Hormone Replacement Therapy in Females Can Decrease the Risk of Lung Cancer: A Meta-Analysis

The purpose of the present meta-analysis was to determine the relationship between hormone replacement therapy (HRT) and lung cancer risk in females. Publications were reviewed and obtained through a PubMed, EMBASE database and Cochrane Library literature search up to May, 2012. The detailed numbers of patients in different groups, odd ratios (ORs) and corresponding 95% confidence intervals (CIs) were collected and estimated using a random-effects model. Twenty five studies entered into the meta-analysis. The total number of participates and lung cancer patients was 656,403 and 11,442, respectively. The OR of all 25 studies was 0.91 (95%CI  = 0.83 to 0.99) and P value was 0.033. In stratified analyses, the positive association between HRT use and decreased lung cancer risk was also found in the patients with BMI<25 kg/m2 (OR = 0.65, P = 0.000), and never smokers patients (OR = 0.86, P = 0.042). However, HRT use in patients with artificial menopause could increase the lung cancer risk, OR = 1.51(P = 0.001). The result of Egger's test did not show any evidence of publican bias (P = 0.069). In conclusion, our meta-analysis on HRT and lung cancer risk suggests that HRT use is correlated with decreased lung cancer risk in female, especially in female with BMI<25 kg/m2 and never smokers.


Introduction
Lung cancer is now the most commonly diagnosed cancer and the leading cause of cancer mortality in both men and women worldwide. Since less females smoke than males or females started smoking later than males, lung cancer trends among females lag behind males [1]. However, lung cancer incidence still keeps increasing [2]. Besides smoking, several occupational and environmental carcinogens such as asbestos, arsenic and radon are known to be risk factors for lung cancer in females [1]. Hormone replacement therapy (HRT) remains the most effective treatment for postmenopausal symptoms in menopausal females and young females who go into early menopause due to surgery or chemotherapy or radiotherapy to the pelvic region [3]. Adami et al. first demonstrated that lung cancer risk increased in women receiving HRT [4]. During the past two decades, studies on the HRT and lung cancer risk have demonstrated contradictory results, making the association still remained uncertain. Two clinical trials from Chlebowski confirmed the correlation between increased lung cancer risk and HRT [5,6], but Chen et al., Clague et al. and Ramnath et al. showed in their studies that HRT use in females can decrease the incidence of lung cancer [7][8][9].
In order to confirm a definite correlation between HRT and lung cancer risk, we conducted a literature search and performed meta-analysis of available studies.

Literature search and identification of the publications
A literature search of the PubMed, EMBASE databases and Cochrane Library (updated to May, 2012) was conducted using combinations of the following terms: ''hormone replacement therapy'', ''HRT'', ''oestrogen replacement therapy'', ''estrogen replacement therapy'', ''progestin replacement therapy'', ''ERT'', ''lung cancer''. We limited the search to studies in human and written in English. Two of the authors (Yanwen Yao and Xiaoling Gu) independently selected the articles with information on the association between HRT and lung cancer morbidity, and further check the reference list of the publications. All studies providing estimates of odd ratio (OR) or relative risk (RR) or hazard ratio (HR) and the corresponding 95% confidence interval (CI), or information sufficient to calculate them, were included in the meta-analysis. Abstracts, reviews, case reports and articles which did not show efficient information were excluded. If multiple studies were published on the same population or subpopulation, the most recent or informative one would be included in the metaanalysis. The flow chart of the selection of publications included in the meta-analysis is shown in Figure 1. The quality assessment of included studies was further conducted according to the Cochrane handbook 5.1.0 for random control trial (RCT) (http:// www.cochrane.org/training/cochrane-handbook) and Newcastle-Ottawa Scale (NOS) for non-randomized study such as cohort study and case-control study (http://www.ohri.ca/programs/ clinical_epidemiology/oxford.asp).

