Hormone replacement therapy and lung cancer risk in Chinese

Authors


Abstract

BACKGROUND.

The association between hormone replacement therapy (HRT) and a reduced lung cancer risk has been reported in previous studies. There is a high female to male ratio in Chinese lung cancer patients, and female patients have different clinicopathological characteristics compared with Western patient populations. The authors investigated whether HRT may reduce lung cancer risk in Taiwan.

METHODS.

The authors used a case-control study design to investigate 826 women with lung cancer and 531 healthy controls. Personal interviews based on a structured questionnaire were performed to collect information on HRT use of at least 3 months, age, ethnicity, active and passive smoking, exposure to air pollution, cooking or incense fumes, body mass index (BMI), menopause, and family history of cancers.

RESULTS.

HRT use was associated with reduced lung cancer risk with a multivariate, adjusted odds ratio of 0.70 (95% CI, 0.53–0.94; P = .019). HRT use was associated with reduced odds ratio of lung cancer in all subset analyses stratified by histology, active and passive cigarette smoking, BMI, history of incense burning, cooking, and motorcycle riding, as well as family history of certain cancers.

CONCLUSIONS.

This study confirmed that HRT is associated with a reduced lung cancer risk. The results appeared to be applicable to Chinese female population groups. Cancer 2007. © 2007 American Cancer Society.

Lung cancer is the leading cause of cancer deaths worldwide including Taiwan.1–2 Although tobacco smoking is a well-established etiologic factor for lung cancer,3 we found that most of the female lung cancer patients in Taiwan were nonsmokers.4 Lung cancer in Chinese women has different clinicopathological features compared with the Western population. The female to male ratio of lung cancer patients was higher than the average ratio worldwide,4, 5 and the cause remains unclear.

Female smokers have higher relative risks for lung cancer than male smokers.6–8 Sex hormones may influence the metabolism of tobacco carcinogens by cytochrome p450 in the liver.8 Estrogen receptors or estrogen binding sites are present in nonsmall cell lung cancer tissues.9–11 Estrogen could also potentially promote lung cancer.12 These findings suggest that sex hormones may play a role in lung carcinogenesis.

Hormone replacement therapy (HRT) is used to treat menopausal symptoms,13, 14 but there is concern that HRT use may increase cancer risk. Prior studies have shown that HRT use is associated with an increased breast cancer risk.15, 16 Conversely, HRT may be associated with a protective role in colon cancer carcinogenesis.17, 18 The association of HRT with lymphoma, ovary, liver, and other cancers remains uncertain.19

The association between HRT use and lung cancer risk is inconsistent. In a Swedish population-based cohort study, HRT use was associated with an increased lung cancer risk,20 whereas a study in Japan21 showed that induced menopause and HRT use were associated with an elevated lung cancer risk compared with normal menopausal women who did not use HRT. A recent Canadian cohort study demonstrated that long-term HRT use is associated with increased risk of lung cancer.22 However, the Women's Health Initiative study17 and a hospital-based case-control study23 found no association between HRT use and lung cancer. A nested case-control study by Elliott and Hannaford found a nonsignificant reduction of lung cancer risk for HRT users.24 Conversely, several studies have demonstrated an inverse association between HRT use and lung cancer risk.25–27

Because of the high female to male ratio of lung cancer patients in Taiwan and the frequent use of HRT in women for postmenopausal symptoms, the hypothesis that HRT may be associated with reduced lung cancer risk is of clinical importance and needs confirmation. In this study, we investigated whether HRT was associated with a reduced lung cancer risk in Chinese women.

MATERIALS AND METHODS

Study Subjects

We enrolled subjects who were participating in the Genetic Epidemiological Study of Female Lung Adenocarcinoma (GEFLAC) in Taiwan, which is a molecular epidemiological study on susceptibility markers for lung cancer in women. The study object of the GEFLAC is mainly focused on female lung adenocarcinoma, because adenocarcinoma is the most common histological type of lung cancer in Chinese women. However, for this study, GEFLAC included patients of all histologies to investigate the genetic epidemiology of lung cancer. Lung cancer patients, aged 18 years or older, with histological or cytological diagnoses were recruited from 6 tertiary medical centers in Taiwan, including National Taiwan University Hospital, Taipei Veterans General Hospital, Chang Gung Memorial Hospital at Linkou, Taichung Veterans General Hospital, National Cheng Kung University Hospital, and Kaohsiung Medical University Hospital. Those with metastatic lung cancer were excluded. Subjects admitted to the health examination departments of these 6 hospitals were recruited as controls. Those with a history of chronic bronchitis and emphysema were excluded. The institutional review boards of all 6 hospitals approved this research.

