Open-Label Placebo Effects in Reducing Depression Severity Among Adults with Major Depressive Disorder: A Systematic Review and Meta-Analysis of Early Evidence

Introduction

Major depressive disorder (MDD) is a common mental illness characterized by persistent sadness and loss of interest or pleasure in previously enjoyed activities. MDD affects about 5% of adults worldwide, approximately 280 million people, and more than 75% of individuals in low- and middle-income countries do not receive treatment (WHO, 2023).

Consequently, MDD is a complex public health problem that affects patients, their families, and their social networks. Beyond its impact on quality of life and productivity, MDD is also associated with an elevated risk of premature death, largely due to suicide and comorbid medical conditions (Herrman et al., 2019; Marwaha et al., 2023).

Psychotherapy and pharmacotherapy with antidepressants and selective serotonin reuptake inhibitors (SSRIs) have long been considered the standard of care for this condition (Simon et al., 2024). However, patients taking antidepressants often experience adverse effects, ranging from gastrointestinal problems, sexual dysfunction, and weight change to increased suicidal thoughts and serotonin syndrome. These side effects can significantly impact treatment adherence and efficacy (Braund et al., 2021). Therefore, developing alternative treatments are critically needed for MDD patients.

Recently, open-label placebo (OLP) has shown positive results in people with clinical disorders and non-clinical symptoms (Schaefer et al., 2023). Randomized control trials (RCTs) demonstrated that OLP can reduce anxiety and acute stress (Schaefer & Enge, 2024; Schaefer et al., 2021); either independently or in conjunction with depressive symptoms (Guevarra et al., 2024). Additionally, other studies reported that patients suffering from a range of diseases such as irritable bowel syndrome, depression, allergic rhinitis, back pain, attention-deficit/hyperactivity disorder, induced pruritus, cancer-related fatigue, and menopausal hot flashes can benefit from OLP as an intervention (Antonelli & Donelli, 2019; Charlesworth et al., 2017; von Wernsdorff et al., 2021).

Despite growing evidence, to date no systematic reviews have comprehensively assessed OLP in the context of MDD, even though it may represent a promising strategy to mitigate the risks of antidepressant overprescription (Krockow et al., 2023).

Given the current limitations of standard care, especially its prevalent adverse effects, and the need to offer a low-risk alternative, we sought to evaluate and synthesize the existing evidence on using OLP to reduce depression severity in patients with MDD. Addressing this critical knowledge gap will provide evidence-based guidance for clinical practice, policy development, and potential future research directions.

Methods

A systematic review of RCTs was conducted following the Cochrane Handbook for Systematic Reviews of Interventions methodology (Higgins et al., 2024) and reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines (Page et al., 2021). The review protocol was developed internally, a priori, to minimize bias and ensure methodological rigor throughout the process. Moreover, it was not registered in any international systematic review registry. The systematic literature search commenced on April 10th, 2025. Comprehensive searches, study selection, data extraction, and quality assessment were conducted until June 15th, 2025, and were performed by multiple independent reviewers to enhance reliability and reduce selection bias.

Search strategy

Following our research question, search strategies were prepared using MeSH terms (open label placebo, non-deceptive placebo, depressive disorder major, major depressive disorder, depressive symptoms), standardized terms, and Boolean operators for the databases Cochrane, PsycNet, PMC, ClinicalTrials.gov, Scopus and PubMed (Supplementary Table 1).

Inclusion and exclusion criteria

We included RCTs involving adults (≥18 years) diagnosed with MDD, treated with OLP, and reporting depression severity using validated scales. Articles had to be published in English, with full-text available online. We excluded studies involving bipolar disorder, psychotic symptoms, double-blind designs, and study protocols.

Selection of studies and data extraction

All authors were involved in the review process. Two reviewers conducted the initial search for articles. A third reviewer intervened if there was a disagreement regarding eligibility, and a consensus was reached. The reviewers independently screened the titles and abstracts after removing duplicates in the Covidence web-based platform (Kellermeyer et al., 2018). Subsequently, full texts were reviewed if a paper was considered relevant based on the inclusion and exclusion criteria.

Extracted items included:

• Study characteristics: author, year, country, design, funding source
• Population: inclusion/exclusion criteria, baseline demographics (age, gender), diagnostic criteria for MDD, depression severity
• Intervention details: group allocation, type and comparator, route of administration, frequency, duration, dosage
• Study setting and recruitment methods, where available
• Outcomes: depression scales used, direction of scoring, range, baseline and 4-week follow-up means and standard deviations

For meta-analysis, the results of means, sample size, and SD were extracted to calculate the standardized mean difference (SMD) and its 95% CI.

