Introduction

Breast cancer is typically diagnosed among older women, with a median age at diagnosis of 62 years in the United States [1]. However, some important risk factors for the disease occur during early life, including birth size, age at menarche, adolescent adiposity, adolescent exposure to tobacco smoke, and physical activity before first pregnancy [2,3,4,5,6,7,8,9]. Evidence suggests that these risk factors may alter circulating hormone levels [10]. Higher endogenous estrogen and androgen levels have been positively associated with breast cancer in postmenopausal women [11,12,13,14,15,16]. High androgen level-related conditions such as polycystic ovarian syndrome (PCOS) have also been associated with incident breast cancer [17].

Higher circulating androgen levels are also associated with acne [18], an extremely common skin condition affecting more than 85% of individuals aged 12–25 and frequently continuing into adulthood [19]. Indeed, 26% of women and 12% of men in their forties report having acne [20]. Higher androgen levels over-stimulate the sebaceous glands and alter the development of skin cells lining hair follicles. Although most acne cases resolve by themselves, about 40% of individuals require clinical treatment for acne during early adulthood [21]. Such “severe acne” is usually treated with isotretinoin (former brand name: Accutane) or systemic antibiotics [22].

Given that higher circulating estrogen and androgen levels are positively associated with both severe acne and breast cancer, history of severe acne (e.g., cystic or scarring acne) could be predictive of breast cancer risk. A recent study by Zhang et al., using data from the Nurses’ Health Study II, reported a 17% higher risk of breast cancer in association with severe teenage acne [23]. Our study seeks to further elucidate the association between severe acne and breast cancer risk by replicating and expanding on Zhang et al.’s study using data from the Sister Study, a large cohort of women in the United States. Sister Study participants self-reported severe acne at any time in their life. Therefore, we were able to evaluate the importance of timing of acne on breast cancer risk. Additionally, we collected information on common acne treatments, which we hypothesized could affect the acne-breast cancer association or function to identify the most severe cases of acne.

Methods

Study sample

The Sister Study is a prospective cohort of 50,884 women ages 35–74, recruited between 2003–2009 in the United States and Puerto Rico [24]. Women were eligible if they had never been diagnosed with breast cancer but had at least one sister who had been diagnosed with breast cancer. Baseline data collection consisted of computer-assisted telephone interviews, self-completed questionnaires, and in-home examiner visits. Follow-up is ongoing, collecting basic health data annually and more detailed data every 2–3 years. Data for this analysis are complete through September 2016 (data release 6.0). More than 90% of participants have responded to the latest follow-up questionnaire. The study was reviewed and approved by the Institutional Review Boards of the National Institute of Environmental Health Sciences and the Copernicus group [24]. All participants provided written informed consent.

We excluded women who developed breast cancer before baseline, had an uncertain history of breast cancer or unknown age at diagnosis (n = 74), or withdrew their consent (n = 2). Our final analysis dataset consisted of 50,808 women. Among these, 3056 women were diagnosed with invasive breast cancer or ductal carcinoma in situ (DCIS) during follow-up.

The primary exposure variable—severe acne diagnosis—was determined from this baseline question: “Has a doctor or other health professional ever told you that you had severe or cystic scarring acne?” Participants who answered “yes” to this question were considered exposed.

Participants who reported severe acne were asked about their age at diagnosis. To evaluate the importance of timing of severe acne exposure, we considered whether women were exposed prior to age 18 years (the median age at severe acne diagnosis) or, separately, before their first pregnancy. To further assess the impact of severe acne in early life, we examined self-reported personal blemish or acne products use at ages 10–13 years (“did not use,” “sometimes,” “frequently,” and, “don’t know.”). Additionally, we considered participants who were diagnosed with severe acne in addition to having frequently used blemish/acne products ages 10–13. We refer to this final exposure variable as “severe and adolescent acne.”

We additionally considered two alternative exposures for treated acne: (1) Accutane and (2) acne-specific antibiotic use. Under these two exposure definitions, participants were considered exposed if they reported ever taking Accutane or, separately, antibiotics to treat their acne. Participants reported their reason for using antibiotics on the baseline questionnaire; those who reported the reason as “acne” or similar were considered exposed under this definition.

Women who reported incident breast cancer diagnosis during follow-up were asked to allow their medical records to be released [25]. We were able to obtain medical records for 82% of incident breast cancer cases [26]. Agreement between self-reported and medical record data was high (99%); therefore, self-reported data were used when medical record data were missing [26].

Statistical analysis

We used Cox proportional hazards regression to model the association between severe acne and breast cancer, using age as the time scale. Women were followed from age at study enrollment to age at breast cancer diagnosis, with censoring at death, loss to follow-up, or September 15, 2016. Because some families included more than one sister in the study, we used robust variance estimators to account for correlations between sisters due to shared genetic and environmental factors. The proportionality assumption was tested using Wald χ2 tests at the 95% confidence level of interaction terms between the exposure of interest and time.