Data Collection

Informed consent was obtained from each study subject. Personal interviews based on a structured questionnaire were carried out to obtain information on age, ethnicity, education level, history of active and passive cigarette smoking, menopausal status and age at menopause, number of miscarriages, family history of various malignancies, history of incense burning, cooking, and motorcycle riding. The history of HRT and contraceptive medication use was recorded, including the age of initiation and the accumulative duration for medication use. If the subject was a current user, the accumulative duration for medication use was calculated from the time between the age of initiation and the age when the subjects were included.

Lung cancer histology was classified according to World Health Organization criteria.28 HRT use was defined as when the subject had ever received either estrogen replacement therapy or estrogen and progestin combination therapy for an accumulative duration of 3 months or more. Contraceptive use was defined as a history of contraceptive medication for an accumulative duration of 3 months or more.

The body mass index (BMI, kg/m2) was derived from the body height (m) and weight (kg). The body weight of the healthy controls was obtained at enrollment. The body weight of the cases was recorded according to the value while the case was in a healthy state so as to avoid underestimating the BMI because of cancer-related body weight loss.

“Ever cigarette smokers” were defined as those who had smoked cigarettes regularly for more than 6 months,25, 29 including current and former smokers. For the case group, “former smokers” were those who had quit smoking for at least 6 months before the diagnosis. In the control group, “former smokers” were those who had quit smoking for at least 6 months before the interview. “Pack-years” were defined as the average number of cigarette packs smoked per day multiplied by the number of years smoked. The history of passive cigarette smoking was defined as a history of inhaling other person's cigarette smoke at the workplace or a history of living with family members who smoked.

Exposure to air pollution was estimated by the history of riding a motorcycle in traffic at least 3 days a week for more than 6 months. Incense fume exposure was estimated by the history of burning incense at least once a month. Exposure to cooking fumes was defined as cooking for at least 6 months continuously. A family history of breast, ovarian, cervical, or endometrial cancers in first-degree female relatives was recorded.

Statistical Analysis

The SAS system for Windows version 8 (SAS Institute Inc, Cary, NC,) and the Statistical Program for Social Sciences program package, version 10.0 (SPSS Inc., Chicago, Ill) were used for statistical analysis. The differences between the case and control groups were analyzed by Pearson's chi-square test. We analyzed ethnicity, education, active and passive smoking status, menopausal status, number of miscarriages, HRT use, contraceptive medication use, family history of certain malignancies, histories of exposure to cooking fumes, exposure to incense fumes, and exposure to air pollution. The differences in mean age, BMI, age at menopause, and cigarette pack-years between the case and control groups were examined by Student's t test.

A multivariate logistic regression analysis was performed to estimate the odds ratio (OR) with its 95% confidence interval (CI) of developing lung cancer for each risk factor. The regression model included known lung cancer risk factors (age, ethnicity, educational level, and active cigarette smoking), hormone-related factors (BMI and menopausal status), potential environmental risk factors (cooking fume exposure, passive cigarette smoking, burning incense, motorcycle riding), as well as family history of certain malignancies. All statistical tests were 2-sided. A P-value < 0.05 was considered to be statistically significant.

RESULTS

Clinical Characteristics

A total of 1357 women were recruited from September 2002 to February 2006 from 6 hospitals; subjects included 826 female lung cancer patients and 531 healthy controls (Table 1). Among the cases, there were 22 (2.66%) patients with small cell lung carcinoma and 804 (97.34%) with nonsmall cell carcinoma, including 655 with adenocarcinoma (79.30%).

Table 1. Characteristics of the Women With Lung Cancer and Healthy Controls
CharacteristicCase no. (%)Control no. (%)P
  1. SD indicates standard deviation (standard error of the mean).

