Association of advanced age and cancer history with autoimmune disease in melanoma patients: a cross-sectional study

We conducted a cross-sectional study of 291,333 patients aged 18–106 years old who were treated at in-patient or out-patient sites of the University of Connecticut Health Center between 2000 and 2018. Patient information was abstracted from a GE Centricity IDX database of the electronic medical record. The sample consisted of 36,219 cancer patients and a random sample of 255,114 patients without cancer history.

Patients were organized into four cancer-related study groups, those with: (1) primary melanoma only (n = 6543), (2) non-cutaneous malignant neoplasms only (n = 27,630), (3) both melanoma and non-cutaneous malignant neoplasms (n = 2046), and (4) patients without a cancer history (n = 255,114). International Classifications of Diseases, Ninth and Tenth Revisions (ICD-9, ICD-10) codes were used as follows: melanoma (ICD-9172.x; ICD-10 D03.xx, C43.x) and non-cutaneous neoplasms (ICD-9140–239 aside from 172 to 173.x, 196-198.x, 216.x. 232.x, 238–239.x; ICD-10 C00-D48 aside from C43-44.x, C77-C79.x, D03-04.xx, D22-23.x, D48.x). Among cancer patients, those having regional or distant metastases at diagnosis were identified by the following codes: ICD-9196-198.x; ICD-10 C77-79.x. These codes only specified the location of the metastatic tumor, not that of the primary tumor. Patients without cancer history were selected randomly and included patients with basal cell carcinoma and squamous cell carcinoma of the skin (ICD-9173.x; ICD-10 C43.x).

Using ICD codes, we ascertained presence of autoimmune disorders identified at or before the cancer diagnosis. A list of 317 ICD codes corresponding to 98 autoimmune conditions were queried in the database (eTable 1). Nonspecific codes that may refer broadly to infectious, allergic, and multiple other causes were excluded, such as “acute pancreatitis, unspecified” (577.0, K85.0). Autoimmune conditions developed after the first recorded date of cancer diagnosis were excluded.

The chi-square test was performed to compare differences in autoimmune prevalence in the four patient groups. Multiple logistic regression was performed to characterize the relationship between autoimmune status and age groups. Relevant demographic and clinical factors were analyzed as potential confounders, which included sex, race, smoking history, cancer type, and presence of metastases. A second regression analysis was performed that considered insurance, as a proxy for age and income, instead of age group; insurance and age group were studied in separate models due to the inherent correlation between age and Medicare status. Odds ratios were calculated with 95% confidence intervals using the coefficient of the regression’s best fit line. Statistical significance was set at p ≤ 0.05; p-values were modified for multiple comparisons using the Bonferroni adjustment. Analyses were performed in RStudio.

The five most prevalent disorders among all patients were: Lichen planus (1.5%), type 1 diabetes (1.2%), rheumatoid arthritis (0.9%), systemic lupus erythematosus (0.7%), and scleroderma (0.5%; eTable 2). In all study groups, women had a higher prevalence of autoimmunity.

Autoimmune prevalence was found to vary by study group. In the absence or presence of melanoma, a history of non-cutaneous cancers was associated with a higher prevalence of autoimmunity (Fig. 1); among those with both melanoma and non-skin cancers, the prevalence of autoimmunity was 20.0% (95% CI 18.3–21.7), and among those with non-cutaneous cancers alone, the prevalence was 16.5% (95% CI 16.1–16.9). Both rates were approximately triple that among patients without cancer history (6.2, 95% CI 6.1–6.3, p < 0.001). Of note, melanoma patients without a history of non-cutaneous cancer had the lowest prevalence of autoimmunity across the cancer groups (9.3, 95% CI 8.6–10.0). Further analyses demonstrated increased prevalence of autoimmunity among patients with benign vs malignant tumors, control patients with non-melanoma skin cancers vs those without, and melanoma patients treated before vs after 2011 (eTable 3).

A total of 2285 patients with evidence of metastatic disease were studied (eTable 4). Patients with metastatic non-cutaneous cancer were considerably older than their group in aggregate (median 69 years vs 62), while metastatic melanoma patients with and without non-cutaneous cancers had comparable median ages to their group in aggregate (65 vs 64 and 60 vs 62, respectively). We show evidence of variability in metastasis prevalence by cancer group. Only 1.8% of patients with primary melanoma alone had metastases, compared to 8.3% of non-cutaneous cancer patients and 24.0% of patients with both melanoma and non-cutaneous cancers (eTable 3). As seen with all other analyses, women with metastases had a higher prevalence of autoimmune conditions compared to men diagnosed with metastatic disease (eTable 5).

