Comorbidities of Keloid and Hypertrophic Scars Among Participants in UK Biobank

This cross-sectional study evaluates established comorbidities of excessive scarring in UK Biobank participants, with comparisons across ethnic groups, and identifies novel comorbidities via a phenome-wide association study.

Key Points

Question

What diseases are people with keloid and hypertrophic scars (excessive scarring) at risk of?

Findings

This cross-sectional UK Biobank study of 972 people with excessive scarring and 229 106 controls identified associations of excessive scarring with atopic eczema and hypertension, with evidence that associations may vary with ethnicity. A phenome-wide scan of this predominantly White cohort identified a range of previously unreported correlations; musculoskeletal disease and pain symptoms were prominent nondermatological associations.

Meaning

Excessive scarring may share underlying predispositions with systemic diseases; future research should aim to understand whether there are causal relationships linking the observed associations.

Abstract

Importance

Keloids and hypertrophic scars (excessive scarring) are relatively understudied disfiguring chronic skin conditions with high treatment resistance.

Objective

To evaluate established comorbidities of excessive scarring in European individuals, with comparisons across ethnic groups, and to identify novel comorbidities via a phenome-wide association study (PheWAS).

Design, Setting, and Participants

This multicenter cross-sectional population-based cohort study used UK Biobank (UKB) data and fitted logistic regression models for testing associations between excessive scarring and a variety of outcomes, including previously studied comorbidities and 1518 systematically defined disease categories. Additional modeling was performed within subgroups of participants defined by self-reported ethnicity (as defined in UK Biobank). Of 502 701 UKB participants, analyses were restricted to 230078 individuals with linked primary care records.

Exposures

Keloid or hypertrophic scar diagnoses.

Main Outcomes and Measures

Previously studied disease associations (hypertension, uterine leiomyoma, vitamin D deficiency, atopic eczema) and phenotypes defined in the PheWAS Catalog.

Results

Of the 972 people with excessive scarring, there was a higher proportion of female participants compared with the 229 106 controls (65% vs 55%) and a lower proportion of White ethnicity (86% vs 95%); mean (SD) age of the total cohort was 64 (8) years. Associations were identified with hypertension and atopic eczema in models accounting for age, sex, and ethnicity, and the association with atopic eczema (odds ratio [OR], 1.68; 95% CI, 1.36-2.07; P < .001) remained statistically significant after accounting for additional potential confounders. Fully adjusted analyses within ethnic groups revealed associations with hypertension in Black participants (OR, 2.05; 95% CI, 1.13-3.72; P = .02) and with vitamin D deficiency in Asian participants (OR, 2.24; 95% CI, 1.26-3.97; P = .006). The association with uterine leiomyoma was borderline significant in Black women (OR, 1.93; 95% CI, 1.00-3.71; P = .05), whereas the association with atopic eczema was significant in White participants (OR, 1.68; 95% CI, 1.34-2.12; P < .001) and showed a similar trend in Asian (OR, 2.17; 95% CI, 1.01-4.67; P = .048) and Black participants (OR, 1.89; 95% CI, 0.83-4.28; P = .13). The PheWAS identified 110 significant associations across disease systems; of the nondermatological, musculoskeletal disease and pain symptoms were prominent.

Conclusions and Relevance

This cross-sectional study validated comorbidities of excessive scarring in UKB with comprehensive coverage of health outcomes. It also documented additional phenome-wide associations that will serve as a reference for future studies to investigate common underlying pathophysiologic mechanisms.

Introduction

Keloids and hypertrophic scars are chronic disfiguring manifestations of excessive cutaneous wound healing considered prototypic of skin fibrosis.^1,2,3 Whether they are distinct entities or a quantitative extreme of normal wound scars remains an active area of debate.⁴ Both conditions involve excess extracellular matrix deposition and raised scar tissue, but unique to keloids are horizontal expansion of scars and absence of clinical regression. Common to both are high symptom burden^5,6 and lack of universally effective treatment.^7,8

There is increasingly a shift from single-disease focused research toward studying disease systems.⁹ Understanding disease comorbidities has both biological and clinical benefits, such as highlighting novel mechanisms and offering opportunities for targeted and early clinical intervention. Previous studies of keloid and hypertrophic scar comorbidities have been limited to candidate diseases, based on speculated biological^{10,11,12,13,14,15} or demographic similarities.^16,17

In this study, we comprehensively assessed disease associations of keloids or hypertrophic scars (henceforth excessive scarring) in UK Biobank (UKB), a multicenter population-based longitudinal observational study of more than 500 000 participants including more than 950 with a diagnosis of excessive scarring. We performed multivariable logistic regression analyses for previously studied comorbidities across several ethnic groups (as defined in UK Biobank) and carried out a phenome-wide association study (PheWAS) over a wide range of systematically defined diseases.

