Odds ratios (ORs) for the binary outcomes of the metabolic syndrome and its component outcomes (diabetes mellitus, hypertension, increased triglyceride [TG] levels and decreased high-density lipoprotein cholesterol [HDL] levels, and general and abdominal obesity). The ORs are adjusted for age, sex, and smoking status. Whiskers indicate 95% confidence intervals.
Ratio of means (RMs) for continuous outcomes of the components of the metabolic syndrome (nonfasting blood glucose level, hemoglobin A1c [HbA1c] level, diastolic and systolic blood pressure, triglyceride [TG] blood level, high-density lipoprotein cholesterol [HDL] level, low-density lipoprotein cholesterol [LDL] level, total cholesterol [TC] level, general obesity, and abdominal obesity). The RMs are adjusted for age, sex, and smoking status. Whiskers indicate 95% confidence intervals. BMI indicates body mass index (calculated as weight in kilograms divided by height in meters squared); IFCC, International Federation of Clinical Chemistry; WC, waist circumference.
eTable 1. Table of crude ORs and age-, sex-, and smoking status–adjusted ORs
eTable 2. Crude means and age-, sex-, and smoking status–adjusted means and ratio of means (RM)
eTable 3. ORs adjusted for potential confounders
eTable 4. ORs and P values assessing the severity of HS among HS subjects only
eTable 5. Missing values
Miller IM, Ellervik C, Vinding GR, Zarchi K, Ibler KS, Knudsen KM, Jemec GBE. Association of Metabolic Syndrome and Hidradenitis Suppurativa. JAMA Dermatol. 2014;150(12):1273-1280. doi:10.1001/jamadermatol.2014.1165
An association between the metabolic syndrome (MetS) and chronic inflammatory diseases, such as psoriasis or rheumatoid arthritis, has been suggested. Hidradenitis suppurativa (HS), a more localized chronic inflammation of the skin, has been speculated to have a similar association. Hidradenitis suppurativa is a substantial burden for the individual and a socioeconomic burden globally. Information about the burden of possible comorbidities is scarce.
To investigate the possibility of an association between HS and MetS.
Design, Setting, and Participants
Cross-sectional population- and hospital-based study of HS and MetS. We identified 32 patients with physician-verified HS from the outpatient clinic at the Department of Dermatology, Roskilde Hospital, and 326 patients with HS and 14 851 individuals without HS from the general population. Individuals with HS were younger, predominantly female, and more often smokers compared with the non-HS group.
Main Outcomes and Measures
Metabolic syndrome and its components of diabetes mellitus, hypertension, dyslipidemia, and obesity.
When compared with the non-HS group, the odds ratios (ORs) for the hospital HS and population HS groups were 3.89 (95% CI, 1.90-7.98) and 2.08 (95% CI, 1.61-2.69), respectively, for MetS; 5.74 (95% CI, 1.91-17.24) and 2.44 (95% CI, 1.55-3.83), respectively, for diabetes mellitus; 6.38 (95% CI, 2.99-13.62) and 2.56 (95% CI, 2.00-3.28), respectively, for general obesity; and 3.62 (95% CI, 1.73-7.60) and 2.24 (95% CI, 1.78-2.82), respectively, for abdominal obesity. With regard to dyslipidemia, significant results were found for decreased levels of high-density lipoprotein cholesterol, with ORs of 2.97 (95% CI, 1.45-6.08) and 1.94 (95% CI, 1.52-2.48) for the hospital HS and general population HS groups, respectively, when compared with the non-HS group. With regard to increased triglyceride levels, only the result for the population HS group compared with the non-HS group was significant, with an OR of 1.49 (95% CI, 1.18-1.87). The OR for hypertension, which was only significant for the hospital HS group compared with the non-HS group, was 2.14 (95% CI, 1.01-4.53). Obesity and inflammation acted as possible confounders. The ORs were higher for the hospital HS group compared with the population HS group. The association between HS and MetS was not influenced by the degree of HS severity.
Conclusions and Relevance
As with more systemic inflammatory diseases, HS appears to be associated with MetS, indicating substantial comorbidities. Because this study is cross-sectional, causality remains to be explored.
