Lung cancer mortality risk among breast cancer patients treated with anti-estrogens

Abstract

BACKGROUND:

The Women's Health Initiative randomized clinical trial reported that menopausal hormone therapy increases lung cancer mortality risk. If this is true, use of anti-estrogens should be associated with decreased lung cancer mortality risk. The authors compared lung cancer incidence and mortality among breast cancer patients with and without anti-estrogen therapy.

METHODS:

Our study included all 6655 women diagnosed with breast cancer between 1980 and 2003 and registered at the Geneva Cancer Registry. Among these women, 46% (3066) received anti-estrogens. All women were followed for occurrence and death from lung cancer until December 2007. The authors compared incidence and mortality rates among patients with and without anti-estrogens with those expected in the general population by Standardized Incidence Ratios (SIRs) and Standardized Mortality Ratios (SMRs).

RESULTS:

After a total of 57,257 person-years, 40 women developed lung cancer. SIRs for lung cancer were not significantly decreased among breast cancer patients with and without anti-estrogens (0.63, 95% confidence intervals [CI], 0.33-1.10; and 1.12, 95% CI, 0.74-1.62, respectively) while SMR was decreased among women with anti-estrogens (0.13, 95% CI, 0.02-0.47, P<.001) but not for women without anti-estrogens (0.76, 95% CI, 0.43-1.23).

CONCLUSIONS:

Compared with expected outcomes in the general population, breast cancer patients receiving anti-estrogen treatment for breast cancer had lower lung cancer mortality. This study further supports the hypothesis that estrogen therapy modifies lung cancer prognosis. Cancer 2011. © 2011 American Cancer Society.

The incidence of lung cancer has increased strongly among women in Switzerland and other industrialized countries because of the smoking epidemic among young women.1 Despite treatment progress, survival remains poor with only <16% of patients surviving more than 5 years after diagnosis.2 Although smoking is the main cause for lung cancer occurrence, other factors may act as modulators of the risk or the natural history of the disease. In particular, hormonal factors that play an important role in lung development, and maturation could also be involved in lung carcinogenesis. There is biological evidence that some lung cancer cells express active hormone receptors,3, 4 and experimental studies report growing evidence that estrogens may promote lung tumor occurrence and progression.5–8 However, studies evaluating the association between menopausal hormone use containing estrogen alone, or combined estrogen and progestin and lung cancer incidence9–20 or mortality,16, 20–25 provide conflicting results. Results from randomized controlled trials showed similar risk of lung cancer incidence13, 20 but reported an increased risk of lung cancer mortality among women with versus without hormone replacement therapy (HRT).20 In particular, the Women's Health Initiative randomized clinical trial reported a 60% (hazard ratio [HR], 1.59, 95% confidence interval [CI], 1.03-2.46) increased risk of dying from nonsmall cell lung carcinoma (NSCLC) among women in the hormone therapy arm versus women in the placebo arm.20

If exposure to estrogen increases a woman's risk of dying as a result of lung cancer, then anti-estrogens may, on the contrary, potentially reduce this risk. Women with breast cancer offer a unique possibility to examine the role of anti-estrogens on lung cancer incidence and mortality because many of these women will receive anti-estrogens (tamoxifen or aromatase inhibitors) as part of their multimodality treatment. In this population-based study, we evaluated lung cancer incidence and mortality risk among breast cancer patients with and without anti-estrogen therapy.

MATERIALS AND METHODS

Study Population

We used information from the population-based Geneva Cancer Registry, which records all incident cancer cases occurring in the population of the canton of Geneva (approximately 440,000 inhabitants) since 1970. The Cancer Registry extracts information from various sources and is considered accurate; it has a very low percentage (<2%) of cases recorded from death certificates only.1 All hospitals, pathology laboratories, and private practitioners in the canton are requested to report every cancer case. Trained tumor registrars systematically extract data from medical and laboratory records. Physicians regularly receive inquiry forms to complete missing clinical and therapeutic data.

