Cohort studies on 71 outcomes among people with atopic eczema in UK primary care data

Eczema (atopic eczema/atopic dermatitis) is one of the most common chronic conditions globally and is associated with substantial morbidity and healthcare costs. Beyond established links with atopic diseases (e.g., allergies and asthma), eczema may be associated with a range of non-atopic conditions through mechanisms such as chronic systemic inflammation, psychological stress, low self-esteem, and sleep disturbance. The American Academy of Dermatology (AAD) 2022 guidelines reviewed studies across 32 adverse outcomes, finding clear evidence for associations with other atopic conditions but mixed or limited evidence for many other outcomes (e.g., mental illness, cardiovascular disease, metabolic disease, osteoporosis and fractures). This study aims to comprehensively evaluate and compare associations between eczema and 71 adverse health outcomes using large-scale UK primary care data, providing evidence on strength of association, absolute risk differences, sensitivity analyses, and variation by eczema severity.

Guidelines from the American Academy of Dermatology (2022) synthesized evidence on 32 adverse health outcomes in adults with eczema, confirming strong associations with atopic/allergic conditions while indicating less certain evidence for many other comorbidities (mental, cardiovascular, metabolic, bone). Prior population-based studies in UK CPRD data have examined specific outcomes such as anxiety, cardiovascular events, renal failure, fractures, and cancers, generally finding modest associations and little to no association with most solid organ cancers. Benchmarking in this study showed adjusted hazard ratios closely aligned with previous CPRD GOLD studies across anxiety (aHR ≈1.16–1.17), cardiovascular outcomes (e.g., myocardial infarction ≈1.06–1.09; stroke ≈1.09–1.10), fractures (e.g., hip fracture ≈1.05–1.13), and cancers (no association with solid organ cancers like lung, breast, prostate; positive associations with nonmelanoma skin cancer and lymphomas). This context supports the validity and generalizability of the current cross-outcome findings while highlighting gaps the present study addresses by assessing a broader outcome set and severity gradients.

Design: Matched cohort study using de-identified UK primary care electronic health records (CPRD Aurum), April 1, 1997 to March 31, 2023. Approvals obtained from LSHTM REC (29781) and CPRD ISAC (23.002665). Exposure definition: Eczema identified via a validated algorithm requiring ≥1 diagnostic code and ≥2 eczema therapy records (emollients, topical glucocorticoids, topical calcineurin inhibitors, or systemic immunosuppressants) on two separate days. Positive predictive value previously reported at ~86% in UK primary care data. Cohorts: Multiple age-defined cohorts were created: any-age, 18+, 40+, a more-severe eczema cohort, and a <18 cohort. For 18+ and 40+ cohorts, exposure could be identified before age eligibility so both incident and prevalent eczema cases were included. Matching and index date: Eczema-exposed individuals were matched without replacement to up to 5 unexposed comparators on age (2-year caliper), sex, and general practice, in calendar date order. Index date was the latest of: meeting eczema definition; 12 months after practice registration; study start (01/04/1997); and cohort age threshold (18 or 40 years where applicable). Unexposed comparators received the same index date as their matched exposed. Comparators were censored if they later met the eczema definition (and could then be re-matched as exposed in other sets). Individuals with the outcome before index date were excluded from outcome-specific analyses. Follow-up: From index date until earliest of outcome, death, deregistration/left practice, study end (03/31/2023), or (for comparators) becoming exposed. Severity analyses: Time-updated eczema severity categorization: mild (default), moderate (receipt of potent topical corticosteroids or topical calcineurin inhibitors after meeting eczema definition), severe (secondary care eczema record, systemic immunosuppressants: azathioprine, ciclosporin, methotrexate, mycophenolate mofetil, or phototherapy). Severity could escalate (mild→moderate→severe) but not regress. To limit reverse causation, hospital admission was additionally adjusted in severity models. Severity analyses were limited to those eligible for linkage to secondary care records. Outcomes: 71 adverse health outcomes spanning atopic/allergic, immune-mediated, skin infections, mental health and substance use, cardiovascular, metabolic, bone health, digestive and neurological conditions. Outcome codelists were adapted from prior studies and mapped to CPRD Aurum codes. Statistical analysis: Cox proportional hazards models stratified by matched set estimated hazard ratios (HRs) with 99% confidence intervals (CIs). Primary results were from comorbidity-adjusted models; minimally-adjusted (by matching) and drug-adjusted models (adjusting for oral glucocorticoid and systemic immunosuppressant prescriptions at index) were also run. Multiple testing was addressed by reporting 99% CIs and indicating Bonferroni significance in supplements (p<0.0007 for 71 outcomes). Absolute rate differences (RD) per 100 or 1,000 person-years were estimated to gauge public health impact. Sensitivity analyses included alternative age cohorts, a more-severe cohort, and exclusion of non-consulters (those without primary care consultation indicators in the prior year). Analyses were implemented in R 4.3.1 using a reproducible targets-based pipeline.