Hidradenitis suppurativa (HS) is a chronic, localized inflammatory skin disease producing inflamed nodules in apocrine gland–bearing skin.1,2 The treatment of HS is often inadequate, and the disease inflicts a significant burden on patients, in whom pain and suppuration from lesions often lead to inactivity, depression, and significantly impaired quality of life.3,4 Changes in patients’ socioprofessional activity indicate the presence of important disease-related impairment.5,6
Hidradenitis suppurativa may be more common than hitherto suggested. The disease is often misdiagnosed or underdiagnosed, and prevalence estimates therefore range from 0.05% to 4%.7- 10 An increasing incidence during the last 20 years has been suggested.11 The comorbidities of HS are poorly described, and the available data are solely hospital based.12,13
The pandemic cluster of cardiovascular risk factors called the metabolic syndrome (MetS) co-occurs more commonly with chronic inflammatory diseases, such as rheumatoid arthritis and psoriasis.14- 18 In contrast to psoriasis, HS may be considered a more localized inflammation of the skin. An association between HS and MetS has been hypothesized to exist and cause significant comorbidity, negatively influencing the overall burden of disease. We therefore investigated a possible association using population- and hospital-based data in a cross-sectional study to explore the association and assess its possible clinical relevance.
This study was accepted by the ethics committee of the Zealand region (project numbers SJ-191, SJ-113, and SJ-114) in Denmark. Written informed consent was obtained from all study participants.
We performed a cross-sectional study of the association of HS (referred to as the exposure) and MetS (referred to as the outcome). We investigated 2 different groups of individuals with HS. The first group was identified in a general population sample (population HS group); the second was identified in a hospital-based sample (hospital HS group).
The population HS group was identified in the Danish General Suburban Population Study (GESUS). GESUS was initiated in January 2010 with ongoing enrollment and is a cross-sectional study of the adult Danish suburban general population in the Næstved municipality (70 km south of Copenhagen).19 All citizens 30 years and older and a random selection of those aged 20 to 30 years were invited to participate. The population HS group was defined based on whether participants reported boils within the previous 6 months and a minimum of 2 boils (in the following 5 possible locations: axillae, groin, genitals, mammae, or other [eg, perianal, neck, or abdomen]) on a questionnaire. The validation of this HS diagnosis is discussed in a separate report5 showing a sensitivity of 90% and a specificity of 97%. The overall participation rate in GESUS was 49.3%. Further details about GESUS can be found in Bergholdt et al.19
The hospital HS group was recruited from the outpatient clinic at the Department of Dermatology at Roskilde Hospital (serving the region of Zealand, which includes Næstved). Inclusion criteria consisted of the diagnostic code for HS from the International Classification of Diseases, 10th Revision (L73.2) and systemic or laser treatment for HS, which indicated moderate or severe disease. The diagnosis of HS was confirmed by results of a physical examination by a physician from the Department of Dermatology (I.M.M., G.R.V., K.Z., or K.S.I.). Eligible patients were invited to undergo the same examination as the population sample. The participation rate was 33 of 98 eligible patients (33.7%). The distribution of age and sex did not differ between participants and nonparticipants (data not shown).
Non-HS participants were defined as participants from GESUS without HS. This population constituted the non-HS group for this study.
Both HS groups and the population-based non-HS group completed the same questionnaire and physical examinations and contributed blood samples. The definitions of the MetS outcome were based on the methods of GESUS involving the GESUS questionnaire (self-reporting), physical examination, and laboratory evaluation of nonfasting venous blood samples.19
The MetS involves the following 4 key components: diabetes mellitus, hypertension, dyslipidemia, and obesity. The methods of this study used to define the outcome are listed in Box 1 and Box 2.20- 24 We used a modified version of the criteria of the National Cholesterol Education Program Adult Treatment Panel III,20 which also accounts for the harmonized definition of MetS.
Metabolic syndrome according to a modified version of the US NCEP ATP III includes ≥3 of the following criteria:
WC of >102 cm for men and >88 cm for women based on the physical examination findings
Decreased plasma HDL levels of <40 mg/dL for men and <50 mg/dL for women based on analysis of the blood sample
Increased plasma TG levels of ≥150 mg/dL, based on analysis of the blood sample
Blood pressure ≥130/85 mm Hg at the physical examination or self-reported use of antihypertensive
HbA1c level of ≥48 mmol/mol (IFCC) (≥6.5% [DCCT]), nonfasting plasma glucose level of ≥220 mg/dL, or self-reported diagnosis of diabetes mellitus
Abbreviations: DCCT, Diabetes Control and Complications Trials; HbA1c, hemoglobin A1c; HDL, high-density lipoprotein cholesterol; IFCC, International Federation of Clinical Chemistry; NCEP ATP III, National Cholesterol Education Program Adult Treatment Panel III; TG, triglycerides; WC, waist circumference.