Data synthesis

Selected clinical trials were summarized in Microsoft Excel tables. Given the small number of eligible trials, the quantitative synthesis was conducted for exploratory purposes. For the meta-analysis, a random-effects model was used to combine the standardized mean differences between studies. Hedges's g method was used to estimate effect sizes, and 95% confidence intervals were calculated. Cochran's Q and I², and Tau-squared statistics were used to assess the heterogeneity of the included studies. An I² > 50% or P < 0.05 indicated statistically significant heterogeneity. All statistical analyses and forest plots were performed using STATA MP version 19 statistical software (StataCorp, 2025).

Risk of bias assessment

Three independent reviewers (ME, MLGS, and SJFG) assessed the risk of bias of the included studies using the Cochrane Risk-of-Bias 2 tool (Sterne et al., 2019). Following Cochrane's recommendations, a traffic light plot was created manually using Excel to visually summarize the risk-of-bias assessments. After completing independent assessments, the reviewers convened to discuss any discrepancies, and consensus was reached through a joint meeting to ensure uniform judgments across all domains.

Results

The search strategy identified a total of 1,886 studies. One duplicate was identified manually, and 49 additional duplicates were detected by Covidence automatic detection tool, resulting in 1,836 studies screened by title and abstract. Of these, 1,812 records were excluded based on the inclusion criteria. A total of 24 studies were selected for full-text review. Twenty-one of these were excluded for the following reasons: wrong patient population (n = 12), wrong study design (n = 5), wrong intervention (n = 2), congress abstract (n = 1), and trial closed (n = 1). No studies were found to be ongoing at the time of the review. Ultimately, three RCTs (Kelley et al., 2012; Nitzan et al., 2020; Schienle et al., 2022) met all inclusion criteria and were included in this systematic review. The study selection process is detailed in the PRISMA flow diagram (Figure 1).

Figure 1. PRISMA flowchart illustrating the identification, screening, and inclusion of studies in the systematic review. Studies were identified from six databases: Cochrane (n = 794), PsycNet (n = 560), PMC (n = 385), ClinicalTrials.gov (n = 91), Scopus (n = 33), PubMed (n = 23). After removing 50 duplicates, 1,836 studies were screened. Following full-text review of 24 studies, 3 RCTs were ultimately included.

Description of the studies

Table 1 summarizes the key characteristics of the three RCTs included in this review. The studies were published between 2012 and 2022 and were conducted in the USA, Israel, and Austria. All trials utilized randomized controlled designs comparing OLP to either treatment-as-usual (TAU) or waitlist controls. Kelley et al. and Schienle et al. used a parallel-group design to assess primary outcomes, while Nitzan et al. used a crossover design after four weeks of parallel-group design. Sample sizes across the studies ranged from 20 to 60 participants.

Population

The three studies enrolled a total of 118 participants (mean age 46.8 ± 14.2 years; females 77.8%) diagnosed with nonpsychotic MDD. Diagnostic criteria varied depending on the study (DSM-IV, ICD-10, DSM-V). Most study participants exhibited moderate depression scores, based on the different scales used, and only the study by Schienle et al. included patients with severe depression (9/60 participants; 15%). No significant differences in baseline characteristics between groups were reported.

Intervention characteristics and effects

The OLP interventions varied in form, delivery, and context, ranging from oral capsules to placebo oil, but in all cases, patients were explicitly informed that they were receiving an inert substance. In the two studies conducted by Kelley et al. and Nitzan et al., patients were instructed to take two OLP pills, twice daily. The third study by Schienle et al. used sunflower oil in a labeled dropper bottle as their OLP and instructed participants to take three oral drops daily prior to home-based relaxation training.

Outcomes

The primary outcomes across the three studies were reduction in depressive symptoms, assessed using standardized tools such as the Hamilton Depression Rating Scale (HAM-D-17), the Beck Depression Inventory-II (BDI-II), and the Quick Inventory of Depressive Symptomatology (QIDS-SR16). These measures were used to evaluate changes in symptom severity over time and between treatment groups.

Secondary outcomes included adherence to therapeutic activities (e.g., frequency of relaxation practice), subjective ratings of relaxation quality, perceived effectiveness of the placebo, placebo intake quantity, clinician-rated global improvement, and dropout rates, and other correlative relationships. Nitzan et al. also explored subgroup effects based on age and onset of depression, non-geriatric adults (<65 y) with early onset of the disorder (<50 years).

Main findings

The three reviewed RCTs were not fully concordant in their conclusions, as their results diverged based on the specific analytical approach employed and the context studied (Table 2). Of these studies, Schienle et al. demonstrated that OLP interventions were associated with a statistically significant reduction in depressive symptoms over time. The reduction in depressive score was measured using the self-reported BDI-II and analyzed with a mixed-model Analysis of Variance (ANOVA), focusing on the group-by-time interaction (group×time).