Population and follow-up: From 46.8 million CPRD Aurum records, ~3.87 million met the eczema definition and were eligible; matching yielded large cohorts across age strata. Individuals were followed a median of ~4–5 years depending on cohort. After matching, age and sex were balanced; comorbidities and behaviors differed between groups. Primary associations: The strongest adjusted HRs were observed for food allergy (aHR ≈4.02; 99% CI 3.95–4.10), esophagitis (aHR ≈2.01; 1.92–2.10), and other allergic rhinitis (aHR ≈1.93; 1.91–1.94). Outcomes near null included prostate cancer (aHR ~1.00; 0.99–1.01), breast cancer (aHR ~1.03; 1.01–1.06), and Parkinson’s disease (aHR ~1.02; 0.98–1.06). Absolute RD estimates were highest for allergy-related infections and asthma (each ~5.4 per 100 person-years) and dermatologic infections (~3.8 per 100 person-years). Comorbidity adjustment generally attenuated HRs relative to minimally adjusted models. Strong associations not in 2022 AAD guidelines: Irritable bowel syndrome (aHR 1.31 [1.21–1.32]; RD 6.07), esophagitis (1.25 [1.23–1.28]; RD 0.48), gastroesophageal reflux disease (1.25 [1.23–1.27]; RD 1.10), thrombophilic disease (1.25 [1.23–1.27]; RD 0.5), sleep disorder (1.22 [1.21–1.23]; RD 0.77), COPD (1.22 [1.20–1.23]; RD 1.15), gastritis/duodenitis (1.21 [1.20–1.23]; RD 0.60), and peripheral neuropathies (1.21 [1.20–1.22]; RD 2.13). Larger HRs with small absolute RDs included Hodgkin’s lymphoma (1.85 [1.66–2.06]; RD 0.02), Crohn’s disease (1.62 [1.54–1.69]; RD 0.09), coeliac disease (1.42 [1.37–1.47]; RD 0.01), ulcerative colitis (1.40 [1.34–1.47]; RD 0.08), and autoimmune liver disease (1.32 [1.21–1.43]; RD 0.01). Category highlights: Confirmed strong associations with atopic/allergic conditions (asthma, allergic conjunctivitis, rhinitis, food allergy) and immune-mediated skin conditions (urticaria, alopecia areata), and with skin infections (e.g., impetigo, herpes simplex, molluscum contagiosum). More common but modestly elevated risks were seen for IBS, gastro-oesophageal conditions, thromboembolic disease, obesity, COPD, gastritis/duodenitis, and peripheral neuropathies. Most cancers and neurological conditions showed little or no association. Severity and sensitivity analyses: Drug-adjusted models were similar to comorbidity-adjusted results. Results were broadly consistent across any-age, 18+, and 40+ cohorts, though some outcomes (e.g., food allergy) varied by age cohort. Excluding non-consulters attenuated HRs across most outcomes, suggesting consultation bias. Severity analyses indicated dose-response for some outcomes, with comparatively larger HRs in severe eczema for certain liver, gastrointestinal, cardiovascular outcomes, osteoporosis, and fractures. Benchmarking: Adjusted HRs aligned closely with prior CPRD studies for anxiety, cardiovascular outcomes, fractures, and cancers. Consistent with prior work, there was little association with most solid cancers, but elevated risks for nonmelanoma skin cancer, non-Hodgkin lymphoma, and particularly Hodgkin lymphoma.

This comprehensive cross-outcome analysis addresses whether and how eczema is associated with a wide spectrum of adverse health outcomes. The study confirms strong links with atopic/allergic and immune-mediated skin conditions and skin infections, and identifies additional associations spanning digestive, metabolic, and thromboembolic outcomes. While some outcomes demonstrate modest relative risks, several are common and result in meaningful absolute risk differences, underscoring potential public health implications. Conversely, most solid organ cancers and many neurological conditions showed little association, refining the understanding of where clinical vigilance is most warranted. Severity analyses suggest that certain comorbidities (e.g., cardiovascular outcomes, osteoporosis, fractures) may be more strongly associated with severe eczema, pointing to potential heterogeneity by disease severity. Comparisons with prior CPRD studies show close agreement, enhancing confidence in the findings. The cross-outcome design improves comparability across conditions, reduces investigator bias, and allows explicit consideration of multiple testing. Nonetheless, causality cannot be inferred; observed associations may reflect residual confounding, healthcare-seeking behavior, or diagnostic practices. The results support multidisciplinary care approaches for people with eczema and provide an empirical basis for updating guidelines and prioritizing mechanistic and targeted outcome-specific research.

Across 71 outcomes, this large, representative UK primary care study provides a comprehensive overview of adverse health outcomes associated with eczema, incorporating sensitivity and severity analyses for comparability and robustness. The cross-outcome approach reduces investigator bias and facilitates benchmarking across conditions. Findings confirm strong associations with atopic/allergic and selected immune-mediated and infectious skin conditions, identify additional associations in digestive, metabolic, and thromboembolic domains, and show that some risks are concentrated among those with severe eczema. These results can inform clinical guidelines, support multidisciplinary management, and guide future research to elucidate mechanisms, the role of systemic inflammation, and the impact of disease severity and healthcare utilization on observed associations.

Key limitations include: (1) Missing data could not be explicitly handled due to the nature of EHR capture; (2) Outcomes were defined in primary care records and may miss diagnoses managed exclusively in secondary care or not recorded following consultation; ascertainment varies by condition; (3) Limited ability to assess and adjust for ethnicity given missingness and potential selection bias; residual confounding by lifestyle factors (e.g., smoking, obesity, alcohol) may remain despite diagnosis-based adjustments; (4) Potential reverse causation cannot be entirely excluded despite excluding individuals with pre-index outcomes; EHR diagnosis dates may not reflect true disease onset; (5) Eczema identification, though based on a validated algorithm, may misclassify milder cases and conflate subtypes; (6) Severity classification relied on treatments and hospital records, which approximate severity and may introduce misclassification; (7) The broad cross-outcome design sacrifices depth on specific causal pathways compared to focused studies, limiting exploration of confounding, mediation, and effect modification for individual outcomes.