SI conversion factors: To convert cholesterol to millimoles per liter, multiply by 0.0259; glucose to millimoles per liter, multiply by 0.0555; and TG to millimoles per liter, multiply by 0.0113.
Self-reported diagnosis or based on analysis of the blood sample: HbA1c level of ≥48 mmol/mol (IFCC) (≥6.5% DCCT) or nonfasting plasma glucose level of ≥220 mg/dL
Self-reported use of insulin or noninsulin antidiabetic
Quantification of HbA1c level (mmol/mol) and nonfasting plasma glucose level based on analysis of the blood sample
Self-reported use of antihypertensives or blood pressure ≥140/90 mm Hg according to definition of hypertension
Quantification based on measurements of systolic and diastolic blood pressure at physical examination
Increased TG levels according to the US NCEP ATP III criteria of ≥150 mg/dL
Decreased level of HDL according to the US NCEP ATP III criteria of <40 mg/dL for men and <50 mg/dL for women based on analysis of the blood sample
Quantification by measurement of TG, HDL, and TC levels, based on analysis of blood sample
General obesity defined as BMI of ≥30, based on physical examination findings
Abdominal obesity defined as WC according to NCEP ATP III criteria of >102 cm for men and >88 cm for women, based on physical examination findings
Quantification of obesity based on BMI and WC, based on physical examination findings
Abbreviations: BMI, body mass index (calculated as weight in kilograms divided by height in meters squared); DCCT, Diabetes Control and Complications Trial; HbA1c, hemoglobin A1c; HDL, high-density lipoprotein cholesterol; IFCC, International Federation of Clinical Chemistry; NCEP ATP III, National Cholesterol Education Program Adult Treatment Panel III; TC, total cholesterol; TG, triglycerides; WC, waist circumference.
SI conversion factors: To convert cholesterol to millimoles per liter, multiply by 0.0259; glucose to millimoles per liter, multiply by 0.0555; TG to millimoles per liter, multiply by 0.0113.
We considered the following possible confounders:
General obesity defined by body mass index (BMI; calculated as weight in kilograms divided by height in meters squared) measured at the physical examination;
Inflammatory load defined by high-sensitivity C-reactive-protein level measured in venous blood samples;
Self-reported level of physical activity at work and in leisure time;
Self-reported intake of atherogenic (ie, saturated) fat, fish, fruit/vegetables, eggs, and alcohol.
The definition of severity of HS for the population HS group was based on self-reported information on numbers and locations of boils and subsequent scarring and was inspired by the Hurley score, which is considered almost static.25 Mild HS was defined as a minimum of 2 boils and no subsequent scarring; moderate HS, a minimum of 2 boils and subsequent scarring; and severe HS, a minimum of 2 boils in a minimum of 2 locations and subsequent scarring. The severity of HS for the hospital HS group was assessed by the Sartorius score based on results of the physical examination.25 We also explored the severity of HS in the population and hospital HS groups as the number of boils to constitute a more dynamic scale.
We compared the HS and non-HS groups using logistic regression adjusting for age, sex, and smoking status, yielding an adjusted odds ratio (OR) for binary outcomes, and linear regression of log-transformed outcomes adjusting for age, sex, and smoking status, yielding an adjusted ratio of means (RM) for continuous outcomes. The binary effects measure (OR) expresses whether an association exists or not and the strength of that association. The continuous effect measure (RM) expresses the quantification of the association and can more easily be translated into clinical management than an OR because RM expresses the difference between the mean values of the HS vs non-HS groups. Thus, the RM provides the expected higher value of an individual with HS compared with a non-HS individual (eg, RM of 1.21 for BMI means that individuals with HS are expected to have 21% higher BMI than those without HS). In comparison, the OR is based on cutoff values and provides the expected higher odds (eg, OR of 2.00 for a BMI > 30 means that individuals with HS have 2 times higher odds of having a BMI > 30). P < .05 was considered to be statistically significant.
The ORs and RMs for the population HS group are expressed as ORpop and RMpop. Similarly, ORs and RMs for the hospital HS group are expressed as ORhos and RMhos.
