Hair dye use, regular exercise, and the risk and prognosis of prostate cancer: multicenter case–control and case-only studies

Study populations

To investigate the risk of prostate cancer, we conducted a hospital-based case–control
study at 2 large medical centers: Kaohsiung Medical University Hospital (KMUH) and
Kaohsiung Veterans General Hospital (KVGH), located in Southern Taiwan. Case patients
comprised men who had been newly diagnosed with and pathologically proven to have
adenocarcinoma of the prostate between August 2000 and December 2008. We matched each
case patient with one healthy man (control) who received a health check-up in the
Department of Preventive Medicine during the same month that the case patient was
diagnosed; the patients and controls were frequency matched according to age (in 2-y
bands), ethnicity, and hospital of origin. The controls had undergone digital rectal
examinations, the results of which were normal, and had serum prostate-specific antigen
(PSA) levels lower than 4 ng/dL.

To investigate the survival rate of patients with prostate cancer, we conducted a
case-only study, recruiting patients newly diagnosed with adenocarcinoma of the prostate
at the Third Medical Center at National Taiwan University Hospital (NTUH) in Northern
Taiwan between August 2000 and December 2007. Because the National Death Registry
of Taiwan has released the personal information, health status, and cause of death
for patients diagnosed before December 2007, we studied only the cases of patients
who were diagnosed before December 2007 in this case-only study. The 3 hospitals are
the main medical centers in their geographic areas and are accessible to patients
from all socioeconomic groups in Taiwan.

Data collection

Participants in the case–control and case-only studies underwent in-person interviews
conducted by trained interviewers using standardized questionnaires. The interviewers
questioned the paired case patients and the controls regarding demographic and lifestyle
characteristics before they were diagnosed with prostate cancer. The questionnaire
included questions pertaining to the demographic characteristics of age, body height
and weight (used to calculate body mass index [BMI]), education attainment, marital
status, blood type, vasectomy history, and family history of cancer. In addition to
the studied exposure factors (hair dye use and regular exercise), we collected other
common and relevant environmental and lifestyle factors, such as diet and habitual
substance use, including cigarette smoking, alcohol consumption, and betel nut chewing.

Cigarette smokers, alcohol drinkers, and betel nut chewers were separately defined
as participants who had smoked 10 cigarettes per week for a minimum of 6 months; or
consumed any alcoholic beverage once per week for a minimum of 6 months; or chewed
one betel nut quid per day for a minimum of 6 months, respectively. The age at which
substance use began and ceased, the type of substance, and the daily consumption amount
and duration of use were documented for identified substance users 31]. The accuracy of information pertaining to substance use that was obtained from the
questionnaires has been validated in our previous studies on esophageal cancer 32]–34].

Assessment of hair dye use and exercise status

Hair dye habit was defined as dyeing the hair a minimum of once every 3 months for
at least 1 year. Detailed information regarding the age at first and final use, frequency,
and years of use were recorded for identified hair dye users. Regular exercise habits
were assessed by asking participants whether they exercised aerobically for a minimum
of 20 min and perspired, performing this activity regularly for at least 1 year. If
participants had regular exercise habits, we asked them to report their average exercise
frequency according to 5 choices (?1 time/d; 4–6 times/week; 1–3 times/week; 1–4 times/month;
and??1 time/month).

Clinical characteristics

The clinical-pathological characteristics, including the Gleason score, stage of disease,
and serum PSA level at diagnosis, were obtained from chart review and are described
in detail elsewhere 35], 36]. Disease stage was determined by analyzing the pathological findings, pelvic computed
tomography or magnetic resonance imaging, and radionucleotide bone scans, according
to the criteria established by the American Joint Committee on Cancer (AJCC) tumor,
node, and metastasis classification system (AJCC Cancer Staging Manual, Fifth Edition,
1997). The pathologic grade was determined according to Gleason scores and was classified
into 3 groups (?6, 7, or 8–10) 37]. Information on death from any cause was obtained from the National Death Registry
of Taiwan, which is accurate and complete because death registration is mandatory
in Taiwan and physicians must issue death certificates 38]. This study was approved by the Institutional Review Board of the Kaohsiung Medical
University Chung-Ho Memorial Hospital, Kaohsiung Veterans General Hospital, and the
Research Ethics Committee of National Taiwan University Hospital. The written informed
consent was obtained from all the study participants of the 3 medical centers prior
to participation.