Methods

This project used UKB data under project No. 15147. The UKB study was approved by the National Health Service National Research Ethics Service (11/NW/0382). This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

Study Population

UK Biobank is a large population-based prospective cohort study that recruited more than 500 000 participants (aged 40-69 years at recruitment) who attended 1 of 22 assessment centers across the UK from 2006 to 2010.¹⁸ Volunteers provided written informed consent for their participation. Rich health-related information is available, including regularly updated self-reported health conditions, lifestyle indicators, and anthropometric and biological measurements. Longitudinal health record data are available, including hospital episode statistics and primary care data. Participants were also asked to report their ethnicity based on a set of predefined categories, further detailed in https://biobank.ctsu.ox.ac.uk/crystal/field.cgi?id=21000. We restricted our analyses to individuals with linked primary care records (limiting the possibility of misclassifying participants as unaffected due to missing data).

Ascertainment of Disease Status

Clinical Code Selection

Disease status was ascertained through the following data sources: self-reported (verbal interview), linked hospital episode statistics (International Classification of Diseases [ICD], ninth and tenth revisions [ICD-9 and ICD-10]), cancer register (ICD-9 and ICD-10), primary care records (Read2 and Read3), and the Office of Population Censuses and Surveys Classification of Interventions and Procedures (OPCS4). Clinical code lists were manually curated by study authors (C.Y.U., A.N.W.) for excessive scarring (described below), vitamin D deficiency, and atopic eczema (eTable 2 in Supplement 1). For conditions with preexisting manually curated code lists (hypertension, uterine leiomyoma), Read2 and ICD-10 clinical code lists were obtained from the CALIBER portal.¹⁹ These clinical code lists (eTable 2 in Supplement 1) were minimally adapted and mapped to Read3 and ICD-9 equivalents, respectively, using UKB Resource 592 (https://biobank.ndph.ox.ac.uk/ukb/refer.cgi?id=592).

Excessive Scarring Status

A broad definition of “excessive scarring” was used, constituting a diagnosis of keloid or hypertrophic scar, which can be difficult to clinically differentiate^20,21 and are not well distinguished by electronic health record coding systems. Primary care data included diagnostic codes specific to either keloid or hypertrophic scar, whereas those from linked hospital episodes statistics could pertain to either scar type. The final codes selected were 7014 (ICD-9), L910 (ICD-10), 7L19A (Read2), M218. (Read2), M214. (Read2 and Read3), M2y11 (Read2), XaC07 (Read3), XaPxn (Read3), and X78TS (Read3; eTable 2 in Supplement 1). Within this group, we additionally identified individuals who had likely received scar-related treatment to define a more homogeneous cohort with moderate to severe scarring (hereafter “treated excessive scarring”), defined by codes for scar excision/refashioning and triamcinolone treatment (eTable 3 in Supplement 1).

Comorbidities/Outcomes

A systematic search for excessive scarring disease associations was performed on Medline using the following query: (case-control studies/ or cohort studies/ or (Risk Factors or (comorbidity or Comorbidity) or comorbidities or (Prevalence or prevalence) or association or predispose or risk).mp.) and (hypertrophic scar.mp. or Cicatrix, Hypertrophic/ or ((Keloid or keloid or keloids) not Acne Keloid).m_titl.). This resulted in 708 references that were independently reviewed for relevance. From the 21 remaining references (eTable 1 in Supplement 1), disease associations selected for analysis were those studied in 2 or more independent reports, namely hypertension, uterine leiomyoma, vitamin D deficiency, and atopic eczema.

PheWAS Catalog

We used the PheWAS Catalog with phecodes that represent a single phenotype and corresponding groupings.²⁰ The same data sources used for clinical code selection were mapped to ICD-10 codes and subsequently mapped to 1518 phecodes using Phecode Map 1.2.

Statistical Analyses

Descriptive statistics are presented as frequencies with percentages for categorical variables and means with SDs for continuous variables. Pearson χ² test and the Welch t test were used for comparisons between categorical and continuous variables, respectively. Statistical significance was set at P < .05.

Significant comorbidity associations were further examined using multivariable logistic regression analyses to estimate odds ratios (ORs) and 95% CIs. In each model, excessive scarring was the exposure variable, and the comorbidity was the outcome. As we were testing 4 primary comorbidities (hypertension, uterine leiomyoma, vitamin D deficiency, atopic eczema), a Bonferroni-adjusted P value threshold (.05/4 = .0125) was used.