Autoimmune prevalence among those with metastases appeared similar to that of their corresponding study group without metastases (Table 2), with the exception of the study group with both melanoma and non-cutaneous cancers: those with metastases had a higher autoimmune disease prevalence of 24.0% (95% CI 20.1–27.9) compared to 19.1% (95% CI 17.1–21.1) in patients without metastases (p = 0.020).

A multiple logistic regression was conducted in order to identify independent factors associated with the presence of autoimmune conditions among all patients studied. Age groups greater than age 34 had increased odds of autoimmunity relative to the 18–34 age group (Table 3). Further, the 75 and older age group, compared to the 18–34 age group, was correlated with the highest odds of autoimmunity out of all age groups (Adjusted OR, 1.78, 95% CI 1.67–1.89; p < 0.001), which was found to be higher than females versus males (Adjusted OR, 1.53, 95% CI 1.48–1.57) or a history of melanoma versus no cancer (Adjusted OR, 1.59, 95% CI 1.46–1.73). The two greatest factors associated with autoimmunity overall were a history of non-cutaneous cancer with and without melanoma (Adjusted OR, 3.22, 95% CI 2.87–3.61; 2.46, 95% CI 2.37–2.56, respectively; p < 0.001). A calculation of the interaction contrast for this model determined that non-cutaneous cancer has an additive effect on the relationship between melanoma history and autoimmunity (IC = 0.31 ± 0.11, eFig. 1). There is additionally a multiplicative effect of non-cutaneous cancer on the relationship between melanoma and autoimmunity as evidenced by the distinct 95% confidence intervals. Non-Hispanic blacks had a significantly higher odds of autoimmunity relative to non-Hispanic white race (Adjusted OR 1.22, 95% CI 1.16–1.29, p < 0.001). Among all races, Native Americans had the highest odds of autoimmunity, but this effect was not significant after performing a p-value adjustment for multiple comparisons (Adjusted OR, 1.72, 95% CI 1.07–2.78, p = 0.4642).

A second multiple regression model that considered insurance type instead of age group showed that Medicare patients had the highest risk of autoimmunity among all insurance types relative to private insurance holders (OR, 1.40, 95% CI 1.36–1.45; p < 0.001; eTable 6). Medicaid status did not significantly affect odds of autoimmunity (OR, 1.02, 95% CI 0.97–1.08; p = 0.383). Both multiple regression models analyzing age and insurance type had similar Akaike information criterion (AIC) values (146,098; 145,682, respectively) and had significant correlation for the entire model (p < 0.001 for both). All univariate comparisons (unadjusted) were consistent with their multivariate counterparts.

Our results suggest that prevalence of autoimmunity generally increases by age, particularly age > 75 years old, consistent with prior work and hypotheses in gerontology [13, 14]. This prevalence appears to vary by tumor type in that patients with a history of non-cutaneous cancers compared to those without a history of non-cutaneous cancers had significantly higher prevalence of autoimmunity. History of having had both melanoma and non-cutaneous cancer exhibited the highest association with autoimmunity compared to the other three study groups (melanoma, non-cutaneous cancer, or no cancer history) .

Our findings appear to be supported by the current understanding of aging physiology. Increased tumor burden, either in metastatic disease or advanced age, are likely associated with pro-inflammatory cytokines that contribute to autoimmunity [15, 16]. Aging is shown to promote the ability of CD4+ T cells to generate an IL-17 response that promote autoimmunity in humans [17]. Tumor neoantigens have been linked to immune responses that have capacity to cross react with host tissues. Scleroderma is a well-studied example wherein patients diagnosed with lung cancer can develop interstitial lung disease through unique antibodies to tumor epitopes [18]. Similarly, there are associations between occult malignancy and dermatomyositis, polymyositis, rheumatoid arthritis, and systemic lupus erythematosus [19, 20]. It has been posited that autoimmunity may be reactive to an occult malignancy in the absence of an infectious state [21]. This theory may provide some rationalization as to why rates of autoimmunity were relatively high in patients with both benign and malignant tumors in our investigation. Additionally, autoimmune disorders can promote an inflammatory state that leads to cancer, particularly in lymphomas.