Despite this statistical significance, the research team reported that these findings were not clinically meaningful, as BDI-II scores did not show a substantial difference between the intervention groups (CBT+OLP versus CBT) following the 3-week program. Furthermore, although most participants (83%) experienced a shift to a lower depression category by the end of the program, there were no significant differences observed between the assigned groups regarding the number of patients within each severity category.

Interestingly, the OLP group exhibited higher adherence to adjunctive therapeutic activities, such as relaxation exercises, when compared to the control groups. Nevertheless, this increased adherence did not translate into significant time-by-group interactions concerning the frequency or quality of relaxation reported.

Studies conducted by Kelley et al. and Nitzan et al. were unable to conclude statistically significant differences between OLP and control groups at their respective primary assessment endpoints using HAM-D-17 and self-reported QIDS-SR16 as the main outcome, respectively. However, Nitzan et al. conducted a subgroup analysis using repeated measures ANOVA, which focused specifically on non-geriatric patients with early-onset depressive disorder. It concluded that for this subgroup, the rate or magnitude of improvement over the course of the treatment was significantly different between the groups, with the OLP arm demonstrating a greater reduction in symptoms, indicated by a significant Time×group interaction.

Lastly, although dropout rates and acceptance of OLP varied across studies, the use of OLP was associated with a higher rate of dropout, particularly among participants who perceived the placebo as ineffective in the Schienle et al. study.

Meta-Analysis

A meta-analysis was conducted using post-intervention mean scores and standard deviations at 4 weeks, as reported by two RCTs (Nitzan et al., 2020; Schienle et al., 2022) (Figure 2). Results from Kelley et al. were excluded from the quantitative synthesis as it reported outcomes only at 2 weeks post-intervention, precluding comparability with the other trials. The pooled analysis yielded a standardized mean difference (Hedges's g) of –0.43 (95% CI: –0.83 to –0.04, p = .03) in favor of open-label placebo. Heterogeneity across studies was negligible (τ² = 0.00, I² = 0.00%, p = .93). Neither individual study reached statistical significance (Nitzan et al.: g = –0.41 [95% CI: –1.04 to 0.22]; Schienle et al.: g = –0.45 [95% CI: –0.95 to 0.06]), although both studies showed effect estimates favoring OLP.

Figure 2. Forest plot of the meta-analysis comparing OLP versus control on depression severity at 4 weeks post-intervention. Nitzan 2020: Hedges's g = –0.41 [–1.04, 0.22], weight 39.18%. Schienle 2022: g = –0.45 [–0.95, 0.06], weight 60.82%. Overall: g = –0.43 [–0.83, –0.04]. Heterogeneity: τ² = 0.00, I² = 0.00%, H² = 1.00. Random-effects REML model.

Risk of bias assessment

As summarized in Figure 3, all three studies raised some concerns regarding bias arising from the randomization process (Domain 1). This was because none of the three studies described the method of allocation, merely referring to their studies as randomized controlled trials, without providing further details on randomization techniques used.

It is important to interpret certain domains in the specific context of open-label placebo (OLP) trials. By design, OLP interventions are unblinded; therefore, domains related to deviations from intended interventions (Domain 2) and measurement of the outcome (Domain 4) require nuanced consideration. However, Kelley et al. and Schienle et al. still presented some concerns related to Domain 2, because caregivers, personnel delivering the intervention, and/or psychologists were aware of participants' assignments.

Similarly, Nitzan et al. and Schienle et al. were judged to be at high risk in the measurement of the outcome (Domain 4) because outcome assessors were aware of participants' treatment allocations.

Schienle et al. showed a high risk of bias in Domain 3 because of a substantial differential dropout rate—27% in the CBT + OLP group compared to 7% in the CBT-only group at follow-up.

Pertaining to domain 5, Kelley et al. raised some concerns as no statistical analysis plan was mentioned in the trial registration or protocol, and multiple outcome measures (HAM-D, QIDS, and SDQ) were used, allowing flexibility in the emphasis of reported results. Finally, although Nitzan et al. used QIDS to assess depression severity, their clinical trial registry (NCT02666989) specified intention to use HAM-D-17 as a readout to be analyzed. This clear shift in the methodology was alarming and indicated potential bias leading to high concerns for domain 5.

Figure 3. RoB2 assessment of included articles (n = 3). Domains: D1 = Bias from the randomization process; D1b = Bias from timing of identification and recruitment; D2 = Bias due to deviations from intended intervention; D3 = Bias due to missing outcome data; D4 = Bias in measurement of the outcome; D5 = Bias in selection of reported results. Judgement: ✕ High risk; – Some concerns; + Low risk.

