Meta-analysis of the Placebo and Nocebo effects associated with Placebo treatment in randomized trials of lipid-lowering therapies

Abstract

Background

Randomized controlled trials (RCTs) of lipid-lowering therapy (LLT) in which the control groups received placebo without background LLT offer unique insights into the placebo and nocebo effects of lipid-lowering RCTs.

Methods and results

Embase and Medline were searched for hyperlipidaemia RCTs with placebo-controlled arms. Placebo arms with background LLT were excluded. A single arm meta-analysis of proportions was used to estimate major adverse cardiovascular events (MACE) and adverse events (AE). A meta-analysis of means was used to estimate the pooled mean differences of total cholesterol (TC), low-density lipoprotein (LDL), high-density lipoproteins (HDL) and triglycerides (TG).

A total of 40 RCTs and 37 668 placebo-treated participants were included. The pooled mean changes for TC, LDL, HDL, and TG were −0.019 mmol/L, −0.028 mmol/L, 0.013 mmol/L, and 0.062 mmol/L respectively among placebo-treated participants, indicating a modest placebo effect. The pooled average nocebo effect among placebo-treated participants was 42.62% for all AEs and 3.38% for musculoskeletal-related AEs, 11.36% for gastrointestinal-related AEs, and 6.62% for headaches. Placebo-treated participants in secondary prevention RCTs had a far higher incidence of these nocebo effects than primary prevention RCTs: any AEs (OR 6.76, 95% CI: 5.56–8.24, P < 0.001), and gastrointestinal-related AE (OR 1.23, 95% CI: 1.00–1.51, P = 0.049). No differences in nocebo effects were found between the placebo arms of statin and non-statin trials.

Conclusion

Our meta-analysis of placebo-treated participants in RCTs with no background LLT indicate a modest placebo effect but prominent nocebo effect of musculoskeletal, headache, and gastrointestinal symptoms that was greatest among secondary prevention RCTs. These findings may inform the design of future LLT RCTs.

Introduction

The World Health Organisation estimated the global prevalence of hyperlipidaemia to be ∼39% in 2008, and identified hyperlipidaemia as a contributing factor to a third of deaths from ischemic heart disease or ischemic stroke.^1,2 The global burden of hyperlipidaemia is expected to rise further due to economic growth, worsening dietary habits and sedentary lifestyles.³ Global data have shown that more patients have and are living longer with hyperlipidaemia, with the years lived with disability growing by 5.71 million from 1990–2019.³ To achieve cholesterol targets recommended by latest guidelines,^4,5 lipid-lowering therapies (LLTs) are often prescribed to patients and statins often being the first-line treatment in most patients.⁵ There is substantial evidence demonstrating the effectiveness of lipid-lowering therapies in the reduction of cardiovascular diseases (CVD)^6–11 with a 20% reduction of major vascular events for every 1 mmol/L decrease in low-density lipoprotein (LDL) levels with statins.¹⁰ Though statins and other LLTs are generally well-tolerated, they are associated with adverse events including musculoskeletal-related symptoms such as myalgia, that may reduce adherence and precede treatment discontinuation.^12,13 Studies have suggested that these AEs may be a result of a ‘nocebo’ effect, which is a neurobiological phenomenon in which patients treated with placebo report side effects similar to patients treated with the pharmaceutically active compound, due to the expectation of these complications arising rather than the intrinsic qualities of the drugs themselves.^14–16

The vigorous methodology and follow-up of randomised placebo-controlled trials, without background LLT, offers us a unique advantage in examining serial lipid profile of these subjects with pharmacologically-untreated hyperlipidaemia over time.¹⁷ As such, the difference in the placebo and nocebo effect on lipid profile and clinical outcomes between primary and secondary prevention cohorts with hyperlipidaemia can be elucidated.¹⁸ Our study sought to examine the placebo-controlled arms in hyperlipidaemia RCTs, through the lens of its placebo and nocebo effects, with the goal of deriving new insights into the design of future LLT RCTs.