Age, y, mean [SD]59.73 [12.30]57.40 [14.16].001
Ethnicity
 Taiwanese655 (79.39)392 (73.82).017
 Hakka people67 (8.12)67 (12.62) 
 Others103 (12.48)72 (13.56) 
Education
 < 6 y513 (62.11)219 (41.24)< .001
 6–12 y219 (26.51)174 (32.77) 
 > 12 y94 (11.38)138 (25.99) 
Smoking status
 Ever59 (7.14)18 (3.39).004
 Never767 (92.86)513 (96.61) 
Passive smoking
 Yes574 (69.49)296 (55.74)< .001
 No252 (30.51)235 (44.26) 
Menopause
 Yes647 (78.52)387 (72.88).018
 No177 (21.48)144 (27.12) 
HRT use
 Yes145 (17.55)134 (25.24)< .001
 No681 (82.45)397 (74.76) 
Miscarriage
 None363 (46.90)199 (41.81).080
 1 fetus218 (28.17)132 (27.73) 
 > 1 fetus193 (24.94)145 (30.46) 
Contraceptive medicine
 Yes56 (7.19)47 (9.36).172
 No723 (92.81)455 (90.64) 
Family history of breast, cervical, ovarian, or endometrial cancers
 Yes63 (7.71)55 (10.44).093
 No754 (92.29)472 (89.56) 
Incense burning
 Ever543 (65.74)318 (59.89).033
 Never283 (34.26)213 (40.11) 
Employed as a cook
 Yes51 (6.17)31 (5.84).907
 No775 (93.83)500 (94.16) 
Exposure to cooking fumes
 Yes741 (89.71)460 (86.63).097
 No85 (10.29)71 (13.37) 
Regular motorcycle riding
 Yes368 (44.55)227 (42.75).538
 No458 (55.45)304 (57.25) 

As shown in Table 1, there were differences between the cases and controls in terms of age (P = .001), ethnicity (P = .017), and educational level (P < .001). Most cases and controls were never-smokers (92.86% vs 96.61%). Cases had a significantly higher proportion than controls for ever-cigarette smoking (7.14% vs 3.39%, P = .004) and passive cigarette smoking (69.49% vs 55.74%, P < .001). The mean number of cigarettes smoked in pack-years was similar among cigarette smoking cases and controls (mean ± standard deviation: 292.0 ± 199.3 vs 281.3 ± 174.7; P = .842).

Differences between cases and controls were also observed for the proportion of HRT use (17.55% vs 25.24%, P < .001), history of incense burning (65.74% vs 59.89%, P = .033), and menopause (78.52% vs 72.88%, P = .018). There was no significant difference between cases and controls in age at menopause, contraceptive medicine use, miscarriage, BMI, history of exposure to cooking fumes, history of motorcycle riding, and family history of breast, ovarian, cervical, or endometrial cancers.

From the multivariate analysis, we found that those with HRT use (P = .019), those without passive smoking (P < .001), and those with higher educational levels (P < .001) were significantly associated with a reduced lung cancer risk. The statistical power of this study was 92.4%.

HRT and Lung Cancer Risk

HRT use was significantly associated with a decreased risk of lung cancer risk showing a multivariate, adjusted OR of 0.70 (95% CI, 0.53–0.94, P = .019) as shown in Table 2. Stratification analyses demonstrated the consistency of the association between HRT and decreased risk of lung cancer among different histological subtypes such as nonsmall cell carcinoma (OR = 0.71; 95% CI, 0.53–0.95), large cell carcinoma (OR = 0.43; 95% CI, 0.21–0.87), adenocarcinoma (OR = 0.76; 95% CI, 0.56–1.03), squamous cell carcinoma (OR = 0.65; 95% CI, 0.29–1.45), and small cell lung cancer (OR = 0.51; 95% CI, 0.11–2.33). However, we may not draw conclusions from these subgroup analyses with the very small sample sizes in some of these subgroups.

Table 2. HRT Use and Lung Cancer Risk
 Case no. (%)Control no. (%)Univariate analysis OR (95%CI)Multivariate analysis* OR (95% CI)
  • OR indicates odds ratio; CI, confidence interval; HRT, hormone replacement therapy.

  • *

    Adjusted for age, ethnicity, smoking, education, BMI, menopause, cooking, motorcycle riding, passive smoking, incense burning, family history. The variable analyzed was not put into the multivariate logistic regression model.