We examined the influence of possible confounders on the association between HS and MetS by including these in the regression model and assessing the effect on the OR. We explored the relationship between the severity of HS and MetS using the same regression method as above analyzing only the HS groups. All statistical analyses were performed using commercially available software (SAS, version 9.3; SAS, Inc).
Owing to differences in background factors between exposed and unexposed groups and knowledge from previous studies that age, sex, and smoking status act as confounders in cardiovascular risk, the effect measures were adjusted accordingly. In eTables 1 and 2 in the Supplement, crude ORs and RMs are displayed.
The data used from GESUS were collected from January 1, 2010, through August 2, 2012. A total of 32 individuals with HS from the hospital, 326 with HS from the general population, and 14 851 non-HS individuals from the general population were identified. The background factors and characteristics of the HS and non-HS groups showed that individuals with HS were predominately younger, female, and smokers (Table).
We investigated the association of HS with the conditions involved in the MetS—diabetes mellitus, hypertension, dyslipidemia, and obesity—and with MetS defined by the criteria of the National Cholesterol Education Program Adult Treatment Panel III.20 The methods used to define the outcome are described in Box 1 and Box 2. The ORs and RMs adjusted for age, sex, and smoking status are illustrated in Figure 1 and Figure 2.
When adjusting for age, sex, and smoking status, the association between HS and MetS was significant for the hospital and population HS groups (ORhos, 3.89 [95% CI, 1.90-7.98]; ORpop, 2.08 [95% CI, 1.61-2.69]). Findings for the associations of MetS components with HS are discussed individually. We found a uniform pattern of the hospital HS group having higher ORs than the population HS group.
When adjusting for age, sex, and smoking status, a significant association between HS and diabetes mellitus was found for the hospital and population HS groups compared with the non-HS group (ORhos, 5.74 [95% CI, 1.91-17.24]; ORpop, 2.44 [95% CI, 1.55-3.83]). When we explored the type of antidiabetic used (insulin or noninsulin), only the association of noninsulin drugs was significant (ORhos, 7.93 [95% CI, 1.75-36.03]; ORpop, 3.50 [95% CI, 2.05-5.98]). When quantifying the association, only the RMhos was significant, with levels 8% (95% CI, 2%-15%) and 10% (95% CI, 5%-15%) higher for the HS groups compared with the non-HS group with regard to glucose and hemoglobin A1c levels, respectively.
When adjusting for age, sex, and smoking status, a significant association between HS and hypertension was found only for the hospital HS group (ORhos, 2.14 [95% CI, 1.01-4.53]). When quantifying this association, only the RMhos for diastolic blood pressure was significant, with a level 5% (95% CI, 1%-10%) higher in the HS hospital group when compared with the non-HS group.
When adjusting for age, sex, and smoking status, an association between HS and high triglyceride (TG) levels was found, but the association was statistically significant for the population HS group only (ORpop, 1.49 [95% CI, 1.18-1.87]). A significant association between HS and low levels of high-density lipoprotein cholesterol (HDL) was found (ORhos, 2.97 [95% CI, 1.45-6.08]; ORpop, 1.94 [95% CI, 1.52-2.48]). When quantifying the association, the RMpop for TG level was 21% (95% CI, 0%-46%) and the RMhos was 11% (95% CI, 4%-17%), significantly higher than in the non-HS group. The RMhos for HDL level was 14% (95% CI, 5%-22%) and the RMpop was 10% (95% CI, 7%-13%), significantly lower than in the non-HS group.
When adjusting for age, sex, and smoking status, we found a significant association between general obesity by BMI and abdominal obesity by waist circumference (WC) for the population and hospital HS groups. With regard to general obesity, the ORhos was 6.38 (95% CI, 2.99-13.62) and the ORpop was 2.56 (95% CI, 2.00-3.28). With regards to abdominal obesity, the ORhos was 3.62 (95% CI, 1.73-7.60) and the ORpop was 2.24 (95% CI, 1.78-2.82).
When we quantified the association, the RMhos for BMI was 21% (95% CI, 14%-29%) and the RMpop was 9% (95% CI, 7%-11%), significantly higher for the HS groups than the non-HS group. The RMhos for WC was 14% (95% CI, 9%-19%) and the RMpop was 7% (95% CI, 6%-9%), significantly higher for the HS groups than the non-HS group.