We noted 3403 individuals (1.2% of the study cohort) with missing data for at least 1 of age, sex, ethnicity, Townsend Deprivation Index (TDI), body mass index (BMI, calculated as weight in kilograms divided by height in meters squared), or smoking status. Participants with missing data were more likely to be male and of non-White ethnicity, with higher TDI and BMI (eTable 4 in Supplement 1). There were no significant differences in missingness between participants with and without excessive scarring diagnoses (eTable 5 in Supplement 1). Nevertheless, for testing associations, the full cohort of 230 078 participants was analyzed, and multiple imputation (20 imputed data sets) was used to account for missing data.²¹

To investigate the independent association with excessive scarring, a “minimal model” adjusting for age, sex (except uterine leiomyoma, which was restricted to female participants only, n = 125 771), and ethnicity, and a “full model” adjusting for the additional potential confounding covariates were fitted for each disease association. The additional covariates were (1) hypertension: BMI, TDI, smoking status, diabetes, hyperlipidemia¹⁹; (2) uterine leiomyoma: BMI,^22,23 TDI, smoking status; (3) vitamin D deficiency: BMI, TDI, smoking status^24,25; and (4) atopic eczema: TDI, BMI, allergic rhinitis, asthma.²⁶

To assess whether the disease associations varied by ethnicity, separate fully adjusted logistic regressions were fitted in the 3 largest ethnic groups. These comprised self-reported White participants (85.9%), Black or Black British participants (henceforth “Black participants,” 6.4%), and Asian or Asian British participants (henceforth “Asian participants,” 5.2%), as described in Study Population. To assess whether any differences in ORs across ethnicities could be attributable to chance, the analysis was repeated in the full cohort with an interaction term between excessive scarring status and ethnicity.

For the PheWAS, logistic regressions were performed to assess the association between excessive scarring and the prevalence of each phecode diagnosis, adjusting for age, sex, ethnicity, smoking status, BMI, and TDI. A Bonferroni-adjusted P value threshold of .05/1518 = 3.3 × 10⁻⁵ was used. Phecodes with fewer than 200 cases or controls were excluded.

All statistical analyses were performed with R, version 4.1.2 (R Foundation for Statistical Computing). Specific R packages included mice,²¹ gtsummary,²⁷ tidyverse,²⁸ flextable,²⁹ PheWAS,³⁰ targets,³¹ ukbwranglr,³² and codemapper.³³

Results

We analyzed 230 078 UKB participants (mean [SD] age, 64 [8] years; 125 771 [55%] were female) for whom linked primary care data were available (eFigure 1 in Supplement 1). A total of 972 participants had a record of excessive scarring (740 with a diagnostic code specific for keloid, 110 specific for hypertrophic scar, 177 for either keloid or hypertrophic scar; eFigure 2 in Supplement 1). The prevalence of excessive scarring for the 3 largest ethnic groups were 1.1% (Asian), 2.4% (Black), and 0.4% (White). Table 1 shows the baseline characteristics for the excessive scarring and comparator groups. In the excessive scar-affected group, there was a higher proportion of female participants than in the unaffected group (65% vs 55%) and a lower proportion of participants with self-reported White ethnicity (86% vs 95%). Of the 972 participants with excessive scarring, 106 had a record of scar-related treatment. There were no notable differences in baseline characteristics between these participants and those without a record of scar-related treatment (eTable 6 in Supplement 1).

Table 1. Baseline Characteristics for the Scar and Comparator Groups.

Characteristic	No. (%)			P valuea
	Study participants	Keloid or hypertrophic scar affected	Keloid or hypertrophic scar unaffected
No. of participants	230 078	972	229 106	NA
Age, mean (SD), y	64 (8)	63 (8)	64 (8)	<.001
Sex
Female	125 771 (54.7)	633 (65.1)	125 138 (54.6)	<.001
Male	104 307 (45.3)	339 (34.9)	103 968 (45.4)
Ethnic backgroundb
Asian or Asian British	4707 (2.1)	53 (5.5)	4654 (2.0)	<.001
Black or Black British	2509 (1.1)	61 (6.3)	2448 (1.1)
Chinese	600 (0.3)	7 (0.7)	593 (0.3)
Mixed	1167 (0.5)	7 (0.7)	1160 (0.5)
White	218 331 (94.9)	829 (85.3)	217 502 (94.9)
Other ethnic groupc	1686 (0.7)	10 (1.0)	1676 (0.7)
Townsend Deprivation Index,d mean (SD)	−1.33 (3.03)	−1.29 (3.12)	−1.33 (3.03)	.65
Ever smoked	135 946 (59.1)	526 (54.1)	135 420 (59.1)	.001
BMI, mean (SD)	27.5 (4.8)	27.9 (5.2)	27.5 (4.8)	.05
Hypertension	79 034 (34.4)	362 (37.2)	78 672 (34.3)	.06
Uterine leiomyomae	14 080 (11.2)	92 (14.5)	13 988 (11.2)	.009
Vitamin D deficiency	5547 (2.4)	50 (5.1)	5497 (2.4)	<.001
Eczema, atopic	13 501 (5.9)	99 (10.2)	13 402 (5.8)	<.001

Abbreviations: BMI, body mass index, calculated as weight in kilograms divided by height in meters squared; NA, not applicable.