Due to the timing of the questionnaire, it was difficult to establish whether several confounders preceded severe acne diagnosis. Therefore, two adjustment models were considered: one minimally-adjusted and one fully-adjusted. The minimally-adjusted model only included race/ethnicity (non-Hispanic White, African-American, Hispanic, other), and the participant’s head of household’s highest educational attainment when the participant was 13 years old (a marker for socioeconomic status; high school or less, some college, bachelor’s degree, graduate degree). In the fully-adjusted model, the HR was adjusted for: age at menarche, age at first birth (as an interaction term with ever/never parity), body mass index (BMI; a restricted cubic spline with knots at the 5th, 35th, 65th, and 95th percentiles), menopausal status at baseline (pre or postmenopausal), history of hormonal contraceptive use (never used, last used within 5 years, used more than 5 years prior), history of hormone replacement therapy (never, estrogen alone, estrogen plus progestin), parity (nulliparous, 1, 2, ≥ 3 births), smoking status (never, former, current), and alcohol use (never/former, current < 1 drink/day, current ≥ 1 drink/day), in addition to race/ethnicity and participant’s baseline educational attainment. An interaction term between BMI and menopausal status was also included, because BMI is associated differently with pre- versus postmenopausal breast cancer [25, 27]. To test for a trend in the personal acne/blemish product use variable, we calculated the association p value for the ordinal form of the variable.

To assess effect measure modification, we examined the association between severe acne and breast cancer risk within strata of race/ethnicity, parity (nulliparous vs. parous), BMI (< 30 kg/m2 vs. ≥ 30 kg/m2), hormone therapy use, oral contraceptive use, age at menarche (≤ 12 years vs. > 12 years), and family history of breast cancer (1 sister/half-sister vs. ≥ 2 first-degree relatives). We also examined modification by menopausal status updated over follow-up. We tested for HR heterogeneity using likelihood ratio tests of the interaction between severe acne and each modifier.

We conducted various sensitivity analyses, including adjusting for oophorectomy status at baseline (both ovaries removed vs. one or neither removed), excluding women who had PCOS, adjusting for use of acne/blemish product use at ages 10–13 (frequent vs. infrequent or never), and excluding women who had had prophylactic mastectomy before baseline (Supplementary Table 2). We repeated these sensitivity analyses when estimating the effect of severe and adolescent acne. We additionally examined the association between acne and specific types of cancer (invasive, DCIS, estrogen receptor [ER] + , and ER −), including case–case comparisons. For these analyses, we censored other breast cancers at their age of diagnosis.

Results

1098 participants (2%) reported a severe acne diagnosis (Table 1). Mean overall follow-up time was 8.4 years. The Sister Study cohort is predominantly non-Hispanic White (84% of unexposed) and highly educated (85% of unexposed had at least some college). Exposed women were more likely to be premenopausal than unexposed women (45% vs. 34%), to have been diagnosed with PCOS (5% vs. 2%), to have used hormonal contraception within the past 5 years (14% vs. 9%), to have had their first pregnancy after age 30 (17% vs. 12%), and to be nulliparous (24% vs. 18%). Differences in covariate distribution by breast cancer case status are reported in Supplementary Table 1.

Table 1 Baseline characteristics of the Sister Study cohort (2003–2009)a by exposure status
Full size table

Among participants with complete data on all potential confounders, two percent (n = 1081) reported a diagnosis of severe acne (Table 2). Fourteen percent (n = 6756) had frequently used acne/blemish products at ages 10–13. Among participants with severe acne, 47% (n = 507) were diagnosed before age 18. One percent of all participants (n = 459) reported both severe acne diagnosis and use of acne/blemish products at ages 10–13.

Table 2 Acne diagnosis, acne treatment and breast cancer risk, NIEHS Sister Study 2003–2009
Full size table

Regression model results were similar comparing minimally with fully-adjusted models (Table 2). For all exposures, Wald χ2 tests of age by acne interaction terms supported proportionality of hazard functions. In the fully-adjusted models, we observed a statistically non-significant positive association between breast cancer and severe acne diagnosis (HR 1.23; 95% CI 0.98, 1.54). Results were stronger for diagnosis of severe acne before age 18 (HR 1.40; 95% CI 1.04, 1.90) versus age 18 or later (HR 1.08, 95% CI 0.77, 1.51; p for heterogeneity = 0.25). Results were similar between those diagnosed with acne before their first pregnancy versus after. We observed only a modest increase in risk for frequent acne/blemish product use at ages 10–13 (HR 1.07; 95% CI 0.96, 1.23; p for trend = 0.25). However, when we considered those who reported both severe acne and frequent adolescent acne product use, relative to everyone else, we observed a higher breast cancer risk (HR 1.35; 95% CI 0.97, 1.87).

When we considered treated acne, we observed a near-null association between Accutane use and breast cancer (HR 1.07, 95% CI 0.84, 1.35). However, risk was higher among women who had ever used Accutane for acne or had been diagnosed with severe acne (HR 1.17; 95% CI 0.98, 1.39). Similarly, women who reported ever using systemic antibiotics for acne treatment were more likely to develop breast cancer than women who did not (HR 1.17; 95% CI 1.00, 1.35).