Data extraction
Information was extracted from all eligible publications independently by two of the authors (Yanwen Yao and Xiaoling Gu). The items included in the data form were as follows: study name (together with the first author's name and year of publication), country, study type, study population, age of the subjects, type of used hormone, study-specific ORs or RRs with 95% CI, detailed numbers of patients for ''non-HRT user versus HRT user'' including ''the short period use versus the long period use'' and matched or adjusted variables in the analysis. For studies from different countries, the ethnicity of subjects was also collected. Most of the estimated associations between HRT use and lung cancer were adjusted for some factors or their combination. According to the suggestion from Chen et al [10], if both the univariate OR/RR and multivariate-adjusted OR/RR were provided, multivariate-adjusted OR/RR was extracted. The lists from the authors were compared, and disagreements were resolved by consensus.

Statistical analyses
OR/RR and corresponding 95%CI or detailed numbers of patients in each included study was pooled to evaluate the association between HRT and lung cancer risk. Whenever available, we used multivariate-adjusted risk estimates; otherwise, we utilized or computed distribution given in the papers the unadjusted ORs. Since not all including studies providing ORs/ RRs but all the detailed numbers of patients were available, we entered the number of patients with HRT use, non-HRT use, lung cancer patients in HRT users and lung cancer patients in non-HRT users to total meta-analysis. Stratified analyses were also carried out by study type, BMI status, histology, smoking status, and menopause type.
Heterogeneity assumption was evaluated with a chi-squarebased Q-test. The summary OR estimate of each study was calculated by a fixed effects model (the Mantel-Haenszel method). If the P value is greater than 0.05 of the Q-test, which indicates low heterogeneity or lack of heterogeneity among the included studies, the fixed effects model is proper, otherwise, the randomeffects model (the DerSimoniane and Laird method) was performed [11,12]. Sensitive analysis was also performed to evaluate the influence of individual studies on the final effect. Potential publication bias would be observed by the funnel plot and formally evaluated with Begg's adjusted rank correlation test and Egger's linear regression test. All statistical analysis were done with the Stata software (version 11.0; StataCorp LP, College Station, TX, USA), using two-sided P values.

Meta analyses on HRT use and lung cancer risk
ORs and 95% CI were not all demonstrated in the included 25 studies. However, the number of lung cancer patients in HRT users and in non-HRT users were provided in all the studies, so the detailed numbers in the two groups were collected and shown in Table 3. In subgroup analyses, ORs and the corresponding 95% CIs were collected and meta-analyses were performed in different subgroups.

Test of heterogeneity
Heterogeneity between studies was observed regarding both overall comparisons and subgroup analyses. According to the I 2 and P for heterogeneity test listed in the Table 4, the random effect model which based on the Mantel-Haenszel method was adopted when I 2 $50% and P for heterogeneity test #0.05. Otherwise, the fixed effect model was adopted.
As shown in Table 4, P for heterogeneity test of lung cancer patients in HRT users vs. non-HRT users for all studies was 0.000 and I 2 was 67.6%, which meant there was heterogeneity for the meta-analysis, so the random effect model was adopted. In the subgroup analyses, meta-analyses on RCTs, BMI,25 kg/m 2 , 25#BMI,30 kg/m 2 , BMI$30 kg/m 2 , small cell lung cancer (SCLC), never smokers, artificial menopause, duration of HRT use#5 years and duration of HRT use.5 years were performed on the fixed effect model, other subgroup analyses were used random effect models. Wu [29] Case-control 4 2 3

