Risk and predictors of psoriasis in patients with breast cancer: a Swedish population-based cohort study

We analyzed two population-based cohorts: a Swedish nationwide cohort of breast cancer patients and a regional breast cancer cohort (Table 1, Additional file 1: Figure S1).

The nationwide cohort of breast cancer patients comprised women who were part of the 1990 national census of Sweden and were diagnosed with primary invasive breast cancer between 2001 and 2011, at age 20–80 years (n = 56,976). Information on breast cancer diagnoses was obtained through the Swedish Cancer Register, which was founded in 1958 and is managed by the National Board of Health and Welfare. Since the focus of our study is on the risk of new-onset psoriasis, patients who had been diagnosed with psoriasis before the date of the breast cancer diagnosis were excluded, leaving 56,235 patients in the cohort (Table 1). To compare the risk of psoriasis, we randomly sampled up to 5 women from the general female population matched on age, county of residence, and social economic status (obtained from the 1990 national census of Sweden, categorized as blue collar workers, white collar workers, self-employed workers, farmers, and others). Each reference individual was alive and free of cancer and psoriasis on the date of the matched patient’s diagnosis (the index date). In total, 321 women could not be matched to an index case, resulting in 280,854 matched reference individuals. Follow-up of the cohorts started from the index date (i.e., diagnosis date for breast cancer patients) and ended on 31 December 2012, date of death, date of emigration, date of a secondary cancer diagnosis, or date of psoriasis diagnosis, whichever came first. Information on death and emigration was obtained through cross-linking the cohorts to the Swedish Causes of Death Register and the Swedish Migration Register, using the unique personal identity number.

All psoriasis cases were identified through the Swedish Patient Register [18], which was established in 1964 and achieved nationwide coverage for inpatient hospitalizations in Sweden since 1987. From 2001, hospital-based outpatient physician visits have also been reported by Swedish counties. Diagnoses were coded according to the 7^th–10^th Swedish Revision of the International Classification of Diseases (ICD) codes (ICD-7 (1964–1968): 706; ICD-8 and ICD-9 (1969–1996): 696; ICD-10 (1997–present): L40). The validity of a psoriasis diagnosis in the inpatient and outpatient registers is 81% [19]. The first psoriasis subtype recorded was defined by the use of the ICD-10 code into: psoriasis vulgaris (L40.0), palmoplantar pustulosis (L40.3), and arthropathic psoriasis (L40.5). To ensure specificity, only primary discharge diagnoses were considered for the present analysis.

A subset of 4365 breast cancer patients within LIBRO-1 who were alive in 2009 consented to participate in a questionnaire-based study and gave a blood sample for genetic analyses. Patients’ lifestyle information on cigarette smoking, body mass index (BMI), alcohol consumption, and physical activity prior to diagnosis was retrieved from the self-reported questionnaire. Genotyping was carried out using a custom Illumina iSelect array (iCOGS) comprising 211,155 single nucleotide polymorphisms (SNPs) [20]. Details of the iCOGS array design, sample handling, and quality control processes have been described elsewhere [20]. To assess genetic predisposition to psoriasis, we selected 35 genome-wide significant SNPs reported by a recent meta-analysis of psoriasis genome-wide association studies (GWASs) [21] for constructing a polygenic risk score (PRS) using a scoring routine in the PLINK software v1.9 [22]. All SNPs were not directly genotyped on iCOGS, but imputed instead using the 1000 Genomes Project March 2012 release as a reference [23]. All SNPs passed quality control (minor allele frequency (MAF) ≥ 0.01 and R²/IMPUTE info-score ≥ 0.5), and for each individual breast cancer patient, a weighted PRS was calculated using the following formula:where β is the per-allele log odds ratio (OR) for psoriasis associated with the risk allele for SNP _k , x _k is the number of alleles for the same SNP (0, 1, 2), and n is the total number of disease SNPs included in the profile. The SNPs and corresponding ORs (weights) used for the derivation of the PRS are summarized in Additional file 1: Table S1. For analysis, patients were categorized by tertiles of the PRS.