Methods

Search strategy

This paper follows the guidelines set by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses,¹⁹ and was registered online with PROSPERO (Registration ID: CRD42022296824). Medline and Embase databases were accessed, and relevant papers were identified from inception to 7 December 2021. Keywords and MeSH terms synonymous with ‘hyperlipidaemia’, ‘ischemic heart disease’, ‘acute coronary syndrome’, ‘lipid-lowering’ and the individual drug classes including ‘statins’, ‘niacin’, ‘fibrates’, ‘bile acids’, ‘PCSK9’, ‘ACL inhibitor’, ‘omega 3 acids’, and ‘ezetimibe’ were utilised in the search strategy. The complete search strategy can be found in the Supplementary Material 1. References were imported into EndNote X9 for the initial sieve with the removal of duplicates. References of related reviews and included articles were also hand-screened to ensure a comprehensive search.^20,21

Study selection and extraction

Eligibility for inclusion was determined by four authors (NWSC, YHC, OZHL, and CXL) who screened articles independently from the initial sieve, with a third independent author involved in the resolution of disputes (MC). The following criteria was used for the inclusion of trials: (1) RCTs with a double-blinded, parallel design (this was done to reduce inter-mixing of groups within the cohort, which would have diluted the placebo drug); (2) studies evaluating adult patients (aged 18 and above) with or at risk of hyperlipidaemia randomised into placebo treatment; (3) prohibited intake of other lipid-lowering medications and (4) reported sufficient data on outcomes of interest including lipid outcomes, cardiovascular outcomes or AEs. Observational studies, case-control studies, reviews, meta-analyses, editorials, commentaries, conference abstracts, and non-English language articles were excluded. As this study focuses primarily on acquired hyperlipidaemia, studies describing familial or genetic causes of hyperlipidaemia were excluded. Other background medications such as antiplatelet medication, antihypertensive medication, and antidiabetic medication deemed to be non-lipid-lowering were allowed. Duplicated studies that obtained results from the same databases were removed, where the latest or the most comprehensive article was retained.

Data were extracted by four authors (YHC, OL, CXL, and YY) in an independent and blinded manner, including data on study characteristics (country, region, type of prevention, presence of a placebo run-in, and duration of follow-up), baseline information [total sample size, mean age, gender, body mass index (BMI), presence of other comorbidities] and outcomes [change in lipid parameters and number of AEs, which included drug related side-effects and major adverse cardiovascular events(MACE)]. Data on AEs in the included trials were primarily collected via (i) self-reporting by participants (especially for more benign AEs) and (ii) on routine follow-up, where the participant was further evaluated if there were derangements in the biochemistry profile. Estimated values of the mean and standard deviation were derived using formulas devised by Wan et al. when they were not provided.²² The units for cholesterol, TG, and glucose were millimoles per litre (mmol/l). The primary study outcome was the mean change on the lipid profile [total cholesterol (TC), high-density lipoproteins (HDL), LDL, and triglycerides (TG)] of patients. Secondary outcomes included the AEs patients experienced and MACE, including myocardial infarction (MI; fatal or non-fatal), stroke, and mortality (all-cause and cardiovascular-related).

Statistical analysis and quality assessment

All analyses were performed in RStudio (version 4.0.3). Three main types of meta-analyses were conducted. First, a meta-analysis of means was used to analyse continuous outcomes for the mean change in TC, HDL, LDL, and TG levels. Next, a single arm meta-analysis of proportions was conducted to analyse the proportion of patients experiencing AEs. A generalized linear mixed model (GLMM) with Clopper–Pearson intervals to stabilise the variance was used for this analysis.^23,24 GLMM has been found in recent simulation studies to be the most accurate transformation method for meta-analysis of single proportions compared with the more commonly used Freeman–Tukey double arcsine transformation.^24,25 Thirdly, a comparative meta-analysis was used to compare the difference in odds of adverse effects occurrence between the active and the placebo-arms included RCTs. Odds ratios (OR) were estimated for dichotomous outcomes respectively using DerSimonian and Liard random effects¹⁷. Statistical heterogeneity was measured using I², tau, Cochran Q test. An I² value of 25%, 50%, and 75% equated to small moderate and large amounts of heterogeneity respectively,²⁶ while a Cochran Q test value of ≤ 0.10 was taken as statistically significant. Hartung–Knapp adjustments were employed to adjust for confidence intervals by controlling for heterogeneity arising from between-study estimations.²⁷ The pooled studies differed in trial characteristics such as inclusion and exclusion criteria, study duration, presence of comorbidities, and drug class. Hence, the random effects model was used to incorporate heterogeneity into the analysis.²⁸ Analyses were only conducted when sufficient data could be extracted for the outcome of interest.