While the connection between aging and cancer is well-established, our study is the first large-scale observational study using real-world data, to our knowledge, reporting associations with autoimmunity. Given the link between autoimmunity and moderate to severe adverse events in ICI therapy, current NCCN Guidelines state that patients with neurologic and life-threatening autoimmune disease are recommended against receiving therapy [22]. Recommendations for senior patients and a full range of autoimmune disorders, however, are not included in recommendations. Further, the marked increase in autoimmunity odds observed at age 75 suggests associated changes in the immune system are seen in this population. This finding also suggest that age of 75 years also represent a critical point for collection of pertinent patient history prior to treatment with ICIs.

Future epidemiologic studies of prevalence of autoimmunity across the life span are suggested as there is a dearth of evidence to date in cancer patients. Given this lack of knowledge, it is difficult to place our prevalence data in context. For example, our overall autoimmunity prevalence of 7.3%, however, is somewhat consistent NIH 2005 estimate in the U.S. (5–8%), but higher than the worldwide estimate from Hayter’s meta-analysis (4.6%) [23, 24]. Our dataset notably looked at 98 conditions instead of a list of 81 conditions reported previously by Hayter. As with all EHR studies, such as ours, there are potential diagnostic inaccuracies that may impact our results. For example, urticaria secondary to an allergen can be incorrectly coded as “idiopathic” (ICD-9: 708.1, ICD-10: L50.1) when the condition should have been coded as “allergic urticaria” (ICD-9: 708.0, ICD-10: L50.0).

Our data demonstrate that cancer patients have a higher prevalence of autoimmunity compared to the non-cancer patient population. Cancer patients have an increased risk of subsequent autoimmunity, such as in paraneoplastic syndromes, and patients with certain autoimmune conditions have an increased risk of cancer [25, 26]. As autoimmune diagnoses were largely coded in the database at the time of cancer diagnosis, it remains uncertain which diagnosis came first. Thus, our work cannot support either directional arrow. Interestingly, while the autoimmune disease prevalence among melanoma patients (9.3%) was significantly higher than that of the non-cancer population (6.2%), the prevalence was significantly lower than those with a history of non-cutaneous cancer, with or without melanoma. The data are roughly consistent with limited past epidemiological studies on cancer patients. In those with lung and renal cancer, the prevalence of autoimmunity, using a shortened list of about 40 autoimmune conditions, is estimated to be 25% [27, 28]. In one study that examined melanoma patients and included 147 different autoimmune diagnoses, the prevalence of autoimmunity was found to be 20.5%, which is close to our rate of melanoma patients with non-cutaneous cancer [29].

Our study showed a significant effect of metastases in patients with both melanoma and non-cutaneous cancers. These patients had a 26% metastases-associated increase in autoimmune prevalence. Our results are similar to a study that that used a looser case definition for autoimmunity and found a 43% relative increase in autoimmune disease prevalence among melanoma patients with metastases compared to those without [29]. The disparate findings suggest that the specificity of autoimmune case definitions may impact results.

We found an odds ratio for autoimmunity of 1.53 for female sex vs male sex, which is slightly less than the relative risk of 2.4 calculated by Hayter et al. [26] Our data also show that non-Hispanic black and Native American race are associated with autoimmunity, which is consistent with a recent large national database study that found an association with these groups and multiracial individuals [30]. We did not find a significant association between Asian-American or multiracial race and autoimmunity, but their results were limited somewhat by small sampling. While the influence of sex and race on autoimmunity is incompletely understood, it is thought to be a combination of genetic and environmental factors [30].

This study is limited by incomplete clinical information in the medical record as well as inherent generalizability concerns from a single-institutional report. For example, severity, control, and activity of autoimmune conditions were not available consistently. Similarly, prognostic factors such as frailty, Charlson Comorbidity Index, ECOG score, and specific cancer stage or grade were not included in analyses. Additionally, the database does not specify which cancer site is linked to ICD codes of metastases, so it is uncertain whether melanoma metastases impact autoimmunity in the combined melanoma and non-cutaneous group. Selection bias might be a concern given that our sample is not population-based. Regarding the link between age and autoimmunity, for example, young patients are less likely to seek routine medical attention in a hospital-based clinic and may receive fewer diagnoses of autoimmunity as a result. However, to illustrate, aged 18–34 years old represented 9.6% of the study group without cancer compared to 3.3–6.0% in the cancer study groups.