of the trials for AE reporting, severe AE reporting, and withdrawal due to AE, but a lower risk of symptom worsening in the active arms for both diseases. This is to some degree surprising given that overall AE reports are high compared to other reports on anti-inflammatory drug classes. This may support the notion that the high global symptom burden and the chronic nature of the IBD let the patients expect and tolerate substantially more AE anyway before leaving RCT.

Nocebo responses, the second aspect in IBD trials, are of a different nature, i.e., the report of increased AE when patients are switched from an established though expensive therapy with biologics to the more reasonably priced therapy with “biosimilars.” PubMed counts (as of Dec 2020) more than 50 publications in which biosimilar usage is linked to the term nocebo, despite the fact that direct (blinded) comparison of biologics with biosimilars usually does not result in higher AE and SAE reporting with biosimilars [e.g., (Boone et al., 2018)]. The nocebo reports in daily clinical routine, however, appears to be driven by media reports, self-aid groups, and Internet-based “fake” news. One may even wonder whether solely the selection of the term “biosimilar”—suggesting similarity but not equality—may have forced or supported this discussion.

However, since this phenomenon dominates medical practice but not medical research, e.g., in RCT, little is known about its mediators and moderators. It is, furthermore, not unique to biologics/biosimilars in IBD, but became visible also in other conditions, e.g., for statin use in blood pressure regulation (Pedro-Botet et al., 2019), lactose usage in individuals claiming lactose intolerance (Vernia et al., 2010), and wheat product consumption in nonceliac gluten hypersensitivity (Biesiekierski et al., 2013) and, in general, when patients are switched from a branded to a generic drug (MacKrill et al., 2019).

However, the pitfalls of this discussion are evident: with any drug taken, separating between an AE and a nocebo response is impossible in the individual patient, even if RCT provides some evidence for an AE being a nocebo response. Furthermore, nocebo responses are not personal characteristics of patients—anxiety may be a sign, but it is not proof of it. And as with the placebo response, the healthcare provider probably contributes as much or even more to a nocebo response as does the patient. In the early days of the biosimilar development, a survey among 1,200 physicians, nearly half of the physicians (and 43.8 in gastroenterology) questioned the equivalence and efficacy of biosimilars across many medical subspecialties (Cohen et al., 2017)—and this may reflect patients concerns and AE reports.