Stratified analyses were consistent with main findings (Table 3), and none of the interaction terms was significant. Sensitivity analyses were also consistent with main findings (Supplementary Table 2). Results largely did not differ by breast cancer case definition (Table 4). However, all associations were stronger among women diagnosed with severe acne before age 18, relative to those with no severe acne.

Table 3 Stratified adjusteda hazard ratios (HRs) for their associations between severe acne and breast cancer in the Sister Study
Full size table
Table 4 Alternative case analyses for severe acne and breast cancer in the Sister Study
Full size table

Discussion

In this large, prospective study we observed a statistically non-significant positive association between breast cancer and severe acne diagnosis, particularly among women who experienced severe acne during adolescence. These results are consistent with the hypothesis that early-life hormone levels may influence breast cancer risk. Elevated sex hormones have previously been associated with both severe acne and increased breast cancer risk [11,12,13, 18].

Our effect estimates were similar to those reported by Zhang et al. for severe teenage acne (HR 1.17; 95% CI 1.03, 1.32) [23]. We expanded on Zhang et al.’s study by examining severe acne exposure at all ages. The Sister Study was well-suited to address this research question because, by recruiting women who all had a family history of breast cancer, it enriched the study population at risk for breast cancer, thus making breast cancer risk factor research more efficient [28].

We also explored the impact of acne treatments on the association of severe acne with breast cancer. We observed that the use of antibiotics for acne was associated with a higher breast cancer risk, while use of Accutane was not. However, misclassification of the antibiotic exposure was likely. For example, some participants reported using antibiotics for a combination of reasons, e.g., “acne and rosacea” or ambiguous reasons, e.g., “zits.” We also assumed that everyone who did not explicitly report using antibiotics for acne was a non-user, which may not be accurate. Accutane first became available in 1982, meaning that most Sister Study participants would not have had access to it during adolescence. Furthermore, because it cost several hundred dollars per month [29], its use might have been limited to wealthier women or women with good health insurance. In our sample, only 28% of women who took Accutane also reported severe acne. This suggests that Accutane was prescribed for less severe acne during the early years of its availability, at least among those who could afford it, which could explain why the association with breast cancer was more modest when Accutane was included in the exposure definition. It is also plausible that Accutane protects against breast cancer by reducing circulating levels of several hormones, including testosterone [30], thereby mitigating the positive association between severe acne and breast cancer risk.

To assess sensitivity to hormonal disorders, we ran analyses excluding women who reported having PCOS, which increases circulating androgen levels in women and is associated with acne [17]. The resulting HR was nearly identical to the original estimate, likely because extremely few women in our study sample (about 2.5%) reported having PCOS.

The association between invasive breast cancer and severe acne was very similar to that for DCIS, indicating no clear difference by disease stage. This provides evidence against the idea that the observed association is driven by screening bias, which could have resulted if women with better access to medical care were more likely to seek both treatment for their acne and regular mammograms. Results from other sensitivity analyses, such as excluding women who had had prophylactic mastectomy, and adjusting for oophorectomy status, were consistent with our main findings.

This study population is predominately non-Hispanic White and well-educated, which may limit the generalizability of our findings. Additionally, all participants in the Sister Study have a family history of breast cancer, which means they are already at higher risk of the disease [31, 32]. However, this risk-based sampling method enriches the Sister Study cohort for breast cancer endpoints and relevant exposures and is well-suited for studying breast cancer risk factors [28]. We observed no clear differences in the effect estimates by degree of family history. Moreover, Zhang et al.’s similar result from the Nurses’ Health Study II, whose participants were not recruited based on their family history, suggests that our findings are externally valid.

Misclassification of exposure was a concern in this study, as severe acne usually occurs during adolescence or early adulthood, meaning that participants were likely reporting their exposure many years after it happened. However, we expect misclassification would be non-differential by case status and thus any related bias would be towards the null.

A strength of our study was its large sample size, which gave us sufficient power to analyze rare exposures such as severe acne and to stratify by age at acne diagnosis. The prospective design, under which only new breast cancer cases were observed, enabled confirmation that exposure (severe acne) preceded disease. This established temporality and avoided recall bias. We also considered numerous severe acne exposure definitions and found consistent estimates of the association with breast cancer, which strengthens our confidence in the association. Finally, the Sister Study collected a wealth of data on potential confounders for which we could adjust our regression models to obtain more accurate effect estimates.

Conclusion

Our study offers further evidence that severe acne may be positively associated with breast cancer risk. Our findings are consistent with prior evidence to support a role for early-life hormonal imbalance in breast cancer development, and suggest more research is needed to elucidate the specific biological mechanisms underlying this association, particularly during adolescence. If this work can be further replicated and validated, severe acne should be considered, along with other risk factors, as a possible predictor of future breast cancer risk.