^a Welch 2-sample t test; Pearson χ² test; statistical significance is declared at P < .05.

^b Ethnic background is based on a set of predefined categories in UK Biobank.

^c Participants selected this category if their background did not fit any of the groups listed.

^d The Townsend Deprivation Index is derived from national census data about car ownership, household overcrowding, owner occupation, and unemployment aggregated for postcodes of residence.

^e Only female participants (total n = 125 771; affected n = 633; unaffected n = 125 138) considered for uterine leiomyoma.

Previously Studied Associations With Excessive Scarring

All previously studied comorbidities (hypertension, uterine leiomyoma, vitamin D deficiency, atopic eczema) were more prevalent for individuals with excessive scarring. Vitamin D deficiency and atopic eczema were around twice as common in the excessive scar cohort: 5.1% vs 2.4% (vitamin D deficiency) and 10% vs 5.8% (atopic eczema). The higher prevalence for uterine leiomyoma and hypertension was less marked: 15% vs 11% (leiomyoma) and 37% vs 34% (hypertension). We found no evidence for differences in prevalence of the primary comorbidities between treated excessive scarring cases and the rest of the excessive scarring group (eTable 6 in Supplement 1); conversely, when considering the treated excessive scarring subgroup against the whole unaffected cohort, the statistically significant associations did not persist (eTable 7 in Supplement 1).

Each comorbidity was analyzed as a disease outcome using 2 multivariable logistic regression models: minimally adjusted (for age, sex, and ethnicity except for uterine leiomyoma, which was tested within female participants only adjusting for age and ethnicity) and fully adjusted (with additional potential confounders for each comorbidity; Table 2). Statistically significant associations with excessive scarring were observed for hypertension and atopic eczema in the minimally adjusted models, while the association with vitamin D deficiency fell short of Bonferroni-corrected significance (OR, 1.42; 95% CI, 1.05-1.93; P = .02). In fully adjusted models, only the association with atopic eczema (OR, 1.68; 95% CI, 1.36-2.07; P < .001) remained significant. Despite a positive effect-size estimate, there was no significant association between excessive scarring and uterine leiomyoma (OR, 1.19; 95% CI, 0.95-1.49; P = .13).

Table 2. Associations Between Excessive Scarring and Selected Comorbidities.

Outcome	Minimal modela		Full modelb
	OR (95% CI)	P valuec	OR (95% CI)	P valuec
Hypertension	1.24 (1.08-1.43)	.002	1.11 (0.96-1.30)	.16
Uterine leiomyomad	1.20 (0.96-1.51)	.11	1.19 (0.95-1.49)	.13
Vitamin D deficiency	1.42 (1.05-1.93)	.02	1.47 (1.08-1.99)	.01
Atopic eczema	1.78 (1.44-2.19)	<.001	1.68 (1.36-2.07)	<.001

Abbreviation: OR, odds ratio.

^a Minimal model adjusts for age, sex (except uterine leiomyoma), and ethnicity (as defined in UK Biobank) for the entire imputed cohort.

^b Full model adjusts for age, sex (except uterine leiomyoma), ethnicity (as defined in UK Biobank), and additional confounders for the entire imputed cohort (see Methods).

^c Statistical significance is declared at P < .05/4 = .0125.

^d Models for uterine leiomyoma are restricted to female participants only without adjusting for sex.