Overall
 HRT use145 (17.55)134 (25.24)0.63 (0.48–0.82)0.70 (0.53–0.94)
 No HRT use681 (82.45)397 (74.76)  
Histology
 Nonsmall cell carcinoma
  HRT use142 (17.66)134 (25.24)0.64 (0.49–0.83)0.71 (0.53–0.95)
  No HRT use662 (82.34)397 (74.76)  
 Adenocarcinoma
  HRT use122 (18.63)134 (25.24)0.68 (0.51–0.89)0.76 (0.56–1.03)
  No HRT use533 (81.37)397 (74.76)  
 Squamous cell
  HRT use9 (15.52)134 (25.24)0.54 (0.26–1.14)0.65 (0.29–1.45)
  No HRT use49 (84.48)397 (74.76)  
 Large cell
  HRT use11 (12.09)134 (25.24)0.41 (0.21–0.79)0.43 (0.21–0.87)
  No HRT use80 (87.91)397 (74.76)  
 Small cell carcinoma
  HRT use3 (13.64)134 (25.24)0.47 (0.14–1.61)0.51 (0.11–2.33)
  No HRT use19 (86.36)397 (74.76)  

HRT use may be associated with a reduced risk of lung cancer in never smokers (OR = 0.70; 95% CI, 0.52–0.94) and those without passive cigarette smoking (OR = 0.61; 95% CI, 0.37–0.99) as shown in Table 3. The P-for-trends for age analysis was .107.

Table 3. HRT Use and Lung Cancer Risk: Age, Active and Passive Smoking
 Case no. (%)Control no. (%)Univariate analysis OR (95% CI)Multivariate analysis* OR (95% CI)
  • OR indicates odds ratio; CI, confidence interval; HRT, hormone replacement therapy.

  • *

    Adjusted for age, ethnicity, smoking, education, BMI, menopause, cooking, motorcycle riding, passive smoking, incense burning, family history. The variable analyzed was not put into the multivariate logistic regression model.

Smoking
 Ever smokers
  HRT use7 (11.86)2 (11.11)1.08 (0.20–5.71)0.17 (0.01–2.02)
  No HRT use52 (88.14)16 (88.89)  
 Never smokers
  HRT use138 (17.99)132 (25.73)0.63 (0.48–0.83)0.70 (0.52–0.94)
  No HRT use629 (82.01)381 (74.27)  
Passive smoking
 Yes
  HRT use105 (18.29)71 (23.99)0.71 (0.51–0.99)0.76 (0.52–1.09)
  No HRT use469 (81.71)225 (76.01)  
 No
  HRT use40 (15.87)63 (26.81)0.52 (0.33–0.80)0.61 (0.37–0.99)
  No HRT use212 (84.13)172 (73.19)  
Age, y
 < 50
  HRT use10 (5.75)12 (8.76)0.64 (0.27–1.52)0.40 (0.15–1.12)
  No HRT use164 (94.25)125 (91.24)  
 ≥50 to < 60
  HRT use64 (29.77)53 (44.92)0.52 (0.33–0.83)0.57 (0.34–0.94)
  No HRT use151 (70.23)65 (55.08)  
 ≥60 to < 70
  HRT use59 (24.48)51 (29.82)0.76 (0.49–1.18)0.95 (0.58–1.55)
  No HRT use182 (75.52)120 (70.18)  
 ≥70
  HRT use12 (6.12)18 (17.14)0.32 (0.15–0.68)0.52 (0.20–1.33)
  No HRT use184 (93.88)87 (82.86)  

HRT use may be associated with a reduced risk of lung cancer in all subset analyses stratified by BMI, history of incense burning, and history of cooking fumes exposure as shown in Table 4. HRT use may be associated with a reduced risk of lung cancer in those with BMI < 25 kg/m2 (OR = 0.66; 95% CI, 0.46–0.95), incense fume exposure (OR = 0.66; 95% CI, 0.46–0.96), or cooking fume exposure (OR = 0.70; 95% CI, 0.51–0.94). The P-for-trend for BMI analysis was .632.

Table 4. HRT Use and Lung Cancer Risk: BMI, Incense Burning, and Exposure to Cooking Fumes
 Case no. (%)Control no. (%)Univariate analysis OR (95% CI)Multivariate analysis* OR (95% CI)
  • OR indicates odds ratio; CI, confidence interval; BMI, body mass index; HRT, hormone replacement therapy.

  • *

    Adjusted for age, ethnicity, smoking, education, BMI, menopause, cooking, motorcycle riding, passive smoking, incense burning, family history. The variable analyzed was not put into the multivariate logistic regression model.