When exploring the possible confounders, we found that adjusting for obesity or inflammatory load reduced the strengths of the associations; however, the associations remained. Thus, obesity and inflammation were identified confounders, whereas level of physical activity and diet were not (eTable 3 in the Supplement). The association between HS and MetS or between HS and the MetS components was not influenced by the degree of HS severity with the exception of general obesity (eTable 4 in the Supplement).
This broad population- and hospital-based study suggests an association between HS and MetS and the individual MetS components of diabetes mellitus, low levels of HDL, and general and abdominal obesity. Positive associations were also found with regard to hypertension and dyslipidemia. Hypertension, however, was only statistically significant in the hospital HS group. When we used binary data (with acknowledged cutoff values) for increased levels of TG, only the association for the population HS group was statistically significant. However, statistically significant differences in the TG levels between the hospital HS group and the non-HS group were found when we examined the continuous data.
The hospital HS group had uniformly higher ORs than the population HS group. This result could indicate that differences in HS severity or the recent suggestion of different HS subtypes may play a part.26 This difference might also be an expression of a dilution of the population-based HS sample due to misclassification bias. The stronger ORs for the hospital HS group may also indicate detection bias; that is, HS patients within the hospital system are more likely to have been diagnosed with the outcome, which influences the self-reported diagnoses. However, detection bias was minimized by the inclusion of self-reported diagnosis and results of the physical examination and laboratory analysis of blood samples.
When we examined the MetS components individually, the ORs for diabetes mellitus indicated a positive association, mainly due to type 2 diabetes mellitus. Quantification analysis demonstrated that nonfasting glucose level was 1% to 8% higher, and hemoglobin A1c level was 1% to 10% higher in the HS groups compared with the non-HS group. Previous studies have suggested that an approximately 1% reduction in hemoglobin A1c level may decrease the risk for myocardial infarction by 14% to 16%, indicating that the abnormalities seen in the HS patients have clinical significance.27 Further indirect support for this association is provided by the observation that metformin hydrochloride treatment may ameliorate HS.28
Furthermore, the ORs for general and abdominal obesity were significant. When quantified, BMI was 9% to 21% greater, and WC was 7% to 14% larger in the HS groups than the non-HS group. Trimming of the WC by 4.4 cm from 102 cm (ie, a 4.3% reduction) may reduce the risk for diabetes mellitus by 58%, similarly indicating clinical relevance and a substantial potential for prevention in this generally neglected disease.29
The association of an atherogenic lipid profile (ie, increased TG levels and decreased HDL levels) was significant with regard to the decrease in HDL levels in both HS groups, but only significant with regard to the increase in TG levels in the population HS group. When quantified, the HS groups had an 11% to 21% higher TG level and a 10% to 14% lower HDL level than the non-HS group. A follow-up study of the effects of statins in 17 802 healthy individuals showed a significant reduction of cardiovascular events when TG levels were reduced by 17% and HDL levels were increased by 4%.30
Having hypertension was only significant with regard to the hospital HS group. When quantifying the hospital HS group, only the diastolic blood pressure was significantly increased by 5%. A recent meta-analysis of preventive treatment of hypertension suggested that lowering diastolic blood pressure from 90 to 85 mm Hg (ie, a 6% reduction) would reduce the risk for cardiovascular heart disease and stroke by approximately 20% and 30%, respectively.31
Our findings are in concordance with and expand the findings of 2 previous hospital-based studies.12,13 In aggregate, these data therefore suggest that the comorbidities of HS are clinically significant and that an increased clinical awareness of the HS diagnosis and its comorbidities might be warranted in this potentially substantial group of patients.
Obesity and inflammatory load were identified as possible confounders, partly but not exclusively explaining the associations and indicating a complex and overlapping relationship. Surprisingly, we found that physical activity level, diet and alcohol consumption, and the severity of HS did not influence the associations. The latter is in contrast to our supposition that the hospital HS group had higher ORs because of more severe disease, implying detection bias as previously discussed or an insufficiently sensitive measure of disease severity. Therefore, one can speculate that HS subtypes may influence the association with MetS more strongly than HS severity.
The association of MetS has also been shown in the chronic generalized inflammatory disease rheumatoid arthritis and the skin disease psoriasis, for which the ORs for MetS are approximately 2.00 compared with almost 6.00 in HS.8 Furthermore, the associations in psoriasis have been suggested to be significant only with regard to a hospital-based HS cohort,8 implying that the burden of these comorbidities is greater for HS than for psoriasis. Therefore, the common etiological factors between inflammatory skin disease and MetS are more readily identified in HS than in psoriasis.
