This meta-analysis was conducted strictly in accordance with PRISMA guidelines.25 This study was prospectively registered in the PROSPERO registry, with the registration number CRD42022307045. The PRISMA checklist is provided in Supplementary Table S1.
Search Strategy
We included RCTs and observational studies (either cohort or case–control studies) that evaluated the preventative or treatment effects of statins on delirium in critically ill or surgical patients in an intensive care unit (ICU). We searched PubMed, the Cochrane Library, and the EMBASE database, limiting the search to human patients and articles in English published until December 31, 2021. The details of the search strategy are provided in Box 1. All identified articles were imported into EndNote software (version X8, Thomson ResearchSoft, CT, USA) to remove duplicate records automatically. Two reviewers (Y. H. C. and J. Y. W.) screened all titles and abstracts independently and evaluated relevant articles. In addition, we searched available bibliographies and review articles for additional potential articles that can be included in this study. Disagreements were resolved through consensus and consultation with a third reviewer (M. J. L.).
Data Extraction and Outcome Measures
By using a standardized data extraction form, we obtained information on the study year, study location, study population, participant characteristics, delirium outcome, and effect sizes and their 95% CIs. For studies that reported only the crude effect size, we contacted authors and requested them to provide data on the adjusted effect size and relevant information. The outcomes assessed were the incidence of delirium as defined using the Confusion Assessment Method (CAM), CAM-ICU, or diagnostic codes from administrative databases. We obtained sufficient published data to calculate the estimates of the crude or adjusted OR with the 95% CI of delirium risk related to statin use in eligible studies. In addition, delirium-free days were counted as an outcome in a few studies.
Risk of Bias Assessment
To examine the quality of the included studies, we used the Cochrane Risk of Bias tool for RCTs26 and the Newcastle–Ottawa scale for observational studies (Supplementary Appendix 1).27 The Cochrane Risk of Bias tool assesses the potential sources of bias including selection bias (random sequence generation and allocation concealment), performance bias (blinding of participants and personnel), detection bias (blinding of outcome assessment), attrition bias (incomplete outcome data), and reporting bias (selective reporting).26 Each study was categorized as having either low risk (green), unclear risk (yellow), or high risk (red) of bias. The risk of attrition bias was considered to be low if the dropout rate was lower than 20% due to the inclusion of few participants in each of the studies. For assessing reporting bias, the protocols for each study were searched on clinicaltrials.gov and biomedcentral.com. The Newcastle–Ottawa scale evaluates three major sources of bias in epidemiological studies by using five items: one for selection bias, two for measurement bias, and two for confounding bias.27 A study is classified in each domain as having high or low risk in accordance with prespecified criteria. All included studies were evaluated by two pharmacists (Y. H. C. and J. Y. W.) independently; any discrepancies were resolved through discussion.
Statistical Analysis
We conducted a meta-analysis to determine the association of statin use with delirium in RCTs and observational studies. The effect size and 95% CI were defined as the OR and 95% CI to indicate the difference in the incidence of delirium between statin use and nonuse. A random-effects or fixed-effects model was selected to calculate the pooled effect size, and a random-effects model was used if the study population had high heterogeneity. We assessed statistical heterogeneity across studies by using the I2 statistic.28 I2 values of < 50%, 50–75%, and > 75% are deemed to indicate low, moderate, and high heterogeneity, respectively. Exclusion sensitivity analysis was conducted to investigate the effect of individual studies. We performed subgroup analysis on the basis of the study design (RCTs versus other observational studies), country in which the studies were performed, and publication before 2015 versus after 2015. We used funnel plots, Egger’s test, Duval and Tweedie’s trim-and-fill approach, and the classic fail-safe N to assess publication bias.29, 30 In addition, we performed meta-regression by adjusting for age and male sex to explore the potential sources of heterogeneity. The Grading of Recommendations Assessment, Development, and Evaluation (GRADE) approach31 was employed to evaluate the quality of evidence; in this approach, the four grades are high, moderate, low, and very low quality. All statistical analyses were performed using RevMan5 and Comprehensive Meta-Analysis Software.
Box 1
PUBMED on December 14, 2021
MeSH terms
1. ‘‘Hydroxymethylglutaryl-CoA Reductase Inhibitors’’ = 32,147
2. ‘‘delirium’’ = 11,025
Text terms
3. ‘‘statin’’ = 61,940
4. “atorvastatin“ = 10,945
5. “cerivastatin” = 789
6. “fluvastatin” = 2,157
7. “lovastatin” = 12,291
8. “pravastatin” = 4,850
9. “rosuvastatin” = 12,912
10. “simvastatin” = 11,544
11. ‘‘deliirum’’ = 20,886
Search strings (all inclusive):
[(1 OR 3 OR 4 OR 5 OR 6 OR 7 OR 8 OR 9 OR 10) =65,927 AND (2 OR 11) = 20,886]=69
Cochrane Library on December 14, 2021
MeSH terms
1. ‘‘Hydroxymethylglutaryl-CoA Reductase Inhibitors’’ = 3,634
2. ‘‘delirium’’ = 898
Text terms
3. ‘‘statin’’ = 7,154
4. “atorvastatin“ = 5,665
5. “cerivastatin” = 196
6. “fluvastatin” = 751
7. “lovastatin” = 987
8. “pravastatin” = 1,967
9. “rosuvastatin” = 2,028
10. “simvastatin” = 3,986
11. ‘‘delirium’’ = 4,290
Search strings (all inclusive):
[(1 OR 3 OR 4 OR 5 OR 6 OR 7 OR 8 OR 9 OR 10) =15,510 AND (2 OR 11) = 4,290]=22
EMBASE on December 14, 2021
Emtree term
1. ‘‘Hydroxymethylglutaryl-CoA Reductase Inhibitors’’ = 169,687
2. ‘‘delirium’’ = 36,008
Search strings:
3. (1 AND 2) =424
4. 3. AND 'human'/de = 418
5. 4. AND ('clinical trial'/de OR 'cohort analysis'/de OR 'controlled clinical trial'/de OR 'controlled clinical trial topic'/de OR 'controlled study'/de OR 'cross sectional study'/de OR 'major clinical study'/de OR 'observational study'/de OR 'prospective study'/de OR 'randomized controlled trial'/de OR 'randomized controlled trial topic'/de OR 'retrospective study'/de) = 173