Meta-analysis results
The result of the meta-analysis on the all 25 studies and the forest plot was shown in Table 3 and Figure 2. OR was 0.91(95%CI = 0.83 to 0.99) and P value of the test for overall effect was 0.033 when analysis performed using random effect model. This result suggested that HRT use was associated with a significant decrease risk of lung cancer compared with non-HRT use in females. A cumulative meta-analysis was further conducted to confirm the result. As shown in Figure 3, the result was shifted to an association between HRT use and a significant decrease risk of lung cancer when the evidence was tracked over time.
Since heterogeneity was revealed in the overall meta-analysis, subgroup analyses were performed after stratifications of data by study type, BMI status, histology, smoking status, oral contraceptive use, menopause type, and duration of HRT use. The overall ORs and corresponding 95% CI of subgroup analyses were demonstrated in Table 4. In analysis of ten case-control studies, the association between significant decreased lung cancer risk and HRT use was also been revealed, while increased lung cancer risk was correlated with HRT use according to the meta-analysis of four RCTs but the difference is not significant. The forest plot of the study type subgroup analysis is demonstrated in Figure 4-A. In the analysis of different types of hormone replacement therapy, estrogen, progestin or the combination of both were used in participates in 17 studies. The sub-analysis result of these 17 studies revealed that this type of hormone use could reduce the lung cancer risk (OR = 0.83, 95%CI 0.76-0.91, P = 0.000) (Figure 4-B). Five studies in which estrogen was used as the replacement hormone demonstrated that although the difference was not significant, only estrogen use could increase the lung cancer risk (OR = 1.16, 95%CI 0.98-1.37, P = 0.079).
As shown in Table 4 and Figure 5, the positive association between HRT use and decreased lung cancer risk was found in the patients with never smokers patients (OR = 0.86, 95%CI 0.75-0.99, P = 0.042), and BMI,25 kg/m 2 (OR = 0.65, 95%CI 0.53-0.79, P = 0.000). However, HRT use in patients with artificial menopause increased the lung cancer risk (OR = 1.51, 95%CI 1.17-1.94, P = 0.001). A subgroup analysis was also done on HRT use in smokers, but the evidence showed negative correlation between smoking and the risk of lung cancer, P = 0.333.

Publication bias
The publication bias of the literatures was assessed by using Begg's funnel plot and Egger's test. The shape of the funnel plots seemed approximately symmetrical ( Figure 6) and the result of Egger's test did not show any evidence of publican bias, t = 1.91 and P = 0.069 for HRT use vs. non-HRT use.

Sensitivity analysis
Sensitivity analysis was also analyzed as previous study (Table 5  and Figure 7) [32]. As shown in Figure 7, none of the studies appears to be an outlier or has results very different from the rest of the studies. After each study was excluded from the overall meta-analysis, the similar results were obtained, which suggested that the result of the meta-analysis was stable.

