We included observational studies with case–control and cross-sectional designs that described the association of HS and MS. The studies could be either prospective or retrospective, and the language of publication could be either English or Spanish. Searches encompassed the time from inception of the databases until January 2016.

Study participants could be either children or adults with HS. Control patients did not have HS. Both sexes were included. The studies had to assess the prevalence or incidence of MS, as defined by any of the following groups: the World Health Organization,11 the National Cholesterol Education Program Adult Treatment Panel III (ATP-III),12 the International Diabetes Federation,13 and the Joint Scientific Statement (of the American Heart Association among others).14 Each definition includes several clinical criteria based on physical examination as well as laboratory findings or patient-reported information in both the HS and control populations.

After making a concerted attempt to contact the authors of articles with missing information, we discarded any studies that still lacked crude or adjusted OR data. If studies included overlapping data, only the one with the larger number of cases was retained.

Searches were done in the following databases: MEDLINE, SCOPUS, SCIELO, Google Scholar, Science Direct, and LILACS. A search strategy appropriate for each database was used. We used Medical Subject Headings and the following free-text terms hidradenitis suppurativa (Spanish, hidradenitis supurativa) and metabolic syndrome (síndrome metabólico). (See online supplementary material, Appendix.) We also hand searched journals for relevant articles mentioned in reference lists.

Two researchers (M.J.M.R.-Z. and H.A.G.-P.) independently read titles and abstracts to assess each study's relevance. The selection criteria were then applied to full texts before articles were finally included. Any disagreement between the readers was resolved by consulting the third author (A.G.O.-L.). The 2 readers then extracted data independently, filling in a standardized questionnaire. Both checked all entries and revised the data sheets at least twice to ensure completeness and accuracy. The following information was extracted: author, year of publication, country/geographic location, study design, number and type of patient; mean age, prevalence or incidence of MS in HS and control groups; and crude and adjusted ORs.

All the included studies applied the ATP-III criteria to define MS.12 Thus, MS was diagnosed when a patient had at least 3 of the following conditions1: fasting glucose levels of 100mg/dL (or under treatment to lower blood sugar levels),2 blood pressure of at least 130/85mmHg (or on antihypertensive treatment),3 triglyceride concentration of 150mg/dL or more (or on treatment),4 high density lipoprotein cholesterol <40mg/dL in men or <50mg/dL in women (or on treatment),5 or a waist circumference over 102cm in men or 88cm in women.

The primary objective was to estimate the association between HS and MS in studies reporting ORs adjusted for confounders such as age, sex, and smoking. Four secondary objectives were also set: 1) to estimate the association between HS and MS based on reported crude ORs, 2) to estimate the association in subgroups, 3) to describe the characteristics of patients with HS (age, sex, smoking habit, etc.), and 4) to assess risk of bias in the included studies.

We used adjusted ORs reported in the studies to calculate the cumulative OR. This OR, reported with its 95% CI, was the outcome measure for the primary objective. Because we expected considerable heterogeneity, the cumulative OR was calculated using a random effects meta-analysis, which is the standard approach.15 Statistical heterogeneity was determined by the Q test, which is based on the χ2 test and the I2 statistic,16 which is considered high if over 50%.17

Descriptive data from clinical trials (percentages and means) were analyzed. All analyses and meta-analyses were carried out with Stata13 software (StataCorp). Some images were created with Review Manager, version 5.1 (Cochrane Collaboration).

We compared results according to design (cross-sectional vs case–control), data collection methods (prospective vs retrospective), geographic location (Europe vs other), MS diagnostic criteria (ATP-III vs other), and patient type (hospital vs another “general” source and adults vs children). Forest plots were used to report the effect sizes and 95% CIs. These subgroup analyses were based on adjusted ORs whenever possible or on crude ORs if no adjusted ORs were reported.

Publication bias was assessed by inspecting funnel plots of study size against standard error. A symmetrical plot indicates that publication bias is unlikely except when very small studies are included. (See online supplementary material, Appendix.)

We used the MINORS tool (Methodological Index for Non-Randomized Studies) to assess methodological quality.18 Two authors (M.J.M.R.-Z. and A.G.O.-L.) completed their assessments independently. Items were scored 0 if not reported, 1 if reported but inadequately, and 2 if reported adequately. The ideal overall score for a comparative study was 24, and we judged that the risk of bias was low if the score was at least 20. (See online supplementary material, Appendix.) When the 2 independent raters differed on any of the items, they attempted to reach agreement. If they failed, the third author (H.A.G.-P.) was consulted for an independent opinion.

