Prevalence of depression and anxiety in systemic lupus erythematosus: a systematic review and meta-analysis

1Systemic lupus erythematosus (SLE) is a multisystem, autoimmune, connective-tissue disorder with frequent psychological comorbidities, of which depression and anxiety are two common manifestations. It has been reported that there were 2 times higher prevalence of depression in SLE patients compared to the general population. In addition, previous study has reported that the anxiety disorders were twice as prevalent among SLE patients as compared to the controls. Depression and anxiety often have profound impacts on SLE patients’ health and well-being including increased incidence of cardiovascular diseases, myocardial infarction, suicidal ideation, physical disability, decreased quality of life, and a higher risk of premature mortality. Therefore, depression and anxiety may be useful targets for interventions aimed at improving subjective health and quality of life in individuals with SLE. However, current epidemiological evidence found that the prevalence of depression and/or anxiety in SLE patients ranged widely from 2% to 91.7% in different studies. This vast inter-study difference was previously attributed to multiple factors, including study quality, unclear definition of depression or anxiety, diverse screening strategies used across studies. Reliable estimates of depression and anxiety prevalence are important for informing efforts to prevent, treat, and identify causes of depression and anxiety among SLE patients. Recent meta-analyses have estimated the overall prevalence of depression and/or anxiety in rheumatoid arthritis and osteoarthritis patients. There has only been one previous systematic review of psychiatric symptoms in SLE; however, no systematic review was conducted to quantify the prevalence of depression and anxiety in SLE using meta-analysis techniques. Our goal was to address this limitation. The objectives of this systematic review were (i) to establish pooled prevalence levels of depression and anxiety among adult SLE patients; (ii) to provide a summary of the methods used to define depression and anxiety in SLE; and (iii) to explore the impacts of study characteristics on prevalence estimates.

2This systematic review was conducted within the Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) statement and followed a predetermined registered protocol (PROSPERO: CRD42016044125).

3A systematic review of published literature in scientific journals that reported on the prevalence of depression and/or anxiety among SLE patients was conducted by two independent reviewers using the following databases from their inception to August 2016: PubMed, Embase, PsycINFO, and the Cochrane database library. The computer-based searches combined terms related to SLE patients and study design with those related to depression or anxiety (see Additional file 1). We conducted citation chasing search strategy with all reference lists of included articles and relevant review papers were considered to identify potentially omitted articles. Finally, we corresponded with the authors for further information if we encountered articles just provided the mean and standard deviation of the depression and/or anxiety assessment scale.

4Studies were included if they met the following criteria: (i) cross-sectional design, baseline cross-sectional data from a longitudinal study or baseline cross-sectional data from a trial, before group allocation; (ii) used validated methods (clinical interviews or self-report instruments) to assess depression or anxiety; and (iii) the sample size was no less than 60.

5Case reports, review articles, animal studies, studies investigating neuropsychiatric syndromes, studies in languages other than English and papers not dealing with SLE patients were excluded. For this meta-analysis, studies using pediatrics sample or screening tools without stating the cut-off thresholds used to detect depression or anxiety were also excluded. Table 2 and Table 3 presented a full list of the eligible methods of detecting depression and anxiety, alongside the numbers of articles utilizing each method and the number of participants assessed.

6Two researchers read the relative studies independently by the titles and abstracts to exclude the references which did not met the inclusion criteria. Then, they read full texts in the remaining studies as mentioned above, and determined whether these references included were final studies or not. When multiple publications spanned the years of longitudinal studies, baseline prevalence levels were reported. The following information was independently extracted from each article by other two trained investigators using a standardized form: year, country, mean disease duration, percentage of female participants, sample size, average age of participants, criteria for detection of depression and anxiety, and reported prevalence of depression and/or anxiety. If we encountered multiple publications from the same cohort, we used the data from the most recent or the paper reporting data from the largest number of participants. The methodological quality of each study included in the present meta-analysis was assessed using a modified version of the Newcastle-Ottawa Scale. Studies were judged to be at low risk of bias (≥3 points) or high risk of bias (<3 points). Any disagreements in data extraction and quality assessment were resolved through discussion between the two reviewers or adjudication with a third reviewer.

7The outcomes were major/minor depression and affective/dysthymic/adjustment/anxiety disorder diagnosed with a structured clinical assessment [e.g., Diagnostic and Statistical Manual of Mental Disorders (DSM)-IV or International Classification of Diseases (ICD)-10] or depression and/or anxiety assessed with validated assessment tools [e.g., the Hospital Anxiety and Depression Scale (HADS), the Centre for Epidemiologic Studies Depression Scale (CES-D)] (see Additional file 2).