Additional analysis was performed based on primary and secondary prevention cohorts in the treatment of hyperlipidaemia. Primary prevention was defined as patients without established CVD who were taking LLT to reduce the risk of the first CVD event and secondary prevention was defined as patients who took LLT following the first CVD event.²⁹ Subgroup analyses were considered for relevant outcomes and performed according to the duration of intervention and follow up (<6 months, 6–12 months, >1 years), and the class of LLT that the placebo was compared with. Lastly, statistical analysis was conducted to compare the difference in AEs, cardiovascular and mortality outcomes between the placebo arms and active arms. A generalized linear model (Binomial family and logit link) was applied with inverse variance weightage to determine the odds ratios (OR) between them.^30,31 A P-value of <0.05 was considered as statistically significant. Publication bias was not conducted due to lack of a suitable assessment tool for single-arm meta-analyses.³²

Risk-of-bias assessment

To evaluate the potential for bias amongst the included trials, the revised version of the Cochrane Risk-of-Bias tool for randomized trials (RoB2) was utilised.³³ The RoB2 evaluates bias across five dimensions: (1) the randomization process, (2) deviations from intended interventions, (3) missing outcome data, (4) measurement of the outcome, and (5) selection of the reported result. Two independent and blinded authors (CXL and YY) assessed the included studies for risk-of-bias, with disagreements being resolved through discussion with a third independent author (YHC).

Results

Summary of included articles

A total of 9338 articles were identified in the initial search strategy, with the inclusion of 2425 duplicates. A further 6655 articles were then excluded, with a full-text review being conducted for 258 papers. A final count of 40 articles involving 42 placebo arms and 37 668 patients were included in the final analysis (Figure 1). The discrepancy between the number of articles and placebo arms arose because two trials investigated placebo arms of different durations, and were considered discrete cohorts and pooled separately.^34,35 Of these trials, 14 were conducted in the United States, 3 each in Japan, the Netherlands, and the United Kingdom, 2 in Canada, 1 each in Israel and Italy, and the remaining 13 trials carried out internationally. Of the 40 studies, 27 were single centre studies and 13 were multinational studies. Placebo served as a control in 19 statin trials, 4 with proprotein convertase subtilisin/kexin type 9 (PCSK9) inhibitors, 9 with fibrates, 3 with ezetimibe, 3 with cholesterylester transfer protein (CETP) inhibitors, and 2 with bile acid sequestrants. At baseline, majority were males (66.3%) and 18.1% were smokers. Of the study population, 17.8% and 37.5% had comorbidities of diabetes mellitus and hypertension respectively. The mean age and BMI of the study population were 57.3 ± 1.0 years and 27.7 ± 0.5 kg/m² respectively, with a mean LDL reading of 4.07 ± 0.12 mmol/L, HDL of 1.17 ± 0.03 mmol/L, and serum blood glucose of 6.18 ± 0.62 mmol/L. The summary of included articles and the detailed references to the included trials and baseline characteristics of the patients can be found in Supplementary Material 2. All included studies were found to have low to moderate risk of bias, with the detailed assessment found in Supplementary Material 3.

Change in lipid levels

Overall

Of the 17 052 patients studied in the placebo arm, the pooled mean change for TC was a mean difference (MD) of −0.019 mmol/L (95% CI: 0.072 mmol/L to 0.033 mmol/L; Figure 2a) over a mean time-period of 2.19 years. When grouped by study follow-up durations, the mean reduction of TC was a MD of −0.040 mmol/L (95% CI: −0.146 mmol/L to 0.066 mmol/L) in studies of <6 months follow-up, −0.047 mmol/L (95% CI: −0.303 mmol/L to 0.208 mmol/L) between 6–12 months, and −0.005 mmol/L (95% CI: −0.074 mmol/L to 0.063 mmol/L) for >1 year after placebo administration (P = 0.795).