To assess the risk of psoriasis in patients with breast cancer, we used stratified Cox regression models (i.e., Cox regression models conditioned on the matching variables age, county of residence, and social economic status) and calculated hazard ratios (HRs) for psoriasis in patients with breast cancer, overall, and by dividing the follow-up time into periods of 0 to < 0.5 year, 0.5 to < 1 year, 1–5 years, and >5 years, and by age at diagnosis (20–44 years, 45–54 years, 55–64 years, and 65–80 years). The underlying time scale for all the analyses was time since diagnosis. The stratified Cox model has been a recommended analysis approach for matched cohort data, and it deals with the presence of potential confounders (measured and unmeasured) as well as any imbalances in the matching scheme caused by censoring [24‐26]. Potential effect modification of age was tested by adding an interaction term of breast cancer (yes vs. no) with age (20–44 years, 45–54 years, 55–64 years, and 65–80 years) to the Cox model. We used Kaplan-Meier analyses to assess the cumulative incidence of psoriasis in breast cancer patients and matched reference individuals. This analysis approach does not account for the competing risk of death. Therefore, to address this potential bias, we also analyzed the data using competing risk models, treating mortality as a competing event.

To identify the risk factors for psoriasis in patients with breast cancer, we analyzed associations with cancer treatment in the regional cohort of breast cancer patients. The basic model (Model 1) was adjusted for age at breast cancer diagnosis and calendar period. In the multivariable model (Model 2), all treatment variables were entered for mutual adjustment as fixed covariates. To investigate potential confounding by disease severity, a sensitivity analysis was conducted in which tumor characteristics were added to Model 2. We further studied the effect of genetic predisposition and lifestyle factors in the subcohort with questionnaire and genotyping data. These analyses were conducted in the same manner as the analysis of cancer treatments, with a basic model (Model 1) adjusting for age at breast cancer diagnosis and calendar period, and a multivariable model (Model 2) including all risk factors, including identified treatment-related risk factors. Ordinary Cox regression was used for all risk factor analyses, and the proportional hazards assumption was tested using Schoenfeld residuals.

Statistical analyses were performed using SAS (version 9.4; SAS Institute Inc., Cary, NC, USA) and Stata software (version 14.0; Stata Corporation, College Station, TX, USA). The study was approved by the Regional Ethical Review Board in Stockholm.

In total, 599 cases of psoriasis were observed during a median follow-up of 5.1 years in the nationwide breast cancer cohort compared to 2795 cases in the matched reference individuals. This corresponded to an incidence rate of 1.9/1000 person-years in breast cancer patients vs. 1.7/1000 person-years in the matched reference individuals, indicating a 0.2/1000 person-years of absolute risk increase (95% confidence interval (CI) = 0.1/1000–0.4/1000). The most common psoriasis subtype was psoriasis vulgaris (298 out of 599), followed by palmoplantar pustulosis and arthropathic psoriasis. The 5-year cumulative incidence of psoriasis in breast cancer patients and in the matched reference individuals was 1.0% and 0.8%, respectively (Fig. 1). These estimates were very similar to those observed in competing risk analyses, indicating that bias due to the competing risk of death is negligible.

Patients with breast cancer experienced a 17% increased risk of being diagnosed with psoriasis during follow-up (HR = 1.17; 95% CI = 1.07–1.28) (Table 2). The increased risk of psoriasis was mainly attributed to psoriasis vulgaris (HR = 1.33, 95% CI = 1.17–1.52), with no overall risk increase being observed for the other psoriasis subtypes. Analyses by time since diagnosis showed that the risk of psoriasis was highest in the second half of the first year after diagnosis (HR = 1.68, 95% CI = 1.30–2.19). The risk of psoriasis vulgaris was long-term increased, with the HR remaining significant between 5 and 12 years after diagnosis (HR = 1.33, 95% CI = 1.06–1.68). Risk of psoriasis did not vary among different age groups, and the interaction between a breast cancer diagnosis and age was not significant (P for interaction = 0.84).

Analyses by treatment characteristics showed no effect of chemotherapy and endocrine therapy on psoriasis risk (Table 3). Radiotherapy, in contrast, was associated with a twofold increased risk after adjusting for surgery, chemotherapy, and endocrine therapy (HR = 2.44; 95% CI = 1.44–4.12). An increased risk of psoriasis was also observed in breast cancer patients who underwent mastectomy, compared to those who had a lumpectomy (HR = 1.54, 95% CI = 1.03–2.31) in the multivariable adjusted model (Model 2). A sensitivity analysis with additional adjustment for tumor characteristics yielded similar results (Additional file 1: Table S2).

