Early Onset Baldness and Prostate Cancer Risk¹

One of five American men will develop prostate cancer during their lifetime (1). Despite the high incidence of prostatic carcinoma, few risk factors other than increasing age, family history,and black race are established (2). A long-standing hypothesis is that prostatic neoplasia is stimulated by testosterone,and therefore, increased androgen levels represent a risk factor for this disease (2, 3). Evidence in support of the role of androgens is provided by the fact that eunuchs rarely develop prostate cancer, that castration has a palliative effect on prostate cancer, and that testosterone alone can produce prostatic adenocarcinoma in rats (2, 4, 5). Likewise, eunuchs will not develop baldness if castrated before age 25, and if supplemented with testosterone, eunuchs will assume age-appropriate hair patterns characteristic of their pedigree (6, 7). Despite these apparent similarities, only a few studies have explored whether hair patterning is associated with prostate cancer, and no associations have been found (8, 9, 10, 11). In the most recent of these studies,Demark-Wahnefried et al. (8) found that men who displayed vertex (crown) baldness had significantly higher levels of free testosterone, as did prostate cancer cases when compared to controls. However, despite strong associations between testosterone levels and disease and between testosterone levels and vertex baldness,no association between present hair patterning and prostate cancer risk was found within the sample of 50–70-year-old men (8). Given the strategy of targeting patients with early disease onset commonly used in studies of genetic susceptibility, we speculated on the emergence of early onset phenotypes that may be associated with disease and hypothesized that vertex baldness may indeed portend risk if assessed during an informative period of time (i.e.,early adulthood). To explore this issue further, retrospective data related to hair patterning at age 30 and age 40 were gathered from men participating in two discrete prostate cancer case-control investigations and are reported herein.

Data were generated from two prostate cancer case-control studies conducted by investigators at Duke University Medical Center during April 1993 to March 1998. For both studies, subjects were limited to mentally competent black or white men who had no prior cancer histories other than nonmelanoma skin cancer. In addition, the items and procedures used to assess and collect hair patterning data were identical for both studies, i.e., subjects were provided with a diagram of the HS³of Baldness as modified by Norwood (12) and asked to select the view pair that best represented their hair patterning at age 30 and again at age 40 (6). Although assessment using the HS requires no special training and concordance between lay and expert scores or repeated ratings have been reported as 98 and 99%,respectively (6), no validation studies have been performed to determine the validity and reliability of hair patterning data that are gathered retrospectively and by self-report. There,however, is no reason to believe that case-control status would influence the patterns, which were reported.

The first study was a prostate cancer case-control investigation that was aimed at determining anthropometric (body fat distribution,skeletal structure, and body musculature) and hormonal (total and free testosterone, dihydrotestosterone, and sex hormone binding globulin) risk factors for prostate cancer. The sample was recruited exclusively from the urology clinics at Duke University Medical Center and was further delimited to men who had been weight stable within 1 year of recruitment (no weight loss or gain >10% of body weight), between the ages of 50 and 70, and who had not received any hormonal treatment (including orchiectomy). Only cases with localized prostate cancer who were within 3 months of diagnosis were included. Controls were patients who came to the clinic expressly for prostate cancer screening (19%) or to be evaluated for benign prostatic hypertrophy (26%), prostatitis (19%), hematuria (6%),dysuria (6%), impotence (5%), nephrolithiasis (5%), nocturia (3%),epididymitis (3%), spermatocele (3%), urinary tract infections (2%),inguinal hernia (1%), renal or seminal vesicle cysts (1%), or retroperitoneal fibrosis (1%). Cases and controls were frequency-matched on age (50% between ages 50 and 60; 50% between ages 60 and 70) and race (80% white; 20% black). Recruitment for this study was conducted in person from April 1993 to January 1995. The refusal rate was 5%, with “lack of time” cited as the primary reason for nonparticipation, which was slightly higher among blacks as well as control subjects (total, n = 315). In November 1997, all surviving patients (per the Duke patient database, n = 287) were sent reprints of resulting papers along with a copy of the baldness scale, questions regarding their hair patterning at ages 30 and 40, and a preaddressed stamped postcard to record and return their response. Control participants also were asked whether they had been diagnosed with prostate cancer in the interval between their initial participation and the time at which they received the mailing (four controls were diagnosed with prostate cancer and were recorded and analyzed as cases). Response cards from 150 men and 42 unopened letters with no forwarding address were received, thus yielding a second tier response rate of 61%. Response rates did not differ between cases and controls; however, they did differ between blacks (34%) and whites (63%).

The second study was a community-based prostate cancer case-control investigation that was conducted within a 63 contiguous county region of North Carolina. The primary aim of the study was to ascertain the effect of environmental and occupational exposures on prostate cancer risk. Cases were identified from 16 hospitals that contributed 75% of cases within the geographic region; community-based controls frequency-matched on age (the age criteria for both cases and controls was ≤67 years of age), race, and county of residence were identified from Department of Motor Vehicle tapes. Recruitment was conducted via a mailed letter, which instructed interested participants to mail back a preaddressed stamped postcard. Given the limited funding that precluded active follow-up, a reliance on active consent was necessary and resulted in an overall acceptance rate of 9.7%. Rates of acceptance varied with regard to race (6.9% among blacks versus 13.6% among whites) and case-control status (6.4%among controls versus 22.3% among cases). Cases and controls, however, did not differ with regard to income and educational status. Cases were interviewed within 1 year of diagnosis, and data for this study were collected via an in person interview. Nurse interviewers showed each participant the HS of Baldness and asked men to indicate the view pair that best “fit” their hair patterning at age 30 and also at age 40; their responses to each question were recorded.