BMI, kg/m2
 < 25
  HRT use91 (17.43)88 (25.21)0.63 (0.45–0.87)0.66 (0.46–0.95)
  No HRT use431 (82.57)261 (74.79)  
 25≤BMI < 30
  HRT use46 (22.66)36 (25.71)0.85 (0.51–1.39)0.89 (0.51–1.56)
  No HRT use157 (77.34)104 (74.29)  
 ≥30
  HRT use7 (20.00)6 (35.29)0.46 (0.13–1.67)0.21 (0.03–1.67)
  No HRT use28 (80.00)11 (64.71)  
Incense burning
 Yes
  HRT use90 (16.57)86 (27.04)0.54 (0.38–0.75)0.66 (0.46–0.96)
  No HRT use453 (83.43)232 (72.96)  
 No
  HRT use55 (19.43)48 (22.54)0.83 (0.54–1.28)0.77 (0.47–1.26)
  No HRT use228 (80.57)165 (77.46)  
Exposure to cooking fumes
 Yes
  HRT use134 (18.08)124 (26.96)0.59 (0.45- 0.79)0.69 (0.51–0.94)
  No HRT use607 (81.92)336 (73.04)  
 No
  HRT use11 (12.94)10 (14.08)0.91 (0.36–2.28)0.60 (0.19–1.87)
  No HRT use74 (87.06)61 (85.92)  

HRT use may be associated with a reduced risk of lung cancer in subset analyses stratified by history of riding motorcycles (OR = 0.65; 95% CI, 0.42–1.00), those without a family history of breast, ovarian, cervical, or endometrial cancer (OR = 0.69; 95% CI, 0.51–0.95) as shown in Table 5. Again, we may not draw firm conclusions from the above subgroup analyses especially with the very small sample sizes in some of these subgroups.

Table 5. HRT Use and Lung Cancer Risk: Motorcycle Riding and Family History of Breast, Ovarian, Cervical, or Endometrial Cancers
 Case no. (%)Control no. (%)Univariate analysis OR (95% CI)Multivariate analysis* OR (95% CI)
  • OR indicates odds ratio; CI, confidence interval; HRT, hormone replacement therapy.

  • *

    Adjusted for age, ethnicity, smoking, education, BMI, menopause, cooking, motorcycle riding, passive smoking, incense burning, family history. The variable analyzed was not put into the multivariate logistic regression model.

Riding a motorcycle regularly
 Yes
  HRT use70 (19.02)55 (24.23)0.74 (0.49–1.09)0.65 (0.42–1.00)
  No HRT use298 (80.98)172 (75.77)  
 No
  HRT use75 (16.38)79 (25.99)0.56 (0.39–0.79)0.74 (0.49–1.11)
  No HRT use383 (83.62)225 (74.01)  
Family history of breast, cervical, ovarian or endometrial cancer
 Yes
  HRT use16 (25.40)18 (32.73)0.70 (0.32–1.56)0.58 (0.21–1.62)
  No HRT use47 (74.60)37 (67.27)  
 No
  HRT use129 (17.11)116 (24.58)0.63 (0.48–0.84)0.69 (0.51–0.95)
  No HRT use625 (82.89)356 (75.42)  

DISCUSSION

We found that HRT use was associated with a statistically significant reduction in the risk of lung cancer. The consistency of the association between HRT use and decreased lung cancer risk in subset analyses further strengthens the hypothesis that HRT may reduce the risk of lung cancer.

Several other studies have demonstrated an inverse association between HRT and lung cancer risk. In these studies, the inverse effect was found in different durations of HRT use. Kreuzer et al. found that more than 7 years of HRT use was associated with a reduced lung cancer risk.25 Olsson et al. demonstrated a reduced lung cancer risk in HRT ever users; long-term HRT use (48 months) was associated with a reduced risk of smoking-related cancers, including lung cancer.26 Schabath et al.27 found HRT use within the previous 6 months was associated with a reduced lung cancer risk. In our study, we defined HRT use as ever use of HRT for 3 months or more, which is similar to the definition by Blackman et al.23 The association of HRT and reduced lung cancer risk could be found in studies with different durations of HRT use. The relation between the duration of HRT use and lung cancer risk is unclear and may required further study.

Some studies have shown an association between reproductive/hormonal factors and reduced lung cancer risk in women. Kreuzer et al. found that oral contraceptive use was associated with a reduced lung cancer risk.25 Seow et al. reported a reduced lung cancer risk for those who had 3 live births or more.30 In a recent study, Kabat et al. found an inverse association between age at first live birth and lung cancer risk in parous women.22 Brenner et al. found an association between an older age at menarche and a reduced risk of lung cancer in China.31 They also found a nonstatistically significant association (Ptrend = .074) between a later age at menopause and reduced lung cancer.31 In our study, we found no significant association between lung cancer risk and hormonal/reproductive factors, such as menopausal status, number of miscarriages, contraceptive medication use, and age at menopause.