Table 4 shows the association of genetic and lifestyle factors with psoriasis risk in patients with breast cancer. A significantly increased risk of psoriasis was found among patients in the highest tertile of the psoriasis PRS compared to those having lower genetic scores (HR = 2.94; 95% CI = 1.57–5.49, P for trend < 0.001). In the multivariable model (Model 2), regular cigarette smoking for more than 1 year prior to diagnosis (HR = 1.59; 95% CI = 1.00–2.52) and a BMI larger than 30 kg/m² at diagnosis (HR = 2.10; 95% CI = 1.20–3.68) significantly increased the risk of psoriasis in breast cancer patients. In the multivariable model, a high level of physical activity (more than 2 h per week) prior to diagnosis was shown to protect breast cancer patients from psoriasis; however, this effect was not significant (HR = 0.59; 95% CI = 0.33–1.03).

The incidence of psoriasis was 17% higher in breast cancer patients compared to the matched reference individuals in this study. The relative risk of psoriasis was seen within the first year after diagnosis and was mainly attributed to an increased risk of psoriasis vulgaris (33% higher). Treatment-specific analyses indicated an increased risk of psoriasis in patients treated with radiotherapy and mastectomy. Apart from treatment-specific effects, we identified genetic predisposition, obesity, and smoking as independent risk factors for psoriasis in patients with breast cancer.

Previous studies have reported a higher risk of skin disorders such as dermatitis and skin infection in patients with breast cancer [27, 28]. We found a slightly increased risk of psoriasis in breast cancer patients as compared to the general population. However, the overall incidence of psoriasis is low, and the absolute risk differences observed were also rather small. Of note, the cumulative incidence estimates of the current study are based on primary clinical diagnoses of psoriasis from the patient register (1%) and are lower than the estimate reported in a previous study (13%) using skin screening results [7]. The use of a clinical definition may have resulted in underestimation of the incidence estimates while capturing the more severe cases.

The increased risk of psoriasis after breast cancer is clinically plausible, as some treatments for breast cancer can cause dermatological side effects. In the regional cohort, we found an association between radiotherapy and risk of psoriasis. Skin trauma (a burn, scratch, bruise, cut, etc.) has been reported as a triggering factor for 43–76% of incident psoriasis cases [12]. The highest risk of psoriasis within the first year after diagnosis could also partly be explained by trauma caused by ionizing radiation. Compared to lumpectomy, we observed a significantly higher risk of psoriasis in breast cancer patients with mastectomy in the multivariable model. Patients with mastectomy usually experience more wound complications and delayed wound healing [29]. As the two main pathogenic features of psoriasis (abnormal keratinocyte differentiation and hyperproliferation of keratinocytes) are also secondary to the altered development of the normal wound healing process [30], prolonged wound healing could trigger psoriasis onset. Additionally, an increased risk of psoriasis could be explained by psychological reactions to the disease diagnosis and treatment decision, since a severe stressful life event has been identified as a potential trigger for psoriasis [31]. Both factors (treatment and stress reaction to the actual diagnosis) could potentially explain the observed peak in psoriasis incidence shortly after diagnosis.

Our study further identified obesity, smoking, and a high genetic predisposition as independent risk factors of psoriasis in patients with breast cancer. Psoriasis is a highly heritable disease; the heritability rate is estimated to be 70% [13]. Previous studies have reported substantially higher relative risks of psoriasis when comparing the highest to the lowest quartile of the PRS, with risks being inversely correlated to the age of psoriasis diagnosis [17, 32, 33]. Although these results are not directly comparable to our findings, lower relative risks with genetic predisposition are anticipated in our study population because of the older age at psoriasis onset (after diagnosis of breast cancer). Smoking and increased BMI are also established risk factors for psoriasis in the general population [34]. The effect of smoking and obesity in the breast cancer cohort was comparable to effects previously observed in the general population [14, 16].

The main strength of our study is the large sample size and the population-based design using Swedish health registers, which minimizes selection and information biases. Other strengths include the abundant information of treatment, lifestyle, and genetic information in the regional breast cancer cohort. We also acknowledge several limitations. The validity (positive predictive value) of psoriasis diagnoses in the Swedish Patient Register is about 81% [19], which indicates a possibility of misclassification (e.g., misclassified radio dermatitis as psoriasis). Although mild cases of radio dermatitis and psoriasis have some symptoms in common (erythema on the skin and sometimes desquamation), severe psoriasis is quite different from radio dermatitis [35], and in contrast to dermatitis, does not disappear after a couple of weeks. As the patient register mostly includes the severe cases of psoriasis, this misclassification should not have influenced our results. Furthermore, because of increased medical surveillance, breast cancer patients may have received a diagnosis that would, in a healthy person, continue to be undiagnosed. However, the long-term risk of psoriasis vulgaris (up to 12 years) argues against surveillance bias being a pure explanation for the increased psoriasis risk observed.