Depression and Anxiety in Rosacea Patients: A Systematic Review and Meta-Analysis

This systematic review and meta-analysis was performed in accordance with PRISMA guidelines [20]. Eligible studies reporting the prevalence rates or/and odds ratios (ORs) were identified by searching the PubMed, Embase and Medline databases using various forms and combinations of search terms, including “rosacea,” “depressive disorder,” “depression,” “anxiety,” “anxiety disorders,” “psychiatric disorders,” “mental disorder,” “phobia” and “suicide.” No language limitation was imposed. The search included all articles available in the databases from database inception to 5 October 2020. Reference lists of relevant publications were screened manually for potential eligibility. Two investigators (SLX and RD) performed independent screenings of literature.

No restrictions were made on study designs with the aim to increase the number of relevant publications. The inclusion criteria were: (1) original study; (2) study assessed the prevalence rates or the relationship of depression or anxiety in rosacea patients; (3) Study was conducted in adults aged > 18 years; (4) study included at least 20 subjects with rosacea. People with rosacea or depression or anxiety were diagnosed by physicians, on the basis of International Classification of Disease (ICD) codes or by self-reported diagnoses from physicians (patient reported that he or she has been diagnosed by a physician as having rosacea). No restriction was made to rosacea subtypes. We also included studies using screening questionnaires to indicate the possible occurrence of depression and anxiety. In present study, we used physically-diagnosed, self-reported and screening questionnaire methods to diagnose study variables. Studies without sufficient data to calculate prevalence or ORs were excluded. When data were duplicated in multiple publications, the most informative article was included in the analysis.

The following data were extracted from each article: first author, year of publication, geographic region of study, study design, study setting, study size, mean age of patients, distribution by sex, methods used to diagnose rosacea, depression and anxiety, prevalence of depression or anxiety, crude and/or adjusted OR estimates with corresponding 95% confidence intervals (CIs). Since the number of studies that stratified rosacea by clinical subtypes was insufficient, we excluded these data. If a study did not provide ORs directly, we calculated this parameter based on raw data extracted from the study. Any disagreement on data was resolved with consensus among all authors. If certain data were unavailable or unclear in published form, the authors of the study were contacted by email and asked to provide the original information.

The quality of each included study was assessed using the Newcastle–Ottawa Scale (NOS), in which studies are judged on eight items regarding representatives of study population, comparability and ascertainment of exposure or outcome of interest [21]. An adapted version was used for cross-sectional studies (Electronic Supplementary Material [ESM] File S1). Higher scores in the NOS indicate better methodical quality, and studies that met ≥ 5 of the NOS items were considered to be of high quality.

The prevalence rates of rosacea patients with depression and anxiety were calculated in both controlled and uncontrolled studies. In controlled studies, ORs and 95% CI for depression and anxiety between rosacea patients and healthy controls were analyzed. If some study did not provide overall ORs for depression risk but presented separate ORs for different exposure levels of rosacea, we combined the corresponding odds estimates on the basis of Hamling’s method [22]. The pooled prevalence and ORs with 95% CIs obtained from included studies were estimated by random-effects models because of significant heterogeneity (I² > 50% for all analyses) [23]. Heterogeneity across studies was assessed by I² statistics, which represents the percentage of total variation contributed by a cross-study variation [24]. A study was defined to be heterogeneous if the P < 0.1 or I² > 50%.

Potential sources of heterogeneity were explored using subgroup and random-effect meta-regression analyses. Subgroup analyses were conducted on study setting, methods used to assess rosacea, depression and anxiety and geographic areas. Meta-regression analyses were conducted only when there were data from > 5 independent datasets. For the meta-regression analyses, we took mean age, percentage of female subjects and NOS scores into consideration. In studies assessing the ORs for depression and anxiety, sensitivity analyses were performed to investigate the influence of the individual study on the pooled estimate by subsequent omitting studies and determining the potential weighted outliers. When > 10 studies were available, we used Egger' regression and Begg rank to assess the publication bias. The trim-and-fill method was used to determine the effect size when publication bias was noted.

The initial comprehensive search yielded a total of 164 articles. The title and abstract of each of these articles were then screened to determine potential eligibility, yielding 37 articles for full-text assessment. Combined with the additional two articles identified from relevant references, 14 articles were ultimately included in the systematic review and meta-analysis. The search strategy is illustrated in the PRISMA flow diagram shown in Fig. 1.

In this meta-analysis of the 14 studies [17, 18, 25‐36], we pooled data on 14,134,021 patients with rosacea and 686,447,139 controls (Table 1). Only one study included > 100,000 patients and three studies included > 10,000 patients. Five studies were conducted on the general population, and nine studies were conducted on a hospital-based population. In most studies, rosacea was diagnosed by a physician or using an ICD code. Only two studies defined rosacea through self-report. All included studies ascertained outcomes of depression or anxiety using screening questionnaires, with the exception of five studies which evaluated depressive symptoms by ICD codes. In all, 14 studies examined the association between rosacea and depression, and seven studies examined the association between rosacea and anxiety.