8Because random-effects models tended to provide wider confidence intervals (CI) and were preferable in the presence of between-study heterogeneity, we used a random-effects meta-analysis to pool studies reporting the prevalence of depression and/or anxiety in SLE patients. Between-study heterogeneity was assessed by the I2 with thresholds of ≥25%, ≥50% and ≥75% indicating low, moderate and high heterogeneity, respectively. The influence of individual study on the overall prevalence estimate was explored by serially excluding each study in sensitivity analyses. Wherever possible, subgroup analyses were planned by overall study quality, sample size, country of origin and publication year, if there was more than one study in the subgroup. Pearson’s and Spearman’s correlation analyses were used to assess the association between variables and prevalence of depression and anxiety in people with SLE. Funnel plots and Egger’s test were combined to explore the potential publication bias in this meta-analysis. Statistical analyses were performed with STATA version 12.0. Statistical tests were 2-sided and used a significance threshold of P < 0.05.

9Fig. 1 provided the details of the study selection process. The initial search identified a total of 3347 potentially relevant articles. After removal of duplicates, titles and then abstracts were screened for potential eligibility. From this, 121 were considered in the full-text review, of which 59 articles met the inclusion criteria, and a full reference list was presented in Additional file 3. Inter-rater reliability of reviewers regarding study relevancy was high (Kappa = 0.87).

10A summary of the included study characteristics was shown in Table 1. A total of 59 identified studies matched the inclusion criteria, reporting on a total of 10828 adult SLE patients. Twenty took place in North America, 18 in Asia, 12 in Europe, 6 in South America, 1 in Oceania, and 1 in Africa. The median of mean ages was 39 years (range, 30.0-50.1), and the median percentage of females represented in the sample was 93% (range, 75%–100%). In addition, the median number of participants per study was 100 (range, 60–1827), and the median of mean disease duration was 9 years (range, 0.22–16.3). Depression was defined in 35 different ways (Table 2). Seventeen studies assessed for depression using the 21 Item-Beck Depression Inventory (BDI), with sixteen different thresholds were presented in the articles. Thirteen articles used the CES-D; six different cut-off points were presented, and the most commonly used being 16. Twelve used the HADS with a cutoff of 8 or more, and 6 used other screening tools. Ten studies assessed for major depression using diagnostic criteria (DSM or ICD). The most commonly used screening questionnaire to assess anxiety was the HADS, with 10 studies using this screening tool with thresholds of 8. The methods employed to assess depression and anxiety and the frequency of their use were presented in Table 2 and Table 3. When evaluated by Newcastle-Ottawa quality assessment criteria, out of 5 possible points, 2 studies received 5 points, 7 received 4 points, 13 received 3 points, 36 received 2 points, and 1 received 1 point. The details of the assessment of individual studies were shown in Additional file 4.

11Prevalence estimates of depression ranged from 2% to 91.7% in individual studies (Table 1). Table 2 indicated the summary of meta-analyses and heterogeneity assessments. Meta-analyses revealed the prevalence of major depressive disorder to be 24% (95% CI, 16%–31%) according to the DSM and/or ICD diagnostic criteria, with high heterogeneity (I2 = 95.2%). Prevalence estimates of depression were 30% (95% CI, 22%–38%, I2 = 91.6%) for the HADS with thresholds of 8 and 38% (95% CI, 32%–44%, I2 = 81.3%) for the CES-D with thresholds of 16, respectively. Prevalence of depression according to the 21 Item-BDI with a cutoff of 14 or more was 39% (95% CI, 29%–49%), with high heterogeneity (I2 = 88.2%) (Fig. 2).

12Prevalence of anxiety alone ranged between 4% and 85% in individual studies (Table 1). Table 3 presented the summary of meta-analyses and heterogeneity assessments. Meta-analyses pooled the prevalence of anxiety to be 40% (95% CI, 30%–49%, I2 = 93.0%) and 37% (95% CI, 12%–63%, I2 = 98.3%) according to the HADS with thresholds of 8 and the DSM and/or ICD diagnostic criteria, respectively (Fig. 3).

13Table 4 suggested depression and anxiety prevalence estimates according to each sensitivity and subgroup analysis, in comparison with the primary analysis. Sensitivity analyses revealed that the exclusion of studies with less sample representativeness tended to decrease dysthymic disorder prevalence estimates according to DSM and/or ICD. The removal of studies with less comparable respondent and non-respondent comparability tended to increase depression prevalence estimates according to the HADS with a cutoff of 8 or more. According to DSM and/or ICD, anxiety prevalence estimates had a trend to decrease by exclusion of studies only using female sample. The subgroup analyses were conducted according to sample size, overall quality, publication year, and country of origin. The results showed that studies with sample size <200 had higher anxiety estimates [43% (95% CI, 31%–55%) vs 28% (95% CI, 16%–40%)] according to the HADS with a cutoff of 8 or more. When evaluated by Newcastle-Ottawa criteria, studies with lower total overall quality scores yielded higher dysthymic disorder estimates [18% (95% CI, 6%–29%) vs 3% (95% CI, 2%–25%)] according to DSM and/or ICD. In contrast with clinical interviews (DSM and/or ICD), more recent publications tended to yield higher depression and anxiety prevalence estimates according to self-report instruments. The subgroup analyses for country of origin showed no clear patterns. There was no particular trend or pattern in any other sensitivity analyses or subgroup analyses.