Discussion
To date, the current and estimated morbidity and mortality of lung cancer in females are still less than that in males [2]. Although much has been learned about the epidemiology, the reason is still unclear. Besides the different smoking status, the kind and altering levels of endogenous and exogenous hormone in males and females has become to be the subject of the epidemiology study.
As the most common therapy for menopausal females and young females with artificial menopause, HRT has been proved to be associated with increased risk of breast cancer and endometrial cancer and decreased risk of colorectal cancer and coronary heart disease [33][34][35][36][37]. The role of HRT in lung cancer incidence has also been investigated in several studies. Wu et al in their study firstly demonstrated that the menopause type and hormone intake were associated with the lung cancer risk in female [29]. Previous studies in past years revealed the controversial association between HRT and lung cancer risk.
Our meta-analysis of total 25 studies on HRT use and lung cancer risk demonstrated that HRT use in female could decrease the risk of lung cancer. The positive correlation was also identified between decreased lung cancer risk and HRT use in females with BMI,25 kg/m 2 and never smokers, increased lung cancer risk and HRT users with artificial menopause.
Estrogen in HRT was considered to be the most likely candidate for mediating growth-promoting effect in lung cancer, because evidence showed that expression of estrogen receptors (ER) in lung cancer mediated transcriptional responses in lung cancer cells [38][39][40]. On the other side, progestin which is also a type of hormone in HRT is reported to take a protect role against lung cancer [40,41]. In the study of Ishibashi, progestin receptor was revealed to be in about half of lung cancer, and progestin mediated pathways to induce apoptosis and reduce growth in lung cancer cells [41]. These results demonstrated that HRT can act tumor growth-promoting or tumor growth-suppressing effects depending on the components of HRT, the expressions of estrogen receptors and progestin receptors.
Moore et al firstly suggested in his study that according to the respective data, young women (age from 31 to 50 years) presented more often with advanced disease [42]. This may help to explain why HRT users with artificial menopause who were always young women had increased lung cancer risk. The mechanism involved may be estrogen plays a deleterious role in lung cancer in younger female patients, especially in early stage disease [42]. The reason for undefined risk in smoking HRT users may be that higher circulating levels of estrogen in women make susceptible to the carcinogenic influence of tobacco smoke [43].
The analyses showed that the ORs and significance were different in case control studies, cohort studies and RCT. As provided in each article, all females in four RCT studies included in this meta-analysis, were postmenopausal women, with the age range or the mean age of 50-79, 50-79, 69 and 63, respectively. While in cohort and case-control studies, females were younger than those in RCTs, suggesting these females were not in the menopause or were close to the menopause.
Vandenbroucke JP [44] indicated that adjustment for previous use of hormones already increased the estimates in the trials. HRT is usually started close to menopause, so the observational studies had shown a relatively true situation. However, this situation was diluted in the randomized trial because fewer women close to menopause were enrolled. The discrepancy between the RCT and observational study was not only in HRT and lung cancer, but also in HRT and coronary heart disease, and breast cancer. It was not due to differences in study design, but to the timing of start of treatment.
Randomized trials are necessary for showing whether the benefit of a medical intervention exists [45]. In contrast, observational study is often used to investigate adverse effects [46]. It was suggested that the same adverse effect for the same treatment can rarely be investigated by observational research and in very large randomized trials [47], such as breast cancer and HRT, and lung cancer and HRT shown in our meta-analysis. Vandenbroucke JP also suggested that observational studies may better reflect the true harm in real-life than selected populations in randomized trials [44].
In the subgroup analyses HRT use was confirmed to be associated with decreased lung cancer risk in non-smoking females. However, HRT use in ever smoking females failed to be associated with lung cancer risk. The fact that smoking can induced increased lung cancer risk in females could help to explain the result. In a       study on smoking-related mortality in the United States, smoking was shown to be also hazardous for women comparable to men [48]. Since smoking and other socio-economic status were proved to be associated with increased lung cancer risk, the effect of HRT may be potentially confounding by these factors.
The present study provides a quantitative analysis of available epidemiologic studies on HRT and lung cancer risk in females. There is a broad time span in the 25 studies, from 1988 to 2012, and the sample size is large (total 656,403 participates). Stratified analyses were conducted. Heterogeneity was tested and randomeffects model was used to estimate the magnitude of the heterogeneity and assign a greater variability to the combined risk estimate to account for the heterogeneity if detected. Publication bias was also examined and corrected if detected.
Some limitations of this meta-analysis should be acknowledged. First, many observational studies which were susceptible to various biases were included in our meta-analyses. The subgroup analysis in different types of studies showed different results. However, taking these results together still shows the positive association between HRT and lung cancer risk. Second, although subgroup analyses was proceed by the kind of HRT, HRT dose which may also affects the result were not taking into analysis. Finally, only published studies were included in our meta-analysis. Therefore, publication bias may be occurred although no publication bias was indicated from both funnel plot and Egger's test.

Conclusion
Taken together, our analysis of currently available studies on HRT and lung cancer risk suggests that HRT use is correlated with decreased lung cancer risk in female, especially in female with BMI,25 kg/m 2 and never smokers, but HRT in female with artificial menopause could increase the lung cancer risk. Considering that lung cancer is one of the leading causes of death in female, further investigations or larger observational study with women closer to menopause rather than postmenopausal was  needed in future to further validate the influence of HRT in lung cancer.

Supporting Information
Checklist S1 (DOC) Author Contributions