According to the NOS system, 13 of the 14 included studies can be considered to be of high quality. The remaining study met only four of the nine NOS items. Information on the NOS scores is presented in more detail in ESM Table S1.

A total of 14 studies reported the prevalence of depression in patients with rosacea (Fig. 2a; Table 2). The pooled random-effects prevalence of any type of depression was 19.6% (95% CI 15.0–24.3%). Heterogeneity was significant (I² = 99.9%, P < 0.01). There was publication bias according to the Egger regression (P = 0.012), but no publication bias was found based on the Begg rank (P = 0.23). The trim-and-fill method showed no study was filled up at final analysis. In the subgroup analysis, the prevalence of depression among the general population was 15.9% (95% CI 11.1–20.7%), whereas in studies in a hospital-based population, the prevalence of depression was 22.0% (95% CI 15.2–28.8%). When depression was stratified by the method used to diagnose rosacea, the prevalence of depression in rosacea patients determined using screening questionnaires was 26.2% (95% CI 19.3–33.2%); however, the proportion of depressed patients decreased when outcomes were measured by ICD codes (9.2%, 95% CI 1.9–16.6%). A significant difference in the prevalence of depression was detected in subgroups diagnosed with depression using different methods (P < 0.01). In rosacea patients diagnosed by physicians or ICD codes, the pooled prevalence of depression was 18.7% (95% CI 13.7–23.7%), while in those diagnosed through self-reported records from physicians (only 2 studies), it was 24.9% (95% CI 19.0–30.8%). Seven studies were performed in Europe, with a pooled proportion of depression of 23.4% (95% CI 19.2–27.5%); three studies were performed in North America, with a pooled prevalence of 14.8% (95% CI 3.6–26.1%); the pooled prevalence of depression in Asia was 11% (95% CI 2.9–19.2%). The prevalence of depression was significantly different between these three continents (P = 0.02). According to random-effect meta-regression analyses, the percentage of women with rosacea (coefficient of variance [CV] 0.18, adjusted [adj.] R² = − 8.9%, τ² = 0.02, P = 0.71) and the mean age (CV − 0.002, adj. R² = − 6.71%, τ² = 0.01, P = 0.50) did not moderate the rate of depression in rosacea patients. However, the NOS scores of included studies explained 41.4% of the variability between studies (CV − 0.06, adj R² = 41.4%, τ² = 0.01, P = 0.01). Therefore, the methodological quality of included studies may be a possible source of heterogeneity.

From the nine studies that included controls, the pooled rate of depression was higher in patients with rosacea than in the control group (OR 2.21, 95% CI 1.79–2.72) (Fig. 2b). When recalculating the pooled estimate by omitting one study each time, sensitivity analysis showed that the ORs for depression remained significant, which indicated that the summary estimates were sufficiently robust to exclude studies with different assessment criteria (ESM Fig. S1a).

A total of seven studies examined the prevalence of anxiety in patients with rosacea (Fig. 3a, Table 2). The pooled random-effects prevalence of any anxiety was 15.6% (95% CI 11.8–19.3%). Heterogeneity across studies was evident (I² = 99.5%, P < 0.01). A further subgroup analysis revealed that the prevalence of anxiety was 12.1% (95% CI 9.6–14.7%) in the general populations, whereas in hospital-based populations, the prevalence was 19.2% (95% CI 9.9–28.6%). When stratified by different methods to assess anxiety, the prevalence of anxiety in patients determined using screening questionnaires was 22.7% (95% CI 12.9–32.5%), and the prevalence of anxoius patients when measuring outcomes by ICD codes was 10.2% (95% CI 5.2–15.3%); these parameters were significantly different (P = 0.03). All rosacea patients in the seven studies were diagnosed by physicians or on the basis of ICD codes; therefore, no relevant subgroups analyses were performed. Three studies were conducted in Asia, and the overall prevalence was 13.4% (95% CI 8.5–18.3%); three were conducted in Europe, with a pooled prevalence of 24.4% (95% CI 11.1–37.8%). Only one study reported the anxiety percentage of rosacea patients in North America: 6.5% (95% CI 6.2–6.8%). In addition, the percentage of women (CV − 0.41, adj. R² = 7.38%, τ² = 0.008, P = 0.28), mean age (CV − 0.003, adj. R² = − 14.53%, τ² = 0.005, P = 0.56) and NOS scores (CV 0.03, adj. R² = 19.99%, τ² = 0.0027, P = 0.22) did not moderate the rate of anxiety in rosacea patients.

Patients with rosacea were significantly more likely to have anxiety (6 out of 7 studies) with a pooled OR of 2.31 (95% CI 1.56–3.44; Fig. 3b). When recalculating the pooled estimate by omitting one study each time, the sensitivity analysis showed that the ORs for anxiety remained significant, which indicated that the summary estimates were sufficiently robust to exclude studies with different assessment criteria (ESM Fig. S1b).