Data from both studies were reviewed and double-key entered (responses from four subjects, one from the Duke-based study and three from the community-based study, were excluded due to failed logic checks and inconsistency). For data analyses, HS scores of baldness at each age were collapsed into three categories: not bald (I/II); frontal bald(IIa/III/IIIa/IVa), and vertex bald (III-vertex, IV/V/Va/VI, and VII;see Fig. 1). To better classify balding individuals with regard to the directionality of their hair loss, i.e., balding beginning at the hairline versus balding beginning at the crown, and also to discriminate between early versus later onset balding, the following five categories were created and compared: not bald at 30/not bald at 40 (not bald: referent group); not bald at 30/vertex bald at 40 (vertex bald later onset); vertex bald at 30/vertex bald at 40 (vertex bald early onset); not bald at 30/frontal bald at 40 (frontal bald later onset); frontal bald at 30/frontal bald at 40 (frontal bald early onset); and frontal bald at 30/vertex bald at 40 (frontal to vertex bald). Given the centrality of vertex balding to our hypothesis, men reporting any vertex baldness were classified as being “vertex bald,” thus excluding the possibility of vertex to frontal hair loss pattern.

Given the differences in study design and proportionate accrual,as well as sample characteristics (e.g., racial composition and stage distribution) that are associated with prostate cancer and could confound androgen-related effects, a decision was made to conduct separate analyses on data from each study. χ²analyses were conducted to determine whether differences in hair patterning existed between cases and controls. Additional analyses using unconditional logistic regression models and adjusting for age and race also were conducted. Age was modeled as a continuous variable in the logistic regression analysis, with age set at age of diagnosis for cases and age at interview for controls.

There were significant differences between the studies with regard to race and stage. The community-based study had proportionately more blacks and more men with later-staged disease than the Duke-based study(Table 1).

Hair patterning data for prostate cancer cases and controls are presented in Table 2, and baldness categories with race and age-adjusted ORs are presented in Table 3 (unadjusted ORs are not presented because of their similarity to these data). Data suggest that vertex baldness is associated with an increased risk for prostate cancer, with an ∼2-fold increase in risk noted among men who develop vertex baldness by age 30. In contrast,data related to frontal baldness suggest that it might be protective or modify the risk associated with vertex baldness; however, data are too inconsistent to draw firm conclusions. For all analyses, the magnitude of the ORs was consistently greater in the community-based study, with effects that approached significance (at the 5% level), as compared to the Duke-based study where weaker trends were observed.

Of the studies that have explored the association between hair patterning and prostate cancer risk, this is the first to suggest differences between cases and controls. Wynder et al.(11), Oishi et al. (10), and Greenwald et al. (9) were all unable to detect differences between prostate cancer cases and controls with regard to body hair, hair thickness, and/or baldness (8). These investigations used different discriminators, such as body hair, and also did not employ standardized tools, such as the HS of Baldness, to assess hair patterning. In addition, some of these prior studies used techniques, such as the use of reunion photographs, which were incapable of capturing hair loss at the crown. However, even using standardized scales and techniques to capture vertex baldness,Demark-Wahnefried et al. (8) also were unable to detect case-control differences in hair patterning when assessed in men over the age of 50. Given the prevalence of hair loss with increasing age and evidence that eunuchs will only demonstrate baldness if castrated after age 25 or if supplemented with testosterone (6), it may be important to capture hair patterning at critical periods of time if its utility as a risk factor is to be adequately tested. This study is unique because: (a) head hair patterning was assessed using a standardized scale; and(b) data were collected on hair patterning during earlier adulthood.

Although the limitations of this investigation included its reliance on self-reported, retrospective data that may be subject to respondent and recall bias and a lack of power to achieve statistical significance,the ORs for vertex baldness were of similar direction and magnitude. The concordance between these results lends strength to our conclusion that early onset vertex baldness may place men at “moderate risk”for prostate cancer (13), with ORs that suggest a >2-fold increased risk among men who develop vertex baldness by age 30.

In summary, given the consistency of our findings, as well as the growing body of research that indicates that there are common biological factors and pathways that are associated with both prostate cancer and baldness [i.e., variations in the A2 allele in CYP17 (cytochrome P450c17α gene on 10q24.3 encoding for 17α-hydroxylase and 17/20-lyase; key enzymes in the androgen steroidogenesis pathway; Refs. 14, 15, 16), expressed levels of androgens and insulin-like growth factor-1 (3, 8, 17, 18), conversion of testosterone to dihydrotestosterone (2, 19), and androgen receptor status (19, 20, 21)], we encourage further study of early vertex baldness as a potential risk factor for prostate cancer.

We gratefully appreciate Jennifer Lucas, Alice Neary, Barbara Matthias, Megs Rosenberg, and Angela Ross for their help with data collection and entry. We also thank Amy Shearin and Jennifer Hoff for their assistance with data cleaning.