Effectiveness of Placebo Interventions for Patients With Nonspecific Low Back Pain

A Systematic Review and Meta-analysis

Rob H.W. Strijkers; Marco Schreijenberg; Heike Gerger; Bart W. Koes; Alessandro Chiarotto


Pain. 2021;162(12):2792-2804. 

In This Article


This systematic review shows the magnitude of the placebo effect in patients with LBP. The number of studies (n = 2) focusing on (sub)acute LBP was too small for conducting meta-analyses and to draw conclusions for this subgroup of patients. In patients with chronic LBP, 19 studies were identified. Meta-analyses showed that placebo interventions are more effective than no intervention for pain intensity at short-term follow-up (SMD −0.37; moderate-quality evidence), but this effect, which corresponds to about 8 point on a 0 to 100 scale, may not be considered as clinically relevant. No significant effect was found on pain intensity at medium-term follow-up (SMD −0.26; low-quality evidence). A statistically significant positive effect on physical functioning at short-term follow-up (SMD −0.19; moderate-quality evidence) was found for placebo interventions as compared with no intervention control, but also this effect may not be clinically relevant. Placebo interventions were more effective than no intervention on short-term HRQoL physical health (MD 2.71 on a 0–100 scale; high-quality evidence) but not more effective for short-term HRQoL mental health (MD −0.49 on a 0–100 scale; high-quality evidence). No long-term follow-up data were available on any outcome.

Results in Perspective of the Available Literature

Multiple nonpharmacological interventions for chronic LBP have previously been studied and recommended, among these individual education, exercise therapy, and multidisciplinary rehabilitation.[49] Hayden et al. recently published an individual patient data meta-analysis on exercise therapy for chronic LBP showing a 10.7-point improvement in pain intensity scores at short-term follow-up in comparison with no intervention or usual care, which met our predefined 20% between-group difference for clinical relevance. Regarding physical functioning, a short-term clinically relevant improvement was also reported for exercise therapy.[25] Only short-term effects on pain were found in our analysis. This can in part be explained by the inability to blind the no intervention group. Literature demonstrates there is an effect of expectations on treatment outcome in patients with LBP, and this effect is not expected to last in the medium-term to long-term follow-up.[23]

Another recent systematic review showed that spinal manipulative therapy (SMT) did not improve pain intensity at short-term follow-up (1 month) compared with sham SMT in patients with chronic LBP, but disability improved only slightly at one month compared with sham SMT.[53]

Nonsteroidal anti-inflammatory drugs (NSAIDs) are currently the most recommended medications for patients with chronic LBP.[49] Interestingly, in a Cochrane review on NSAIDs, authors found an 8-point improvement on a pain scale of 0 to 100 compared with placebo.[15] Therefore, the magnitude of improvement on pain and functioning of NSAIDs vs placebo intervention is similar to the magnitude of effect for placebo interventions vs no interventions found in this review, both being not clinically relevant effects. However, this is not a direct comparison because the comparator groups are different. It would be more meaningful to compare the placebo effect found in this review with the effect of NSAIDs as compared to no intervention; this indirect comparison would be possible using network meta-analysis but it was outside the scope of this review. This systematic review further contributes to the literature discussion on the magnitude of placebo effects. In 2010, the latest version of the Cochrane review of Hrobjartsson et al. concluded, on basis of 156 trials, that there is no powerful placebo effect, only a potential small effect on subjective outcomes (eg, pain). Contrary to the review of Hrobjartsson, a recent narrative review on the placebo and nocebo effects of Colloca et al. concluded that placebo effects are "powerful and pervasive," based on the results of laboratory studies.[10] Such powerful effects were not found in our meta-analyses that focused on placebo effects on subjective outcomes in RCTs in the field of nonspecific LBP. We argue that laboratory studies, although representing a good platform to modulate placebo effects, may not have a sufficiently controlled setting to determine the magnitude of these effects, which was the focus of this systematic review. In fact, laboratory research is not usually used to assess the effectiveness of health interventions.[1] In this review, a larger effect size was found for open-label placebo RCTs as compared to 3-arm RCTs not including an open-label placebo treatment. A possible explanation for the discrepancy in effect size might be that the researchers in the open-label studies are focused on purposively delivering an enhanced placebo intervention, increasing its effectiveness. Given all these observations, we conclude there is a short-term placebo effect on pain intensity in patients with chronic LBP. Although previous research suggested that the magnitude of effect may depend on how exactly the placebo is delivered,[59,60] our subgroup analysis separating open-label placebo from placebo controls did not confirm this assertion, at least for patients with chronic LBP. However, it should also be acknowledged that the small number of included studies (n = 2) prevents definite conclusions at the moment. Most recently, another small open-label placebo trial in patients with chronic LBP was published, showing again a small effect in favor of the placebo intervention as compared to no intervention.[34] Another large and adequately powered clinical trial is probably needed to draw firm conclusions on the clinical relevance of the effect of open-label placebos in patients with chronic LBP.