A total of 29 592 patients were included in the analysis of LDL change, there was a mean reduction of MD −0.028 mmol/L (95% CI: −0.078 mmol/L to 0.021 mmol/L; Figure 2b). When analysed by study follow-up durations, the average change in LDL levels was a MD of −0.023 mmol/L (95% CI: −0.141 to 0.095) in studies of <6 months follow-up, −0.050 mmol/L (95% CI: −0.235 mmol/L to 0.136 mmol/L) between 6–12 months, and −0.026 mmol/L (95% CI: −0.084 mmol/L to 0.032 mmol/L) >1 year after placebo administration (P = 0.948).

Next, 23 208 patients were examined in the analysis of HDL, a MD of 0.013 mmol/L (95% CI: 0.002 mmol/L to 0.0.024 mmol/L; Figure 2c) over a mean follow-up duration of 1.91 years was found. In the subgroup analysis based on study follow-up duration, the average increase in HDL was a MD of 0.017 (95% CI: −0.014 mmol/L to 0.047 mmol/L) in studies of <6 months follow-up, 0.005 mmol/L (95% CI: −0.029 mmol/L to 0.038 mmol/L) between 6–12 months, and 0.019 mmol/L (95% CI: 0.007 mmol/L to 0.024 mmol/L) >1 year after placebo administration (P = 0.708).

Of the 25 845 patients analysed for TG, the overall mean change in TG after placebo administration was a MD of 0.062 mmol/L (95% CI: −0.070 mmol/L to 0.193 mmol/L) over a mean duration of 2.09 years (Figure 2d). The mean change in TG based on the study follow-up duration was a MD of 0.039 mmol/L (95% CI: −0.384 mmol/L to 0.462 mmol/L) in studies with <6 months follow-up, 0.037 mmol/L (95% CI: −0.073 mmol/L to 0.146 mmol/L) between 6–12 months, and −0.034 mmol/L (95% CI: −0.221 mmol/L to 0.154 mmol/L) >1 year after placebo use.

Primary and secondary prevention

A visual representation of the analysis can be seen in Figure 3. Overall, in the placebo arms of primary prevention cohorts, there was little change in TC (MD: 0.007 mmol/L, 95% CI: −0.090 mmol/L to 0.103 mmol/L; Figure 2a), LDL (MD: −0.004 mmol/L, 95%CI: −0.067 mmol/L to 0.060 mmol/L; Figure 2b), HDL (MD: 0.012, 95%CI: −0.008 mmol/L to 0.033 mmol/L, Figure 2c) and TG levels (MD: −0.058 mmol/L, 95%CI: −0.256 mmol/L to 0.141 mmol/L; Figure 2d). Analysis by the different follow-up timepoints found no significant differences in the mean change of TC (P = 0.248), LDL (P = 0.388), HDL (P = 0.204) and TG levels (P = 0.337).

For secondary prevention studies, we found that the average change in TC, LDL, HDL, and TG was a MD of −0.079 mmol/L (95% CI: −0.186 mmol/L to 0.028 mmol/L; Figure 2a), −0.105 mmol/L (95% CI:−.259 mmol/L to 0.049 mmol/L, Figure 2b), 0.017 mmol/L (95% CI: −0.007 mmol/L to 0.040 mmol/L, Figure 2c), and −0.105 mmol/L (95% CI: −0.259 mmol/L to 0.049 mmol/L; Figure 2d), respectively. When analysed by different follow-up timepoints, we found that larger improvements of lipid levels were noted (e.g. decrease in lipid levels for TC and LDL, and increase in HDL levels for HDL) for studies with <6 months and 6–12 months of follow-up as compared with follow-up periods of <12 months for TC, LDL, and HDL (Figure 2).

Moreover, subgroup analysis by the placebo arms of various drug classes were also conducted and the summary of their findings can be found in Supplementary Material 4. The analysis revealed no significant differences in lipid changes across the different categories of lipid-lowering medications, and between statin and non-statin drug classes.

