A literature search of Medline (PubMed search engine) and Embase (Scopus search engine) database was conducted through inception up to December 16 ^th, 2022, following the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines. ²⁸ The literature search was independently performed by two reviewers, using the following algorithm: (carpal tunnel syndrome OR CTS OR entrapment neuropathy OR median nerve compression) AND (“dentists” OR “dental surgeon”).

The reference lists of all identified eligible studies were evaluated by both reviewers for potentially missed articles from the initial literature search. Following the aforementioned procedure, all studies were stored in the Zotero reference management software (version 6.0.18) and the duplicate citations were removed. ³⁶ The remaining articles were independently screened by two investigators to identify studies that met the pre-determined inclusion criteria. The study selection was conducted in two stages. First, article titles and abstracts were reviewed and those that did not meet our inclusion/exclusion criteria were removed. Secondly, the full texts of the remaining articles were retrieved and evaluated. If an absence in studies selection procedure was notified, the final decision was reached by team consensus.

Articles that examined specifically the prevalence rates of CTS among dentists were included. Only observational studies written in English language were inserted with no restriction on publication date. Case reports, case series with less than ten participants, review articles, clinical trials, animals studies, letters to the editor, books, expert opinion, conference abstracts, studies with no full-text available, studies not written in English language, articles reported solely the prevalence of CTS’ symptoms, studies regarding dental laboratory technicians and dental hygienists were excluded. In articles with overlapping populations, the most recent or most complete publication was considered eligible. The following variables were obtained from each study: the first author’s name, year of publication, study design, continent of origin, study period, total number of patients, proportion of males, mean age, participants with CTS and diagnostic procedures.

Quality appraisal was independently performed by two investigators using the National Heart, Lung, and Blood Institute (NHLBI) Quality Assessment Tools. The NHLBI quality assessment tool for Observational Cohort and Cross-Sectional Studies was employed. Individual studies were assessed for potential flaws in accordance the study methodology or the conduct of each survey that could jeopardize internal validity. For each of the fourteen questions, investigators could select one of the following answers: “yes”, “no”, “cannot determine” (e.g. data were unclear or contradictory) or “not reported” (e.g. missed data) or “not applicable” (e.g. not relevant question regarding this type of study). Study quality was defined as “low”, “moderate” or “high” risk of bias. ²⁵

Statistical analysis was carried out using RStudio (version: 022.12.0+353) software (RStudio Team (2022)). ³² The meta-analysis was conducted through metafor package. ³³ The DerSimonian and Laird random-effects model was used to estimate the pooled prevalence and its respective 95% confidence intervals (CI). Logit transformation was performed. Heterogeneity presence between studies was evaluated through visual inspection of the forest plot and by using the Cochran’s Q statistic and its respective p value. The Higgins I ² statistic and its respective 95% CI were used for quantifying the magnitude of true heterogeneity in effect sizes. An I ² value of 25%, 50%, and 75% indicated low, moderate, and high heterogeneity, respectively. To determine if the potential outlying effect sizes (as evaluated in the forest plot) were also influential, screening for externally studentized residuals with z-values larger than two in absolute value and leave-one-out diagnostics were performed. ³⁴ Due to high heterogeneity remaining, a moderator analysis was performed. In the conducted subgroup analysis, the continent of origin and the diagnostic procedure (verified during the implementation of each study or previously diagnosed) were chosen as the categorical moderators on effect sizes. In the performed meta-regression analysis with continuous variables, the year of publication and the proportion of males were assessed as moderators on effect sizes. Owing to the limited availability of data (less than ten studies for each covariate) regarding other variables (e.g mean age, obesity, diabetes, hypothyroidism, pregnancy, autoimmune diseases), these data were not included in this analysis. ¹⁷ Unless otherwise stipulated, the statistical significance was established at p=0.05 (two-tailed). Tests to evaluate publication bias, such as Egger’s test, ¹⁰ Begg’s test ⁵ and funnel plots, were developed in the context of comparative data. They assume studies with positive results are more frequently published than studies with negative results, however in a meta-analysis of proportions there is no clear definition or consensus about what a positive result is. ⁴ Therefore, publication bias in this current meta-analysis was assessed qualitatively.

As reported in the relevant section (Criteria for study selection and data extraction), manuscript that were only related to the prevalence of CTS’ symptoms (such as the study conducted from Prasad, D.A. et al. ³⁰ ) and studies regarding dental hygienists (such as the study conducted from Anton D., et al ³ and Cherniack M., et al. ⁸ ) were excluded. In total, ten (n=10) eligible studies (3,547 participants) were finally included in this analysis (see Figure 1 for the PRISMA flow chart). ³⁷ The descriptive characteristics of the incorporated research are presented in Table 1. All articles were published from 2001 to 2021 (conducted from 1997 to 2019). All of them were found to be of cross-sectional design. Most studies were contemplated in Asia (Iran, Lebanon, Saudi Arabia, n=6), followed by America (USA, Brazil, n=2) and Europe (Czech, Germany, n=2). The average percent of males was 54.22% while the mean age of participants ranged from 35 years to 46.4 years (median=38.2 years). Lastly, two studies were estimated as high quality (low risk of bias) and the remaining ones as moderate quality (moderate risk of bias).

A random-effects model analysis yielded an initial overall CTS prevalence of 9.87% (95%CI 6.84%-14.03%) with significant heterogeneity between studies I ²=90.55% (95%CI 79.29%-97.31%, p<0.01) ( Figure 2). The influence diagnostics are presented in Figure 3. The forest plot illustrating the results of the leave-one-out analysis is presented in Figure 4. As per them, no study was identified as being influential. In other words, there was no study identified that was capable of turning the effect of the analysis into some direction.

To investigate the effect of potential risk factors in the heterogeneity, a moderator analysis was performed. Forest plots of the subgroup analysis are illustrated in Figure 5 and Figure 6. The prevalence was 7.02% (95%CI 1.44%-27.99%) among studies conducted in Europe, 8.06% (95%CI 2.88%-20.60%) among studies conducted in America and higher among those conducted in Asia (11.71%) (95%CI 8.25%-16.35%). The prevalence was 12.47% (95%CI 6.38%-22.95%) for the group identified as having CTS through medical history and at least clinical examination or electrodiagnostic testing and 8.56% (95%CI 5.53%-13.01%) among those who identified solely through questionnaire (previously diagnosed, self-reported). Heterogeneity remained high in the subgroup analysis by both continent of origin and type of diagnostic procedure. In the meta-regression analysis with continuous variables, the year of publication and the proportion of males, no statistically significant (positive or inverse) modification was found as presented in Table 2.

The main strength was the comprehensive methodology applied for literature search, study selection, specific inclusion/exclusion criteria, screening for eligibility, quality assessment and pooling analysis of prevalence data from ten studies. Nonetheless, the present study had several limitations. It should be noted that the unidentified heterogeneity remained on high levels, therefore, the results should be interpreted with caution. The highly heterogenous outcomes across the included studies were expected due to the nature of this type of studies. Owing to the limited availability of data (less than ten studies for each covariate) regarding variables such as mean age, obesity, diabetes, hypothyroidism, pregnancy, autoimmune diseases, these data were not included in this analysis. Lastly, only observational studies written in English language were included resulting in the occurrence of reporting bias.