The major strengths of our study are the large population-based HS group and the inclusion of individuals with HS from hospital- and population-based groups. The broad recruitment reduced selection bias with a broader range of disease severities and thereby aided the generalization of the results. To explore misclassification bias of individuals with HS, we validated the HS definition with a sensitivity of 90% and a specificity of 97%.9 Because the self-reported questions used to identify HS patients refer to symptoms (ie, boils) rather than the actual diagnosis (ie, do you have HS?), we strove to optimize the inclusion of individuals with undiagnosed and misdiagnosed HS in the population, which is particularly pertinent for underdiagnosed diseases. The diagnosis of HS among hospitalized participants was physician verified. The combined methods of self-reporting, laboratory analysis of blood samples, and physical examination aimed to reduce false-negative findings with regard to the outcome (MetS). Finally, essential possible confounders were recognized and explored systematically.
Potential limitations merit consideration. First, one must recognize that, because this study is cross-sectional, we cannot prove causality between HS and MetS. Furthermore, the population is suburban, most of the participants were white, and the group aged 20 to 30 years was underrepresented, which may limit the generalizability of our findings. In addition, an age bias was found between the HS and non-HS groups. However, we accommodated this age bias by age adjustments. The low participation rate of the hospital HS group increased variation and reduced power. The low participation rate may be an expression of the limited resources of HS patients due to the physical and mental burden of the disease. Nonfasting blood samples were used. However, differences in fasting and nonfasting lipid levels have been shown to be minimal.32 Furthermore, we did not include information on HS medical treatment, which might confound the results. Last, as with any questionnaire survey, the risk for recall bias was present.
The data suggest an association between HS and the MetS in the hospital and population HS groups. This allegedly increased disease burden due to comorbidities indicates that HS patients require general medical attention beyond the skin.
Future longitudinal studies with similar methods are needed to explore the temporal relationship of these associations. These studies should be large, include individuals with HS from the general population and hospital, and explore the differences between these groups and additional possible confounders.
Accepted for Publication: May 1, 2014.
Corresponding Author: Iben Marie Miller, MD, Department of Dermatology, Roskilde Hospital, Køgevej 7-13, 4000 Roskilde, Denmark (email@example.com).
Published Online: September 17, 2014. doi:10.1001/jamadermatol.2014.1165.
Author Contributions: Drs Miller and Jemec had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Miller, Ellervik, Vinding, Ibler, Jemec.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Miller, Ibler, Jemec.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Miller, Knudsen.
Obtained funding: Miller, Ellervik, Jemec.
Administrative, technical, or material support: Miller, Vinding, Ibler, Jemec.
Study supervision: Ellervik, Ibler, Jemec.
Conflict of Interest Disclosures: Dr Jemec has received consulting fees from Abbott Laboratories, AstraZeneca, MSD, Novartis, Pfizer, and Dumex-Alpharma; lecture fees from Abbott Laboratories, Galderma, Pfizer, and Roche; grant support from Abbott Laboratories, Pfizer, Photocure, and LEO Pharma; equipment on loan from Michelson Diagnostics; and reimbursement for travel expenses from Abbott Laboratories, Galderma, and Photocure. No other disclosures were reported.
Funding/Support: This study was supported by a grant under the Industrial PhD Programme and grants from the Danish Agency for Science, Technology, and Innovation and LEO Pharma (Dr Miller) and by the Jacob Madsen & Olga Madsens Foundation. The GESUS study was supported by the Johan and Lise Boserup Foundation, Trygfonden,Det Kommunale Momsfond, Johannes Fogs Foundation, Region Zealand Foundation, Næstved Hospital, Næstved Hospital Foundation, and the National Board of Health.
Role of the Funder/Sponsor: Dr Knudsen, who is an employee of LEO Pharma, which provided part of the grant supporting this study, was involved in analysis and interpretation of data and manuscript revision. The funding organizations had no other role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Additional Contributions: Claus Bay, MSc Stat, Jørgen Iversen, BSc, and Torsten Skov, MD, PhD, Department of Biostatistics, LEO Pharma, provided statistical, technical, and epidemiological support. Birgitte Schnack Nielsen, clinical nutritionist, Department of Pediatrics, Roskilde Hospital, Denmark, provided information on diet. The contributors did not receive compensation.