14We used Pearson’s and Spearmen’s correlation analyses to assess the association between variables including mean/medium disease duration, proportion of female participants, mean/medium age, representativeness, sample size, comparability, overall quality, country of origin, publication year, and the prevalence of depression and anxiety. Table 5 indicated that more recent publications was significantly associated with increased depression prevalence (r = 0.26, P = 0.04). No study characteristics presented a significant association with anxiety prevalence estimate.

15Assessment of publication bias indicated significant publication bias, according to the Egger’s test, in studies reporting depression according to HADS with thresholds of 8 and CES-D with a cutoff of 16 or more [Egger: bias = 0.81 (95% CI: 0.04, 1.58), P = 0.04, and Egger: bias = 2.79 (95% CI: 0.61, 4.97), P = 0.02, respectively]. There was no significant evidence of publication bias in any other analyses (see Additional file 5).

16This systematic review and meta-analysis of 59 studies involving 10828 adult SLE patients demonstrated that a few studies using gold standard clinical interviews (DSM and/or ICD) reported that major depression and anxiety were presented in 24% and 37% among SLE patients, respectively. The majority of studies using screening tools found that significant depression were presented in 30% using the HADS a cutoff of 8 or more and 39% using the 21 Item-BDI with thresholds of 14. This study also found that more recent publications was significantly associated with increased depression prevalence among SLE patients. Furthermore, the prevalence of anxiety was 40% according to the HADS with thresholds of 8. These prevalence estimates are significantly higher than those observed in the general population and other rheumatic and connective tissue diseases. Furthermore, these findings demonstrated that SLE patients tended to have a higher prevalence of anxiety than depression, which was in line with previous studies. Such discrepancy could be explained by the differences in time frames when these studies were performed, disease characteristics, social and cultural contexts of the lupus patients and tools used for assessing depression or anxiety. Because the development of depression and/or anxiety could result in increased incidence of cardiovascular diseases, decreased quality of life, and a higher risk of premature mortality among SLE patients, these findings highlighted an important issue in health education for this population.

17Neuropsychiatric (NP) disorders appeared in about 70% of the patients diagnosed with SLE. Previous meta-analyses have assessed the prevalence of the 19 NP syndromes defined by the American College of Rheumatology (ACR) in 1999 among SLE patients. However, there were a wide variety of neurologic and psychiatric manifestations of SLE, which extended beyond those identified in the 1999 ACR classification criteria for SLE. Several attempts have been made to devise a classification of NP-SLE manifestations because there were controversies regarding the inclusion of mood disorders in the 1999 ACR NP-SLE criteria. That’s why we excluded the studies investigating neuropsychiatric syndromes among SLE patients in this meta-analysis.

18Although studies varied widely in terms of quality, our sensitivity analyses suggested that depression and/or anxiety prevalence estimates (except dysthymic disorder estimates) were reasonably stable. Variation in study sample size contributed importantly to the observed heterogeneity in the data. Studies with sample size <200 had higher anxiety estimates according to the HADS with thresholds of 8. Furthermore, studies with lower total overall quality scores yielded higher dysthymic disorder estimates according to DSM and/or ICD. Country, publication year, age, and gender also contributed to the heterogeneity between studies.

19In this meta-analysis, many methods were used for data extraction and synthesis. The gold standard method was diagnostic interviews using DSM or ICD criteria, which were often time consuming and expensive. Therefore, it was not ideal for examining patients in a busy hospital environment. Alternatively, self-report screening tools might be used, because they were quick and easy to complete and cheaper to use than diagnostic interviews. However, prevalence estimates using screening tools were often overestimated, because such tools tended to prioritize sensitivity over specificity. Furthermore, there have not been validation studies to determine the best cut-point for screening tools in SLE patients, and several cut-off scores on self-report tools were often used in many studies. It indicated that the rheumatologists should always report prevalence at conventional cut-points, and screen for depression and anxiety among SLE patients according to the social and cultural contexts of the rheumatologists and SLE patients in clinical practice.

20There are, however, additional important shortcomings in the evidence on prevalence of depression in SLE that need to be addressed. First, a substantial amount of the heterogeneity among the studies remained unexplained by the variables examined. Unexamined factors, such as gender, age, disease duration, might contribute to the risk for depression and/or anxiety symptom among SLE patients. Second, the data were derived from studies that used different designs and involved different groups of patients (e.g., from different countries), which might result in heterogeneity among the studies. Third, we did not look for healthy subjects in each study reporting the prevalence of depression or anxiety in SLE patients, which should be addressed in future research.

21The prevalence of depression and anxiety was high in adult SLE patients. It indicated that rheumatologists should screen for depression and anxiety in their patients, and they should refer them to mental health providers in order to identify effective strategies for preventing and treating depression and anxiety among SLE patients.