Clinical endpoints

We analysed the prevalence of MACE in the placebo arms of hyperlipidaemia, with a summary of the findings displayed in Figure 3 and Supplementary Material 6. We found that the prevalence of the overall, fatal, and non-fatal MI was 2.29% (95% CI: 1.20–4.33), 0.45% (95% CI: 0.18–1.13%), and 2.57% (95% CI: 1.23–5.3%) respectively. The prevalence of stroke was 1.97% (95% CI: 1.00–3.81%), and the prevalence of all-cause and cardiovascular-related mortality was 2.11% (95% CI: 1.06–4.18%) and 0.86% (95% CI: 0.32–2.31%) respectively. When analysed by the type of prevention, secondary prevention studies had increased odds of overall and non-fatal MI, stroke, all-cause, and cardiovascular-related mortality, when compared with primary prevention studies (Figure 3 and Supplementary Material 5). When compared with the active arms of the studies, the placebo arm had increased odds of overall and non-fatal MI, stroke, all-cause mortality, and cardiovascular mortality. There was no statistical difference between the placebo and active arms for fatal MI (Supplementary Material 6 and 7).

Nocebo effect

Overall

A total of 25 608 patients were pooled across 22 trials for the nocebo effect analysis, with a pooled prevalence of 42.62% (95% CI: 28.75–57.76%) patients experiencing any AE after taking the placebo drug. Across the various study time-points, a larger proportion of the AEs were observed within 1 year, with 47.96% (95% CI: 38.13–57.95%) of patients experiencing any AE before 6 months, 36.22% (95% CI: 8.73–77.14%) at 6–12 months, and 41.56% (95% CI: 10.57–81.05%) beyond 12 months (Figure 4 and Supplementary Material 8). We found that the pooled prevalence of musculoskeletal-related AE, gastrointestinal AE and headache after placebo administration was 3.38% (95% CI: 1.66–6.74%), 11.36% (95% CI: 4.98–23.86%) and 6.62%(95% CI: 4.15–10.40%), respectively. For serious AEs, there was a pooled prevalence of 1.68% (95%CI: 0.48–5.72%, Figure 4). The prevalence of discontinuation of treatment due to AE was found to be at 4.54% (95% CI: 3.25–6.30%). A summary of the other AEs can be found in Figure 4. There were no statistical differences between the placebo and active arms for all of the AEs, except gastrointestinal AEs (Supplementary Material 6 and 7).

Subgroup analysis based on the class of LLT that the placebo was compared with, significantly affected the prevalence of overall AEs (P < 0.001), with ezetimibe-placebo having the highest prevalence of AEs (65.75%, 95% CI: 56.21–74.16%) followed by fibrates-placebo (57.86%, 95% CI: 0.83–99.56%), PCSK9 inhibitors-placebo (49.93%, 95% CI: 38.78–61.10%), statins-placebo (31.21%, 95% CI: 10.91% to 61.24%) and CETP inhibitors-placebo (30.25%, 95% CI: 0.20–98.93%). No significant differences were noted for serious AEs, treatment discontinuation and musculoskeletal-related AEs across the various lipid-lowering classes (Supplementary Material 9).

Primary and secondary prevention studies

Secondary prevention studies reported higher rates of AEs when compared with primary prevention studies (Figure 4), with increased odds of any AEs (OR:6.76, 95% CI: 5.56–8.24, P < 0.001), severe AEs (OR: 2.43, 95% CI: 2.00–2.96, P < 0.001) and gastrointestinal-related AEs (OR:1.23, 95% CI: 1.00–1.51, P = 0.049) when compared with primary prevention studies (Supplementary Material 5). No significant differences were noted for treatment discontinuation, musculoskeletal-related AEs, headache, back pain, myalgia, and myopathy (Supplementary Material 5).

Discussion

This meta-analysis offers a novel perspective of the placebo and nocebo effect in randomised subjects with hyperlipidaemia who did not receive pharmacologically-proven lipid-lowering agents. Interestingly, the largest improvements in lipid profile were found within a year of placebo treatment and mitigated beyond a year of follow-up. This magnitude of the placebo effect was more apparent in the secondary prevention cohorts compared with primary prevention ones. We also found that there was a substantial nocebo effect demonstrated in the control arms of RCTs, with up to 42.62% of patients experiencing AE, of which we identified three major nocebo effects, namely musculoskeletal-related AEs, gastrointestinal AEs, and headaches. The prominent nocebo effect was greatest amongst secondary prevention RCTs but did not discriminate between statin-placebo and non-statin-placebo lipid-lowering agents. These findings have important implications on the design of future RCTs of LLT.