A previous systematic review in patients with osteoarthritis concluded that contextual effects of interventions are very large and that the more invasive is an intervention, the larger is the placebo effect.[65] This conclusion was not confirmed by the subgroup analyses undertaken in this review (supplement E, supplemental digital content available at https://links.lww.com/PAIN/B345). In fact, an arguably more intensive intervention such as placebo injections did not show a significant reduction in pain intensity compared with no intervention and did not show a larger effect than placebo taping and placebo spinal manipulation (SMD 0.03, SMD 0.09, and SMD −0.28, respectively; see supplement C, supplemental digital content available at https://links.lww.com/PAIN/B345). Only placebo acupuncture and open-label placebo pills exhibited a significant result in reduction of pain intensity compared with no intervention (SMD −0.37, 95% CI −0.64 to −0.10; N = 2; P = 0.007, and SMD −0.33, 95% CI −0.61 to −0.05, P = 0.02, respectively), but this effect may not be clinically relevant. Hence, in contrast with Zou et al. we could not identify a specific intervention that elicits more profound effects on short-term pain intensity. This difference could be due to the fact that Zou et al.[65] did not rule out the effect of no intervention interventions in their review and the difference in patient population (ie, LBP vs osteoarthritis). Therefore, for patients with chronic LBP, the effect of placebo intervention may not be dependent on the type of intervention but more on other factors such as interaction with researchers or physicians. It is also important to note, however, that the number of studies was very small in these subgroup analyses, and certainly the power to detect significant differential findings was limited.

Strengths and Limitations

The use of no intervention control groups corrected for factors such as regression to the mean and natural course of disease thus allows for a valid estimate of symptom improvement that may be attributed to the administration of placebo interventions. Nevertheless, even including a waiting list or no intervention group in a trial gives rise to potential bias.[18] For instance, patients in the no intervention group may seek more alternative interventions than patients in the placebo group. Our RoB analyses did not control for the potential impact of co-medication or other co-intervention, and this may have slightly underestimated the magnitude of placebo effects. Conversely, considering that it is not possible to blind personnel and patients in a trial with a no intervention group, there may have also been some overestimation of the effectiveness of placebo interventions, although there is conflicting evidence on the difference in intervention effects between study with or without blinding.[47,48] Furthermore, the heterogeneity of placebo interventions summarized in our meta-analyses might be considered a limitation. However, in all analyses the τ2 estimate of between study heterogeneity indicated only low variation between effect sizes from individual studies, which allows for meta-analytic summary across the pool of studies despite the obvious diversity in placebo interventions approaches. Another limitation of our study is that by choosing to conduct meta-analyses on aggregate data only we were not able to investigate differential placebo effects with respect to certain patient characteristics. It is known from laboratory research that some individuals experience larger effects of placebo intervention than others depending on personality traits,[20] and this aspect could be further explored within individual patient data meta-analyses of the trials included in this review.

Future Research

Future research should focus on the factors that moderate and cause the effect in the placebo groups of RCTs. Most of the RCTs included in this systematic review reported pain intensity, but fewer studies reported on physical functioning or disability outcomes, and an even smaller number of studies reported HRQoL as an intervention outcome. We stress the need for uniformity in reporting outcomes, as previously suggested.[8] All studies on LBP should at least report on these 3 outcomes. Other outcome measures such as pain interference and pain self-efficacy can also be of use. Furthermore, we observed considerable variation regarding the follow-up period. We advise to use the standard follow-up moments as suggested by the Cochrane Back and Neck review group.[19] Only few data were available for medium-term follow-up, and only one study reported on long-term follow-up. However, from a clinical perspective it would be most relevant to see whether the effects caused by placebo interventions sustain over time in patients with LBP. More research is needed in (sub)acute LBP. With the overview on the current evidence based on the effects of placebo interventions from our meta-analysis and other literature, we suggest that future intervention trials in LBP should consider using a 3-arm design, which includes an intervention group, a placebo group (which closely resembles the intervention group), and a waiting list or no intervention group. We recognize recruitment of patients might prove difficult with such a design.[56] A potential solution to this problem could consist of an adaptive clinical trial design,[51] in which the no intervention group could be dropped if there is an apparent great difference in effectiveness. Alternatively, a trial with multiple baseline assessments might be considered if the use of a no intervention group is unfeasible or considered unethical.