Lifestyle factors and hand eczema: A systematic review and meta‐analysis of observational studies

Abstract Evidence regarding the association between lifestyle factors and hand eczema is limited.To extensively investigate the association between lifestyle factors (smoking, alcohol consumption, stress, physical activity, body mass index, diet, and sleep) and the prevalence, incidence, subtype, severity, and prognosis of hand eczema, a systematic review and meta‐analysis were conducted in accordance with the Meta‐analysis Of Observational Studies in Epidemiology consensus statement. MEDLINE, Embase, and Web of Science were searched up to October 2021. The (modified) Newcastle‐Ottawa Scale was used to judge risk of bias. Quality of the evidence was rated using the Grades of Recommendation, Assessment, Development and Evaluation approach. Eligibility and quality were blindly assessed by two independent investigators; disagreements were resolved by a third investigator. Data were pooled using a random‐effects model, and when insufficient for a meta‐analysis, evidence was narratively summarized. Fifty‐five studies were included. The meta‐analysis (17 studies) found very low quality evidence that smoking is associated with a higher prevalence of hand eczema (odds ratio 1.18, 95% confidence interval 1.09‐1.26). No convincing evidence of associations for the other lifestyle factors with hand eczema were found, mostly due to heterogeneity, conflicting results, and/or the limited number of studies per outcome.

skew the immune system toward T helper (Th) 2 immunity, which could, at least in theory, increase the occurrence, severity, and/or worsen the prognosis of HE. Likewise, obesity is associated with a chronic low-grade inflammatory state, which might also influence HE. 13 In addition, for smoking, also non-immunologic effects such as cutaneous vasoconstriction with delayed wound healing might play a role in the severity and/or prognosis of HE. 14 Following this hypothesis, it is possible that more inflammatory subtypes such as vesicular HE, or HE accompanied by AD, are more influenced by lifestyle factors compared to other subtypes. The aim of this systematic review and metaanalysis was to assess the association between lifestyle factors (including smoking, alcohol consumption, stress, physical activity, body mass index [BMI], diet, and sleep) and prevalence, incidence, subtype, severity, and prognosis of HE. Because lifestyle factors relate to human behavior and exposing subjects to unfavorable conditions might be harmful and unethical, this study focused on observational studies including subjects with all subtypes of HE.

| METHODS
This study was conducted in accordance with the Meta-analysis Of Observational Studies in Epidemiology (MOOSE) consensus statement. 15,16 A protocol was registered prospectively in PROSPERO (International Prospective Register of Systematic Reviews) (CRD42020207731).

| Data sources
A literature search was conducted in MEDLINE, Embase, and Web of Science from inception to October 14, 2021, supervised by an experienced research librarian. Search terms included all terms regarding HE and synonyms, combined with synonymous and related terms for lifestyle factors (see Appendix S1 in the supplement for the full search strategy).

| Study selection
After de-duplication, 17 all studies were uploaded in RAYYAN (http:// rayyan.qcri.or/) 18 for blinded and independent screening for eligibility based on title, abstract, and keywords by two investigators (LL, MJB).
Disagreements were resolved by reaching consensus or otherwise treated as a provisional inclusion awaiting full text. References of included studies and possibly relevant reviews were searched manually for additional studies. Broad inclusion criteria were applied, and all human studies that assessed the association between lifestyle factors and HE, regardless of the underlying etiology, were included. Excluded were studies without primary data, case reports, case series (n < 10), reviews, studies that only assessed the association between HE and second-hand smoking, tobacco allergy, skin exposure and food (substances), or the use of topical alcohol as disinfectant. We applied no language restrictions. Subsequently, full texts were retrieved of all the (provisionally) included studies. Studies published in languages other than English, Dutch, or German, were translated by colleagues with sufficient knowledge of the particular language. Abstracts from unpublished studies were also included if sufficient data were provided. Multiple papers from a single study were included if each presented unique data. Final inclusion was assessed independently by two authors (LL, MJB), and any disagreements were resolved by consulting a third author (BWMA).

| Data extraction
From each included study, first author name, publication year, country, study setting and design, number of total subjects, and number of subjects with HE were extracted. For study outcomes, assessment of HE, outcome ascertainment, instruments used, and primary study outcomes were recorded. A list of excluded studies based on full text, including justifications, was maintained. Authors were contacted in case of insufficient information or in case full text was not available.