Although placebos are designed to be an inert control arm, the study of placebo arms from RCTs have been used to study the natural history of other diseases.^36,37 The placebo effect on the lipid profiles observed in the placebo arms provide an important insight into the natural history of hyperlipidaemia in the setting of randomised trials, which could be attributed to a multitude of reasons. It is possible that the non-specific changes in attitude, environment, and psychosocial components lead to an improvement in patient education in clinical trials. This could in turn result in a better understanding of the disease process, evoking lifestyle modifications leading to reductions in weight, BMI, waist-to-hip ratio and blood pressure.^38,39 The changes could also be the result of a selection bias where more compliant individuals are recruited in clinical trials or a result of the Hawthorne effect.⁴⁰ Improvement in lipid markers generally occurred within the first year of being enrolled and were more pronounced in secondary prevention trials (Figure 3). The plateauing of this effect, however, showcases the potential limitations of diet and lifestyle modifications. Major studies in lifestyle changes have also shared a similar trend with small but significant improvements in cholesterol within 12 months and reversal beyond 12 months.⁴¹ Crucially, while there was a larger improvement in lipid profiles apparent in secondary prevention trials, adverse cardiac events were not mitigated in these individuals. The more pronounced risk of adverse cardiac outcomes in secondary prevention cohorts raise concerns regarding the appropriateness of utilising non-pharmacologically proven lipid-lowering therapies in secondary prevention clinical trials.

Recent evidence has also suggested that up to 78% of AEs caused by statins may potentially be contributed by the nocebo effect.^15,42 In our meta-analysis, we found that up to 42.62% of the population experienced AEs with 4.54% leading to treatment discontinuation. Additionally, comparison between active and placebo arms of trials found no significant differences in all AEs, apart from gastrointestinal events. This has implications on daily clinical practice and warrants questioning on the degree of premature discontinuation not based on medical grounds, but the nocebo effect. Long-term complications could include unnecessary increase in cardiovascular-related morbidity and mortality risks, as our paper has also found.^43,44 These findings mirror the recent Self-Assessment Method for Statin Side-effects Or Nocebo (SAMSON) trial¹⁵ which showed that the act of taking prescribed tablets rather than the content was the predominant cause for AEs. Apart from the commonly reported musculoskeletal AEs, our present study highlights two other important components of the nocebo effect—namely, gastrointestinal AEs and headache—which are common reasons for discontinuation of LLTs.⁴⁵ Moreover, studies have suggested that the nocebo effect is largely associated with statin use, with the consideration of switching to non-statin lipid-lowering therapy should side-effects arise.⁴⁶ However, the present analysis highlights that the contrary is true—non-statin-placebo lipid-lowering therapy (such as ezetimibe-placebo, fibrates-placebo and PCSK9 inhibitors-placebo) had higher overall rates of AEs compared with statin-placebo. Physicians should be cognizant of the high prevalence of the nocebo effect that is non-discriminatory to all classes of lipid-lowering therapy; and that current recommendations suggestive of the alternative use of non-statin therapy for individuals burdened by statin-related AEs and eventual treatment discontinuation, should be considered cautiously on a case-by-case basis.⁴⁷

Our current study emphasises the need for a better risk-benefit ratio analysis when designing the placebo-controlled arm in lipid-lowering trials. While double blinded placebo-controlled RCTs are currently considered the gold standard in examining the effectiveness of different therapies in various diseases,⁴⁸ it should be carefully implemented based on the current body of evidence. Compelling methodological reasoning is warranted such that participants are not unjustifiably exposed to the risk of no treatment.⁴⁹ This will help future LLT clinical trialists design their studies, estimate event rates, effect sizes of LLT treatment, and perform sample size calculations.