Carcinogenic polycyclic aromatic hydrocarbons from cigarette smoke can bind to estrogen receptors.32, 33 For women who received HRT, the exogenous hormones may bind to the estrogen receptors instead of the polycyclic aromatic hydrocarbons, thereby limiting lung cancer carcinogenesis. However, this does not explain the protective effect of HRT in a female population comprising > 90% nonsmokers in our study. The prevalence of ever smokers for women in Taiwan was 4.76%.34 We have also found a low prevalence of ever smokers among female lung cancer patients in our previous study.4 The cause for the very low prevalence rate of ever smoking in female lung cancer patients remains unclear.

From the multivariate analysis in this study, we found that passive smoking was associated with an increased lung cancer risk, which suggests that passive smoking may play a role in female lung carcinogenesis. The association of passive smoking and increase lung cancer risk has been discussed in other studies.35, 36 HRT may have protective effects for subjects with and without passive smoking. Nonsmoking factors may be involved in HRT and reduced lung cancer risk.

HRT use may be associated with a reduced risk of lung cancer for those with BMI < 25 kg/m2 in this study. The explanation for this is unclear. Obesity was associated with decreased lung cancer risk in 2 studies37, 38 but with increased lung cancer risk in another study.39 Obese people tend to have higher estrogen levels converted from androgens in fat tissue40, 41 and higher unbound-estrogen levels with lower levels of steroid hormone-binding globulin.41 Whether the circulating estrogen status of obese people was influenced by HRT remains unclear.

Exposure to combustion-related carcinogens from incense burning or frying pan fumes while cooking was associated with higher risks of lung cancer in nonsmoking Chinese women.42–44 However, a more recent study in Hong Kong did not show such correlations.45 No previous studies have investigated the effects of HRT use concerning incense burning and cooking. Our finding that HRT is associated with reduced lung cancer risk in women with a history of exposure to incense or cooking fumes suggests that the interaction between estrogen and incense burning or cooking fumes in lung carcinogenesis may need further investigation.

A familial aggregation of cancers of the reproductive system and the breast among female lung cancer patients has been reported.46, 47 An increased risk of lung cancer in association with cancers of the reproductive organs and breast48–54 was also described. These findings suggest that there may be some correlation in carcinogenesis among lung cancer and cancers of breast, cervix, endometrium, and ovary. Subgroup analyses from this study suggest that HRT may have a protective effect in patients with a family history of these cancers (OR = 0.58), but the sample size is too small in this subgroup to draw any conclusions.

Treatment of the A549 lung cancer cell line with estrodiol resulted in down-regulation of the epidermal growth factor receptor (EGFR). Treatment with fulveststrant, an antiestrogen agent, increased the EGFR expression in A549 cells.55 EGF treatment reduced ERβ expression, whereas gefitinib treatment, an EGFR tyrosine kinase inhibitor, increased the expression of ERβ.55 Combining fulvestrant and gefitinib gave maximum antiproliferative and proapoptotic effects in vitro.55 Hormone and EGFR interactions may have important clinical implications for targeted therapy of female Chinese lung cancer patients, because they have a good response to EGFR targeted-therapy such as gefitinib.56–58

There are several limitations to this study. This was a secondary analysis from a molecular epidemiological study on susceptibility markers of lung cancer in women. This type of study may also be subjected to selection or recall bias. There was incomplete information regarding the dose, duration, or various estrogens used in HRT. This case-control study is not case-matched. However, we adjusted for unmatched variables such as age, ethnicity, and education levels in our analysis. We may not draw firm conclusions from our subgroup analyses, especially with the very small sample sizes in some of these subgroups.

In conclusion, HRT was associated with a reduced lung cancer risk in Chinese women. Its multivariate, adjusted OR remained statistically significant in subset analyses stratified by other risk factors. The consistency of these findings appeared to support the hypothesis that HRT may reduce lung cancer risk.

Acknowledgements

The authors thank Dr. Schabath of the Division of Epidemiology, University of Texas School of Public Health, for kindly reviewing this article. The authors also thank Dr. Wing-Kai Chan of the Department of Medical Research, National Taiwan University Hospital, for helpful discussions and assistance in writing this article.

Ancillary