| Assessment of risk of bias and overall quality of the evidence
We used the Newcastle-Ottawa Scale (NOS) for cohort and case-control studies for quality assessment. 19 An adapted version of the NOS was used for cross-sectional studies, as per Quaade et al. 2 Each study was assessed independently in pairs of two among three authors (LL, MJB, AALMP); a fourth author (BWMA) was asked to resolve differences. If studies reported results on multiple lifestyle factors and/or outcome measures, quality assessment was conducted per reported outcome measure. Cohort and case-control study outcomes with ≥6 points on the NOS, and cross-sectional study outcomes with ≥7 points, were considered as low risk of bias. To easily identify the NOS items that were low (one or two stars) or high risk of bias (zero stars), we used Cochrane's RevMan version 5.4.1. 20 for a visual presentation, as was done previously by Papola et al. 21 Of the three NOS scales used (case-control, cohort, and cross-sectional) we divided the item comparability in two (sex and age) so that these are also easily identifiable. For the metaanalyses the Grades of Recommendation, Assessment, Development and Evaluation (GRADE) 22 methodology was used to assess the overall quality of the evidence (high, moderate, low, or very low).

| Data synthesis
In case of missing summary statistics, odds ratios (ORs) were calculated from raw data, if possible. From the studies included in the meta-analyses, forest plots with estimated pooled ORs with 95% confidence intervals (CIs) were generated using Cochrane's RevMan version 5.4.1. 20 We chose the random-effects model for meta-analyses as the variability between studies was assumed to be high. Overlap of CIs or point estimates were used to judge inconsistency. To explore the possible causes of heterogeneity among study results, a subgroup analysis based on setting (occupational vs non-occupational) was performed. To assess the robustness of the synthesized results, a subanalysis was performed including only studies with an overall low risk of bias. If the number of studies allowed it, funnel plot visualization was used to inspect potential publication bias. When data were insufficient to conduct a meta-analysis, the evidence was narratively summarized.

| Literature search
The search identified 5686 records for screening after de-duplication, 17 of which 140 met the criteria for full-text extraction. Of these, 91 were excluded, based mainly on not reporting (separate) data for HE or lifestyle factors (see Table S2 in the Supplement for a list of excluded studies with justification). Six studies were included from references. See Figure 1 for the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 flowchart. 15

| Smoking
Fifty-six patients followed the diet. Thirty-six patients cleared or improved after 1 mo of dieting. Twenty-eight of them followed the prescribed diet rigorously or intermittently for at least a year, because they experienced recurrence of the dermatitis if they stopped. Six noted no long-term benefit and two were lost to follow-up.

3
Note: Studies occurring multiple times due to multiple studied lifestyle factors or outcome measures. When studies reported both unadjusted and adjusted results, adjusted results are reported. Cohort and case-control study outcomes with

≥6
points on the NOS, and cross-sectional study outcomes with ≥7 points, were considered low risk of bias.
Twelve studies could not be included in the meta-analysis due to lack of data (n = 8), 5,28,29,37,40,44,48,49 categorization of smoking habits other than yes/no (n = 3), [24][25][26] or outcome measures not convertible to ORs (n = 1). 35 Of these, four reported a positive association between smoking and prevalence of HE, 24,26,35,44 and eight reported no association between smoking and HE. 5,25,28,29,37,40,48,49 Two studies on smoking and HE incidence found no association. 51,52 Regarding the association between smoking and specific subtypes of HE, no consistent results were found. 44,[53][54][55][56][57] Of the seven studies reporting on smoking and severity of HE, results were contradictory, two studies reporting a positive association, 58,59 one a negative association, 39 and four no association. 5,50,56,60 Two studies reported a worse prognosis for patients who smoked, 61,62 and two studies reported no association between smoking and longstanding HE or persistence of HE. 29,57 A possible publication bias was detected for the outcome smoking and prevalence of HE, based on asymmetric funnel plots (Figures S7 and S8). In combination with the findings above, this warrants caution regarding generalization.