Study limitations

This is the first study examining the effects of nocebo and placebo on the lipids profile and outcomes of participants in hyperlipidaemia RCTs. However, there are some limitations to our study. First, we were unable to analyse the possible dose dependent placebo effect as there were no available data on the dose-related outcomes. Second, AEs could be confounded by the concomitant medications or comorbidities of the study population, and such studies are often unable to differentiate the impact of the placebo compared with these confounders. Third, most of the included studies were conducted before 2010 which may affect the reliability and generalisability of our findings. We mitigated this through a subgroup analysis based upon year of publication and no significant differences were found. Another important consideration would be the drop-in and drop-out rates of the trials as it would have been unethical for patients with suboptimal lipid profiles to carry on in the placebo arm. This could have led to bias arising due to exclusion of patients with suboptimal lipid control during the trial process. With regards to the drop-out rates seen in the included trials, our results revealed that the number of discontinuations was far below the reported overall rates of ‘nocebo’. Although this observation may be due to the difference in the nature of the outcomes (i.e. discontinuation is a ‘hard’ outcome of intolerance while nocebo rates encompass more subjectivity), one should be cognizant that the methods of data collection of AEs can also affect the ‘nocebo’ rates reported in the clinical trials. However, the lack of granularity in the methodology of the included articles did not allow for subgroup analysis based on the methods of AE assessment (e.g. self-reporting, specific questionnaire), and future studies exploring this possible confounder will be of interest.

Conclusion

Our meta-analysis of placebo-treated participants in RCTs with no background LLT indicate a modest placebo effect but prominent nocebo effect of musculoskeletal, headache, and gastrointestinal symptoms that was greatest among secondary prevention RCTs, which may inform the design of future RCTs of LLT. These findings also reinforce the need for physicians to remain cognizant of the high prevalence of the nocebo effect observed in all classes of lipid-lowering therapies with the goal to reduce unnecessary discontinuation of treatment.

Acknowledgment

All authors have made substantial contributions to all the following: (1) the conception and design of the study, or acquisition of data, or analysis and interpretation of data, (2) drafting the article or revising it critically for important intellectual content, (3) final approval of the version to be submitted. No writing assistance was obtained in the preparation of the manuscript. The manuscript, including related data, figures and tables has not been previously published and that the manuscript is not under consideration elsewhere.

Funding

This research did not receive additional support from organizations beyond the authors’ academic institutions.

Registration and protocol

The current study was registered with PROSPERO, ID: CRD42022296824.

Conflicts of interests

G.A.F. receives funding from the National Health and Medical Research Council (Australia), New South Wales Office of Health and Medical Research, and Heart Research Australia. She reports personal consulting fees from CSL, Janssen, Amgen and Boehringer Ingelheim and grants from Abbott Diagnostic outside the submitted work. In addition, G.A.F. has a patent Biomarkers and Oxidative Stress awarded USA May 2017 (US9638699B2) issued to Northern Sydney Local Health District.

M.Y.C.: Speaker's fees and research grants AstraZeneca, Abbott Technologies and Boston Scientific.

All other authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability

All articles in this manuscript are available from Medline and Embase.

Author contributions

Conceptualization—Yip Han Chin, Oliver Lim, Chaoxing Lin, Cheng Han Ng, Mark Chan, Nicholas WS Chew

Data curation—Yip Han Chin, Oliver Lim, Chaoxing Lin, Yu Yi Chan, Gwyneth Kong, Bryan Chong, Nicholas Syn

Formal analysis—Yip Han Chin, Oliver Lim, Chaoxing Lin, Yu Yi Chan, Gwyneth Kong, Cheng Han Ng, Bryan Chong, Nicholas Syn

Supervision—Nicholas Syn, Mark Muthiah, Shadab Siddiqui, Jiongwei Wang, Gemma Figtree, Nicholas WS Chew, Mark Chan

Validation—Yip Han Chin, Gwyneth Kong, Cheng Han Ng, Nicholas Syn, Nicholas WS Chew, Mark Chan

Writing, original draft—Yip Han Chin, Oliver Lim, Chaoxing Lin, Yu Yi Chan, Gwyneth Kong, Cheng Han Ng, Bryan Chong, Nicholas WS Chew, Mark Chan

Writing, review, and editing—Yip Han Chin, Oliver Lim, Chaoxing Lin, Yu Yi Chan, Gwyneth Kong, Cheng Han Ng, Bryan Chong, Nicholas Syn, Mark Muthiah, Shadab Siddiqui, Jiongwei Wang, Gemma Figtree, Nicholas WS Chew, Mark Chan

All authors have read and approved the final version of the manuscript for submission.

References

Author notes