The meta-analysis included studies with ORs adjusted for confounding factors as well as crude ORs. The 2 types of data were analyzed separately. The primary objective was met by analyzing only studies reporting adjusted ORs, in keeping with Cochrane recommendations.20 Our analysis showed that HS patients had approximately 80% higher risk of MS compared to control patients. This result was similar to the findings of an earlier meta-analysis, which reported a cumulative OR of 2.22 (95% CI, 1.62–3.06; I2 = 77%).7

Observational studies generally suffer from a higher likelihood of bias from confounding factors. Therefore, and because the observational studies we included had different designs, we preferred adjusted ORs in order to reduce effects of bias and heterogeneity.8

However, we did analyze crude ORs for our secondary objectives, following a method similar to that of an earlier meta-analysis.7 Our results with crude ORs similarly showed that HS patients had a 2-fold higher risk for MS. It is believed that the similarity between adjusted and crude ORs is attributable to the small number of published studies. However, the similarity does suggest there is an association between HS and MS.

The HS–MS association is related to a shared origin in chronic inflammation. As in psoriasis,21–23 chronic inflammation in HS leads to insulin resistance and, subsequently, endothelial cell dysfunction, promoting cardiovascular disease. Although there is no consensus about which cytokines and immune pathways are responsible for inflammation in HS, studies have demonstrated the presence of increased concentrations of interleukin (IL) 17, IL-1b, IL-10, tumor necrosis factor, and IL-23, which are considered the main cytokines involved.24,25

We also found a stronger association between HS and MS in hospital-treated patients versus “general” patients. These findings reflect diagnostic bias between these 2 groups. However, we believe that tertiary care hospital clinics are more likely to be treating HS patients with more severe symptoms, such as a larger number of lesions or highly inflamed lesions, which would have a greater impact on quality of life.

This indirect indicator of disease severity is reflected in studies reporting that patients with more severe forms of HS seek care in dermatology clinics, increasing the burden on the health care system.26,27 Hence, we believe this pattern explains the stronger HS–MS association in groups with more severe disease7,28 mediated by a higher degree of inflammation in severe HS.29 Hessametal30 found that HS severity correlated positively with higher concentrations of serum markers of inflammation such as C-reactive protein and neutrophils.

Each component of MS (arterial hypertension, obesity, dyslipidemia, and diabetes mellitus) will also show an association with HS. A higher body mass index has been linked to greater severity in HS symptoms,31,32 and a significant correlation between HS and obesity in contrast with overweight has been reported (OR, 4.4 vs 2.1).28 The prevalence of obesity in patients with HS (70%) was even higher than in patients with psoriasis (50%) in one study.33

Patients with HS were 2-fold more likely than control patients to be smokers. There is considerable epidemiologic evidence for a link between HS and smoking,32 and it is important to remember that smoking will contribute considerably to inflammation in MS, exacerbating and accelerating atherogenesis.34 Nonsmoking patients with HS have also been reported to improve with treatment more than smokers.35 Our findings support the notion that smoking addiction is a common concomitant condition among patients with HS. Thus, physicians should encourage these patients to quit smoking in order to improve both their HS symptoms and their metabolic profile.

HS symptoms, which usually appear after puberty and are less common during childhood,36 can nevertheless develop at young ages.37 Our analysis shows that there is a significant association between HS and MS in studies that include pediatric patients as well as adults. However, there is scarce literature on the relationship between obesity and MS in children with HS.

Deckers et al38 found that the body mass indexes of patients whose HS symptoms developed after the age of 30 years were similar to those whose symptoms appeared earlier. When Shalometal39 analyzed their results by age groups in a study comparing age- and sex-matched patients with HS, the prevalence of obesity was significantly higher in HS patients under the age of 20 years (20%) than in older patients (12%) (P=.005).

Finally, current guidelines state that the possibility of endocrine disorders and obesity should be borne in mind when treating young patients with HS.40 MS, which is a condition that includes diabetes mellitus, hypertension, and dyslipidemia among its components, does not have specific criteria for making the diagnosis in children.41 Studies specifically in pediatric cohorts should be carried out to confirm the association between HS and pediatric metabolic disorders.

The funnel plots we constructed suggest the presence of publication bias. (See online supplementary material, Appendix.) This detection of probable bias contrasts with an earlier meta-analysis.7 However, heterogeneity among the studies we included was low, as reflected in the consistency of their findings. Likewise, the risk of bias was low in a considerable number of the individual studies (Figs. 2A and 2B), supporting the validity of our findings. These results should nevertheless be interpreted cautiously.

In addition, we found few studies about HS and MS and were unable to obtain missing data from authors in spite of repeated attempts to reach them. Finally, potentially eligible studies may have been left out because we restricted our search to English and Spanish, leaving out such languages as German and Chinese.