| Stress
Twelve studies reported the association between stress and HE on fourteen outcomes, of which seven 1,35,41,45,59,62,63 outcomes had an overall low risk of bias. Stress was assessed in different ways. Five studies analyzing stress and prevalence of HE found a positive association 1,35,41,45,63 and one reported no association. 50 In four studies stress was mentioned as an aggravating factor, 29,50,64,66 and four studies did not find an association between stress and severity of HE. 50,59,60,65 One retrospective cohort study reported more persistence of occupational HE in subjects with higher self-reported frequencies of stress. 62

| BMI
Twelve studies reported results on 13 outcomes of body mass index (BMI) and HE. Eight outcomes proved to have a low risk of bias. 35,43,44,59,[67][68][69] Five studies reported results on BMI and prevalence of HE, of which four reported a positive association between BMI ≥30 and HE. 35,41,43,45 One study found no association between BMI and prevalence of HE. 42

| Physical activity
Eleven studies reported results on 12 outcomes of physical activity and HE, of which seven 35,37,42,44,48,62,71 outcomes proved to have a low risk of bias. Eight studies included results on physical activity and prevalence of HE, of which four 5,37,48,71 found no association between physical activity and HE. One reported increasing HE frequencies with higher physical activity levels at work only in women. 70 Two reported a negative association between physical activity during leisure time and HE. 35,41 Another study reported varying outcomes in multiple categories of physical activity with various associations with HE. 42 One study found no association between physical activity and hyperkeratotic HE compared to other subtypes of HE. 44 Two other studies found no association between physical activity and severity of HE. 5,60 One clinical occupational retrospective cohort study found that moderate physical activity >4 hours weekly or regular strenuous exercise was associated with less-persistent HE. 62

| Alcohol consumption
Seven included studies reported on the association between alcohol consumption and HE, of which four 31,35,48,51 of the outcomes were assessed as having a low risk of bias. These studies used varying definitions of alcohol consumption. Six of them reported results on HE prevalence, of which five 28,31,42,48,73 did not find an association between alcohol consumption and the prevalence of HE. One study reported a negative association between alcohol consumption and HE. 35 No associations between alcohol consumption and incidence or severity of HE were reported. 51,60

| Diet
Three studies reported an association between diet and HE, of which one study 33 was of low risk of bias. That study did not find an association between a vegetarian/vegan diet and the prevalence of HE in upper-secondary school children. 33  Note: The Quality of evidence was accessed using the GRADE approach. Explanations: GRADE Working Group grades of evidence. High quality: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate quality: We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is markedly different. Low quality: Our confidence in the effect estimate is limited: The true effect may be markedly different from the estimate of the effect. Very low quality: Very little confidence in the effect estimate: The true effect is likely to be markedly different from the estimate of effect. Observational studies, such as the ones included in this systematic review, are assumed to have low quality and can be up-or downgraded based on the GRADE criteria. Abbreviations: CI, confidence interval; OR, odds ratio.
a Prevalence was either clinical diagnosed or self-reported.
b Downgraded for serious inconsistency: high heterogeneity. Clinically, observed in studies with participants from different settings (occupational, clinical, or general population) and outcomes (clinically confirmed or self-report); Statistical heterogeneity observed as studies with inconsistent point estimates and low extent of 95% CI overlap with the meta-analysis calculation.

| Sleep
No studies could be identified reporting the association between the amount of sleep and HE.

| Interpretation
Two previous studies, one meta-analysis and a systematic review, also reported on smoking and HE. 76,77 No association between smoking and the prevalence of HE was found in the meta-analysis. However, this conclusion was based on only three studies conducted in the same country. 76 On the other hand, the systematic review indicated that smoking might cause an increased prevalence and severity of HE, especially in high-risk occupations. 77 The exact mechanism behind the association between lifestyle factors and HE remains unknown. It has been described that lifestyle factors such as smoking 11 and stress 12 may influence the immune system toward Th2 immunity. In addition, obesity is associated with a chronic low-grade inflammatory state, which might also influence HE. 13

| Heterogeneity and confounders
In

| Strengths and limitations
This thorough and robust systematic review with meta-analysis gives a comprehensive overview of the existing literature on a possible associa- Due to the limited number of studies for each outcome, it was not possible to incorporate all these aspects into subgroup analyses.
Heterogeneity did not permit, besides for smoking, meta-analyses and assessment of pooled outcomes of major concern, and use GRADE for rating of the evidence. In addition, visual inspection of the funnel plots detected a possible risk of publication bias for smoking and HE. Furthermore, reporting bias could not be excluded.
The majority of the studies, which could not be included in the metaanalysis, due to lack of numerical data, reported no association between smoking and the occurrence of HE. If data were reported sufficiently it is not unimaginable that including those (in some cases large) studies would alter the overall estimated effect. Finally, information bias should also be taken into account, as self-reported lifestyle behavior might be prone to influence by socially desirable answers.
This extensive systematic review and meta-analysis found very low-quality evidence that smoking is associated with the preva-