We then performed association testing within subgroups of participants defined by self-reported Asian (53 excessive scar affected, 4654 unaffected), Black (61 affected, 2448 unaffected), and White (829 affected, 217502 unaffected) ethnicity (Table 3). With the exception of atopic eczema, our analysis suggested a divergence between ethnic groups in the prevalence of each comorbidity and its association with excessive scarring (Table 3). The associations with hypertension and uterine leiomyoma were nominally significant in Black participants (OR, 2.05; 95% CI, 1.13-3.72; P = .02 [hypertension] and OR, 1.93; 95% CI, 1.00-3.71; P = .05 [uterine leiomyoma]) and not significant in Asian or White participants. Vitamin D deficiency was only significantly associated with excessive scarring in Asian participants (OR, 2.24; 95% CI, 1.26-3.97; P = .006). For atopic eczema, the association with excessive scarring was highly significant in White participants (OR, 1.68; 95% CI, 1.34-2.12; P < .001), nominally significant in Asian participants (OR, 2.17; 95% CI, 1.01-4.67; P = .048), and although not statistically significant in Black participants, exhibited a similar trend (OR, 1.89; 95% CI, 0.83-4.28; P = .13). Finally, to formally assess these differences, we fitted a full logistic regression model incorporating an interaction between excessive scarring status and ethnicity, finding statistical evidence for ethnicity-specific effect sizes in the case of hypertension in Black participants (relative to White participants).

Table 3. Summary of Comorbidities and Fully Adjusted Models Investigating Associations Between Excessive Scarring and Selected Comorbidities Within the 3 Main UK Biobank Self-reported Ethnic Groups.

Ethnic groupa	No. (%)
	Hypertension	Uterine leiomyomab	Vitamin D deficiency	Atopic eczema
Asian participants
Controls (n = 4654)	1995 (43)	291 (13)	1041 (22)	323 (7)
Cases (n = 53)	23 (43)	4 (14)	20 (38)	8 (15)
P value, observedc	.99	.99	.01	.04
Fully adjusted OR (95% CI)	0.95 (0.48-1.86)	1.04 (0.36-3.01)	2.24 (1.26-3.97)	2.17 (1.01-4.67)
P value, fully adjusted modeld	.87	.95	.006	.048
Black participants
Controls (n = 2448)	1056 (43)	458 (33)	263 (11)	144 (6)
Cases (n = 61)	38 (62)	18 (47)	6 (10)	7 (11)
P value, observedc	.004	.08	.99	.12
Fully adjusted OR (95% CI)	2.05 (1.13-3.72)	1.93 (1.00-3.71)	0.88 (0.37-2.08)	1.89 (0.83-4.28)
P value, fully adjusted modeld	.02	.05	.77	.13
White participants
Controls (n = 217 502)	74175	12843 (11)	3866 (2)	12621 (6)
Cases (n = 829)	290	63 (12)	20 (2)	83 (10)
P value, observedc	.62	.60	.21	<.001
Fully adjusted OR (95% CI)	1.08 (0.92-1.28)	1.07 (0.82-1.39)	1.32 (0.85-2.07)	1.68 (1.34-2.12)
P value, fully adjusted modeld	.33	.63	.22	<.001

Abbreviation: OR, odds ratio.

^a Ethnic group is based on a set of predefined categories in UK Biobank.

^b Only female participants considered for uterine leiomyoma (controls [White] n = 119 053, cases [White] n = 544; controls [Black] n = 1404, cases [Black] n = 38; controls [Asian] n = 2157, cases [Asian] n = 29). Logistic regression models for uterine leiomyoma are restricted to female participants only without adjusting for sex.

^c Pearson χ² test; statistical significance is declared at P < .05.

^d Statistical significance is declared at P < .05/4 = .0125.

Discovery Analysis

We screened 1518 phecodes across 17 disease groups, identifying 110 diseases significantly enriched among participants with excessive scarring (Figure, Table 4; eTable 8 in Supplement 1). There was strongest evidence of association for several dermatological diseases, most prominently sebaceous cyst (OR, 2.56; 95% CI, 2.17-3.03; P = 9.45 × 10⁻³⁰), nonepithelial skin cancer (OR, 2.89; 95% CI, 2.36-3.54; P = 2.03 × 10⁻²⁵) and the umbrella phenotype “diseases of hair/hair follicles” (OR, 2.3; 95% CI, 1.94-2.73; P = 1.50 × 10⁻²²), as well as infections of skin/subcutaneous tissue, seborrheic keratosis, actinic keratosis, acne, and notably, atopic/contact dermatitis (all with OR > 1.9 and P < 1.0 × 10⁻¹¹). Similarly strong evidence was observed for pain-related symptoms, particularly for joint pain (OR, 1.84; 95% CI, 1.61-2.1; P = 1.87 × 10⁻²⁰) but also back pain, cervicalgia, enthesopathies, and mastodynia (all with OR > 1.6 and P < 1.0 × 10⁻¹²). Significant associations with the largest effect sizes were abnormal weight gain (OR, 3.97; 95% CI, 2.37-6.66; P = 9.32 × 10⁻⁸) and heart valve replacement (OR, 3.9; 95% CI, 2.25-6.73; P = 6.65 × 10⁻⁷).