Statistical analysis

Demographic and clinical characteristics were tabulated for the cases and controls
in the case–control study. Multivariable unconditional logistic regression models
were used to estimate the odds ratios (ORs) and 95 % confidence intervals (CIs) for
the relationships among hair dye use, regular exercise, and the risk of prostate cancer
after adjustment for other covariates. Initially, we included the variables that have
been considered significant risk factors or protective factors for prostate cancer
in previous studies, including age (65 y, ? 65 y), education attainment ( high school,
high school, high school), and family history of prostate cancer (yes, no). Missing
data were classified into an additional category in the models to maximize the study
participants. The additional variables were then added to the models according to
forward stepwise selection, and were included in the models if they caused a minimal
10 % change in the association between hair dye use or regular exercise and prostate
cancer risk for the risk of or protection against prostate cancer. The selected variables
included marital status, BMI (24 kg/m
²
, 24–26 kg/m
²
, ? 27 kg/m
²
), cigarette smoking (yes, no), alcohol consumption (yes, no), betel nut chewing (yes,
no), blood type, vasectomy history, and food or nutrient intake (multivitamin supplements,
tea, coffee, milk, soy products, and instant noodles). Because the intake frequency
of specific nutritional supplements—including vitamins A, B, C, D, and E, and calcium—was
less than 5 % among the study participants, we merged them into the multivitamin supplement
category. The covariates used in the final model of the case–control study comprised
age, marital status, blood type, education attainment, family history of prostate
cancer, cigarette smoking, alcohol consumption, and betel nut chewing. We used an
additional model, which only adjusted with age and family history of prostate cancer
that is the risk factors for prostate cancer with sufficient evidence as race.

In addition to analyzing whether participants used hair dyes (yes, no), we categorized
hair dye use based on age at first use (50 y, 50–60 y, or???60 y), duration of use
(?10 y and??10 y), frequency of use (6 times/y and??6 times/y), and year of first
use (before and after 1980, which is the approximate year of the reformulation of
dye products). We categorized exercise by frequency (1–6 times/week and???7 times/week).

For the case-only study, Kaplan–Meier analysis and log-rank testing were used to examine
the relationship between personal hair dye use or regular exercise and the prostate
cancer patient survival rate. Cox proportional hazards modeling was employed to compute
hazard ratios (HRs) and 95 % CIs for prostate cancer deaths after adjustment for other
covariates. The covariates included in the model were the clinical stage, PSA level,
and the same covariates used in the case–control study. Each participant accumulated
person-time beginning from the prostate cancer diagnostic date and ending on the date
of prostate-cancer-related death or the end of this study in December 2007. If patients
died from other causes, they were censored to account for the competing death attributable
to other causes 39]. In addition, we analyzed the effect that exposure (hair dye use or regular exercise)
exerted on the death attributable to other causes by censoring the study participants
with prostate-cancer-specific deaths. Data analysis was performed using the SAS 9.1
statistical package; all P values were 2-sided and significant below the .05 level.

Sensitivity analysis

Sensitivity analyses were conducted in both the case–control study and case-only study.
In the matched case–control study, we excluded the case patients with a hair dye exposure
history of less than 5 years before the occurrence of prostate cancer to account for
a latent period, and examined the robustness of ORs for hair dye use. Because we did
not inquire about the beginning or cessation of regular exercise among the study participants,
we could not analyze how this variable affected the risk of prostate cancer. For the
case-only study, we analyzed the case patients who were recruited before December
2006. For the missing data, in addition to treating them as an additional category,
we also analyzed the participants without missing data 40].