Discussion

This systematic review summarizes evidence from three RCTs evaluating OLP for MDD in adults, which collectively suggest modest symptom improvements compared with TAU or control conditions, though not all results reached statistical significance. While previous research has studied the potential benefits of OLP for a variety of clinical conditions including chronic low back pain (Flavio-Reis et al., 2025), cancer-related fatigue (Hoenemeyer et al., 2021), and irritable bowel syndrome (Charlesworth et al., 2017), to our knowledge, this is the first systematic review to focus exclusively on the use of OLP for depression.

Despite an extensive search strategy and a meticulous screening process conducted through Covidence, the evidence base remains scarce. Only three eligible studies met the inclusion criteria, underscoring the novelty and underexplored nature of OLP as a treatment for MDD. Nevertheless, the studies available provide initial insight into the feasibility, safety, and potential efficacy of this approach. A key limitation of the review is relying exclusively on results in the English language. This potentially entails losing studies conducted in other languages.

There was substantial heterogeneity among the included RCTs, particularly with respect to the diagnostic criteria used to define the study population and the assessment of primary outcomes. The varying diagnostic tests employed across these studies have direct implications for the comparability of their findings. The shift in diagnostic criteria from the pilot study by Kelley et al. suggests that subsequent studies may have included a more heterogeneous MDD population. DSM-IV criteria were often considered more stringent for MDD, emphasizing core depressive symptoms. In contrast, DSM-V and ICD-10 include broader presentations such as prolonged grief or anxiety features. This may lead to the inclusion of individuals with less severe or more complex, co-morbid symptom profiles under the definition or spectrum of MDD.

Similarly to inclusion criteria which were based on different diagnostic tests, outcome assessment suffered substantial heterogeneity. Kelley et al. (2012) and Schienle et al. (2022) incorporated a mix of clinician-rated and self-report instruments, while Nitzan et al. (2020) relied solely on the self-reported version of the QIDS. Although interviews were conducted at 0, 4, and 8 weeks, the lack of clinician-rated outcome measures may have increased the risk of measurement bias in their study.

While both clinician-rated and self-reported scales are generally sensitive in detecting treatment effects in clinical trials, they capture distinct features of depression. Clinician-based rating scales, such as the HAM-D-17, offer an objective, expert assessment of observable signs. Conversely, self-report scales, including the BDI-II and QIDS-SR16, capture patients' subjective experiences, encompassing internal states, cognitive distortions, and personal distress that may not be readily apparent to an observer.

Hence, it is important to point out that the choice of primary outcome measure profoundly shapes the narrative of OLP efficacy. Given that OLP effects are hypothesized to involve psychological mechanisms such as expectation and conditioning, the patient's subjective experience, as captured by self-report measures, might serve as a more direct or sensitive indicator of the impact of OLP. However, clinician ratings are often perceived as more robust indicators of clinical significance.

Although the analysis by Nitzan et al. did not draw out significant differences between OLP and TAU groups, subgroup analysis on non-geriatric participants with early onset highlights an important variable that could potentially influence a patient's response to OLP interventions and prompts further investigation. The results of the meta-analysis indicate that OLP may offer a modest but consistent reduction in depression severity relative to control conditions. The absence of heterogeneity suggests that the effect was stable across studies.

While the use of RCT designs across all three studies strengthens internal validity by reducing confounding, several limitations warrant caution. The meta-analysis included only two small trials, and all three studies enrolled modest sample sizes (20–60 participants), limiting statistical power and the reliability of heterogeneity estimates. Notably, one was a pilot study focused on feasibility rather than efficacy.

Risk of bias was also a concern, as assessed using the Cochrane Risk of Bias 2 (RoB2) tool. Two of the three studies were rated as having a high overall risk of bias, while Kelley et al. (2012) raised some concerns. The most frequent sources of potential bias were related to the randomization process, deviations from intended interventions, missing outcome data, and the measurement of outcomes.

Based on the three available randomized controlled trials, there is currently insufficient evidence to conclude that OLP reduces depressive symptoms in adults with MDD. Although all studies reported trends toward symptom improvement—particularly when OLP was combined with usual care or behavioral interventions—these findings were derived from small, heterogeneous samples. The study populations were predominantly female and included mostly patients with mild to moderate depression, leaving uncertainty about OLP's applicability to men or those with more severe or comorbid conditions. Despite these limitations, OLP remains an ethically transparent, low-risk intervention worth further investigation. Future studies should recruit larger and more diverse samples, standardize OLP delivery, assess both clinician- and self-rated outcomes, and evaluate the durability of effects.