Figure. Multivariate Logistic Regressions to Estimate the Association of Excessive Scarring Status With the Risk of Each Phecode Diagnosis, Adjusting for Age, Sex, Ethnicity, Smoking Status, Body Mass Index, and Townsend Deprivation Index.

Dots represent phecodes, grouped into systemic categories denoted by alternating colors. Statistical significance was set at P < .015/1518 (based on the number of phecodes tested), indicated by the horizontal hatched line. HPV indicates human papillomavirus. An interactive version of this Figure can be found at https://cyu06.gitlab.io/ukb_excessive_scarring_phewas/.

Table 4. Phecodes Significantly Associated With Keloid and Hypertrophic Scar Status^a.

Group (No. of phecodes)	Phecode description
Genitourinary (15)	Mastodyniab; hypertrophy of breast (gynecomastia); inflammatory disease of breast; cystitis; frequency of urination and polyuria; lump or mass in breast; irregular menstrual cycle; excessive or frequent menstruation; ovarian cyst; noninflammatory disorders of vagina; noninflammatory disorders of vulva and perineum; premenstrual tension syndromes; urinary tract infection; kidney failure NOS; inflammatory diseases of uterus, except cervix
Dermatologic (14)	Sebaceous cystb; diseases of hair and hair folliclesb; atopic/contact dermatitis due to other or unspecifiedb; other local infections of skin and subcutaneous tissueb; seborrheic keratosisb; actinic keratosisb; acneb; rash and other nonspecific skin eruption; other hypertrophic and atrophic conditions of skin; hyperhidrosis; disturbance of skin sensation; carbuncle and furuncle; pruritus and related conditions; diseases of nail, NOS
Respiratory (12)	Acute upper respiratory infections of multiple or unspecified sitesb; coughb; acute sinusitisb; other diseases of respiratory system, NEC; acute pharyngitis; acute laryngitis and tracheitis; influenza; shortness of breath; pneumonia; pneumonia due to fungus (mycoses); chronic sinusitis; chronic pharyngitis and nasopharyngitis
Infectious diseases (11)	Postoperative infectionb; tuberculosis; viral warts and HPV; dermatophytosis of nail; sexually transmitted infections (not HIV or hepatitis); dermatophytosis; candidiasis; herpes simplex; viral infection; mycoses; gram-negative septicemia
Sense organs (11)	Conjunctivitis, infectiousb; dizziness and giddiness (light-headedness and vertigo); inflammation of eyelids; eustachian tube disorders; disorders of lacrimal system; otalgia; otitis externa; infection of the eye; hearing loss; conjunctivitis, noninfectious; tinnitus
Neoplasms (8)	Other nonepithelial cancer of skinb; melanomas of skin; malignant neoplasm, other; lipoma; Other benign neoplasm of connective and other soft tissue; benign neoplasm of lip, oral cavity, and pharynx; malignant neoplasm of female breast; nevus, nonneoplastic
Symptoms (8)	Malaise and fatigueb; back painb; cervicalgiab; abdominal pain; nausea and vomiting; swelling of limb; sciatica; edema
Circulatory system (7)	Angina pectoris; nonspecific chest pain; other chronic ischemic heart disease, unspecified; coronary atherosclerosis; heart valve replaced; endocarditis; orthostatic hypotension
Musculoskeletal (7)	Pain in jointb; peripheral enthesopathies and allied syndromesb; enthesopathyb; bursitis; other and unspecified disc disorder; other disorders of soft tissues; juvenile osteochondrosis
Digestive (6)	Hemorrhage of rectum and anus; symptoms involving digestive system; irritable bowel syndrome; diseases of lips; anal and rectal polyp; dysphagia
Neurological (5)	Acute pain; organic or persistent insomnia; other headache syndromes; other peripheral nerve disorders; migraine
Injuries and poisonings (3)	Sprains and strains; contusion; certain early complications of trauma or procedure
Endocrine/metabolic (2)	Abnormal weight gain; hypercholesterolemia
Mental disorders (1)	Adjustment reaction

Abbreviations: HPV, human papillomavirus; NEC, not elsewhere classified; NOS, not otherwise specified.

^a Significant associations with excessive scarring (P < .05/1518 = 3.3 × 10⁻⁵), grouped by category.

^b The 20 most significantly associated phecodes.

Associations were identified with hypertension (OR, 1.26; 95% CI, 1.08-1.46; P = 2.05 × 10⁻³) and vitamin D deficiency (OR, 1.47; 95% CI, 1.08-2.00; P = 1.34 × 10⁻²) as expected, but these did not meet Bonferroni-corrected significance. Other previously reported associations that did not meet our selection criteria for specific analysis were explored, including obesity, osteoporosis, skin cancers, pancreatic cancer, migraine, and asthma (eTable 1 in Supplement 1). Of these, statistically significant associations were observed for skin cancers (melanoma, OR, 3.17; 95% CI, 2.24-4.50; P = 3.33 × 10⁻¹¹) and migraine (OR, 1.53; 95% CI, 1.27-1.84; P = 6.29 × 10⁻⁶) (eTable 9 in Supplement 1).

Discussion

To date there have been few large-scale association studies for excessive scarring.^{10,11,12,13,15} This cross-sectional study aimed to both validate previously studied associations (hypertension,^16,34,35 uterine leiomyoma,^13,17 vitamin D deficiency,^14,15 and atopic eczema^11,12) as well as scan for excessive scarring associations across the phenome. Our ethnicity-specific analysis includes a comprehensive study in White people, the most-represented ethnic group in UKB, and an understudied group in keloid literature.

In the transethnic UKB population, we replicated associations with 3 (hypertension, vitamin D deficiency, and atopic eczema) of the 4 primary comorbidities previously studied 2 or more times. Only the associations with hypertension and atopic eczema were statistically significant after correcting for multiple tests. The association with hypertension was attenuated after adjusting for additional risk factors (BMI, TDI, smoking status, diabetes, hyperlipidemia). This suggests that the observed difference in hypertension prevalence is associated with differences in these risk factors between people with and without excessive scarring. However, in our cross-sectional analysis, we are unable to determine the causal relationship between excessive scarring and increased hypertension risk factor burden.

Subgroup analyses revealed possible ethnic variations in comorbidity risk, particularly for hypertension in Black participants, in whom we found a significantly larger effect size than in White participants. Although our relatively small subgroup sample sizes make robust interpretation challenging, given the well-established disproportionate burden of diseases within the respective ethnic groups,^22,36,37,38 we propose that there may be ethnicity-specific risk factors that are shared by these pathologies.

The only association that showed nominally significant evidence for association in multiple ethnic groups was atopic eczema. Interestingly, this has previously been reported in Taiwanese¹² and Korean¹¹ populations, and we now observe this association across all 3 of our broadly defined ethnic groupings; however, regional variations in disease prevalence and associations may still emerge. Atopic eczema skin is more likely to be excoriated, and scars may thus be an epiphenomenon. Nevertheless, our result adds epidemiological support to the hypothesis that the Th2 inflammatory axis contributes to keloid pathogenesis.^{39,40,41,42,43,44}

Previous nongenetic applications of PheWASs^45,46,47,48 have been based only on ICD diagnostic codes. This would have excluded a large proportion of individuals, who were only identified through primary care codes. Through our comprehensive strategy to maximize identification of people with excessive scarring, we highlighted numerous significant disease associations, potentially indicating an increased risk of poorer health outcomes.

The frequent female genitourinary disease associations are consistent with suggestions that sex hormones may play a role in keloid pathophysiology.^49,50,51 Potentially, this could be explained by the overrepresentation of female participants within the excessive scar-affected group; however, our models adjusted for sex. The highly significant associations with dermatological conditions and neoplasms may represent true predispositions or reflect ascertainment bias (ie, if a patient presents with a dermatological condition or is reviewed postsurgically, a scar-related diagnosis is more likely to be recorded).

Of the dermatological associations, diseases of the pilosebaceous unit (sebaceous cyst, diseases of hair/hair follicles, acne) support the “sebum hypothesis,”⁵² which is based on high sebaceous gland density observed in keloid-prone skin⁵³ and sebum being intrinsically proinflammatory.⁵⁴ The association of keloid with skin cancer has been previously reported.¹⁰ Although plausible reasons have been proposed, including similar bioenergetics (reliance on glycolysis)^55,56 and signaling pathways including TGFβ/Smad^57,58,59 and Wnt/beta-Catenin,^60,61,62,63 this finding is interpreted cautiously, again considering the risk of ascertainment bias.

The associations with musculoskeletal disorders (enthesopathy, pain in joint, back pain, cervicalgia) may support the observation of chondrogenic misdifferentiation in keloids⁶⁴ and the shared significance of TGFβ in joint pathologies.^65,66 Interestingly, associations with pain symptoms spanned disease categories (nonspecific chest pain, irritable bowel syndrome, mastodynia, acute pain, headache syndromes, pain in joint, back pain, cervicalgia). Pain is known to debilitate some patients with keloids⁶⁷; whether there is shared underlying biopsychosocial dysfunction with other pain entities or whether they may be mutually reinforcing is speculative. Nonetheless, chronic pain represents a major global burden of disease,⁶⁸ and proactive identification of these conditions may aid patient counseling and treatment decisions.

Finally, whether an individual whose skin scars excessively is at risk of excessive internal scarring remains unanswered. In our study, the association of peritoneal adhesions with excessive scarring carried an OR of 3.68 (95% CI, 1.87-7.25; P = 1.19 × 10⁻⁴), which is intriguing but this is based only on 617 cases of peritoneal adhesions, 9 of whom had excessive scarring.

Limitations

Although this study used a large biobank cohort, a relatively small sample size of participants with excessive scarring were identified (n = 972). This is particularly relevant when attempting to dissect differences between ethnic groups, as there is less statistical power to detect significant associations in groups with lower sample numbers. However, it is reassuring that disease prevalence in our data set is in keeping with currently available epidemiological studies.⁵³ As most participants report White ethnicity, the main findings, particularly from the PheWAS, may not be generalizable to other ethnic groups for whom excessive scarring is a more prominent issue.

From our PheWAS, there are 2 further discussion points. The lack of significant associations with other fibrotic/scarring comorbidities (eg, lung or liver fibrosis) supports previous reports,^13,69 although low case numbers may mean our investigations were insufficiently powered to detect them. Second, it was striking that all significant associations were positive (ie, increased prevalence of comorbidities in people with excessive scarring), potentially a result of coverage bias whereby participants with more complete coverage of linked health data may be more likely to have a record of excessive scar diagnosis as well as a diagnosis of any other comorbidity. This may mean that the effect sizes are overestimated; however, the relative order of the associations remains informative. We do not draw causal conclusions from our findings; rather, we inform on coexisting associations between diseases that may not have previously been appreciated.

Our inclusion of keloids and hypertrophic scars in the definition of excessive scarring results in a study population that may be heterogeneous with respect to disease severity (and potentially pathophysiology). It might be expected that more severe cases of excessive scarring should be associated with a higher burden of comorbidity. We undertook to formally analyze this by distinguishing a subset of treated excessive scar cases. This attempt to refine the study population to those with moderate to severe keloid scarring yielded 106 individuals. Notably, the prevalence of all 4 primary comorbidities was lower in the treated excessive scarring group than in other excessive scarring cases, substantially so in the case of vitamin D deficiency, although differences were not statistically significant and could represent a sampling effect (eTable 6 in Supplement 1). Consequently, the increased prevalence rates for uterine leiomyoma, vitamin D deficiency, and hypertension detected using a minimally adjusted model in the larger heterogeneous group were no longer observed in this subgroup (eTable 7 in Supplement 1). Taken together, this may indicate that the associations do not hold for individuals with a more severe excessive scarring phenotype, or may be a reflection of limited power due to the low case numbers. This remains an important question for future study. Moreover, our work highlights how data availability and information or misclassification bias are key challenges of phenotyping based on electronic health records, and emphasizes the need for greater dermatological presence and the integration of dermatology-relevant data within existing biobank studies. Using currently available clinical codes within UKB, we were only able to establish heterogeneous excessive scarring cases based on potentially subjective clinical assessment, without clinicopathological correlation. Nonetheless, our results add to what is already known in the literature, as evidenced by the detection of both previously reported and novel associations.

Conclusions

In this cross-sectional study, we report a comprehensive observational analysis of a heterogeneous cohort of UKB participants with excessive scarring, replicating previously reported disease associations for excessive scarring with atopic eczema and hypertension. Only the association with atopic eczema showed a similar trend across the 3 major ethnic subgroups (Asian, Black, and White participants). Our PheWAS implicates a range of unreported associations for reference when studying the pathophysiology of excessive scarring and may prove valuable in studying the associated disease areas. There is a need for further research among diverse populations to examine whether these findings generalize to other groups.

Supplement 1.

eTable 1: Summary of systematic search for published keloid and hypertrophic scar disease associations.

eTable 2: Clinical codelists for diseases in specific multivariable analyses.

eTable 3: Clinical codes used to identify individuals who are likely to have been treated for excessive scarring.

eTable 4: Summary of individuals with and without missing data.

eTable 5: Missing data in excessive scarring (Case) versus Control individuals.

eTable 6: Baseline characteristics for excessive scarring cases with and without scar-related treatment codes.

eTable 7: Baseline characteristics for treated excessive scarring cases versus control individuals without excessive scarring.

eTable 8: Detailed summary of phecodes significantly associated with excessive scarring status.

eTable 9: Phecodes of previously reported disease associations with excessive scarring status.

eFigure 1: Flowchart of participants included in study cohort.

eFigure 2: Venn diagram of patients with keloid and or hypertrophic scar diagnosis codes in their linked electronic health records.

Supplement 2.

Data Sharing Statement