Recorded data include sociodemographic information, method of detection, type of confirmation, tumor characteristics (coded according to the International Classification of Diseases for Oncology, ICD-O),26 hormone receptor status, stage of disease at diagnosis, treatment during the first 6 months after diagnosis, survival status, and cause of death.27

The Cancer Registry regularly assesses survival. The index date refers to the date of confirmation of diagnosis or the date of hospitalization when it preceded the diagnosis and was related to the disease. In addition to passive follow-up (routine examination of death certificates and hospital records), active follow-up is performed yearly using the files of the Cantonal Population Office in charge of the registration of the resident population. Cause of death is systematically recorded and validated by consulting medical files or, when necessary, by sending a specific questionnaire to the patient's physician. Cause of death is coded according to the International Statistical Classification of Diseases and Related Health Problems (ICD-10).27

In the current study, we included all patients diagnosed with invasive breast cancer between 1980 and 2003. We excluded breast cancer cases diagnosed at death (n = 60). The final cohort consisted of 6655 breast cancer patients.

Because of differences in data availability, we separated the period of diagnosis 1980-1990, when information was less exhaustive, from the period 1991-2003. Socioeconomic status was regrouped in 3 levels: low (manual employees, skilled and unskilled workers, including farmers), middle (nonmanual employees and administrative staff), and high (professionals, executives, administrators, entrepreneurs) based on the patient's last occupation or, if unemployed, that of the spouse. Familial risk was categorized as high (at least 1 first-degree relative with breast or ovarian cancer diagnosed before the age of 50 years), low (no affected first- or second-degree relatives with breast or ovarian cancer), or moderate (all other known family histories). Staging was based on the pathologic tumor-node-metastasis (TNM) classification or, when absent, the clinical TMN classification.28

Method of discovery was considered as consultation following symptoms, mammography or clinical screening, and other. Breast cancer histology was classified as ductal carcinoma, lobular carcinoma, other, and unknown (no microscopic confirmation). Hormone receptor status was classified as positive (≥10% of tumor cells expressing receptors) or negative (<10% tumor cells expressing receptors). Treatment was classified as surgery (breast-conserving surgery, mastectomy), radiotherapy (yes, no), chemotherapy (yes, no), and anti-estrogen therapy (yes, no). During the study period anti-estrogen therapy consisted mainly of tamoxifen because aromatase inhibitors were prescribed in Switzerland only from 2006.

As information on smoking exposure is not routinely recorded in the cancer registry database, we retrospectively extracted this data from medical files. Women were classified as never or ever smokers and within the latter group as former smokers (defined as individuals who stopped smoking at least 1 year before the breast cancer diagnosis) or current smokers (defined as individuals who smoked at the time of breast cancer diagnosis). Tobacco consumption was expressed as number of pack-years smoked (calculated by multiplying the number of packs of cigarettes smoked per day by the number of years the person has smoked).

Statistical Methods

We compared patient and tumor characteristics among women with versus without tamoxifen by chi-square test of heterogeneity.

Patients were followed for lung cancer incidence and mortality from 6 months after the date of breast cancer diagnosis until December 31, 2007. Person-years at risk were calculated to the date of end of follow-up, date of departure from the canton, date of lung cancer diagnosis, or date of death, whichever came first.

Expected numbers of lung cancer cases and lung cancer deaths were calculated on the basis of the cantonal lung cancer incidence and mortality rates for each 5-year age group and calendar year. Standardized Incidence Ratios (SIRs) and Standardized Mortality Ratios (SMRs) were then calculated by dividing the observed numbers by the expected numbers. Statistical significance and 95% confidence intervals were estimated assuming a Poisson distribution.

We also modeled SIRs and SMRs by fitting Poisson regression models using the natural logarithm as a link and including the natural logarithm of the expected number of events as a fixed offset.29 Variables significant in univariate analysis or most biologically relevant for developing or dying from lung cancer were entered into the multivariate Poisson models. In addition to anti-estrogen treatment, we included age and period at diagnosis, estrogen receptor status, radiotherapy and chemotherapy use. The model fitted for the SMRs did not include estrogen receptor status, as the variable had strata with 0 events.

RESULTS

Of the 6655 breast cancer patients, 3066 (46%) received anti-estrogen therapy and 3589 (54%) did not (Table 1). Anti-estrogen therapy was more frequently administered in the most recent period of the study (86% in 1991-2003). Compared with women without hormonal therapy, those with anti-estrogen therapy were significantly older (mean age at diagnosis 64 and 60 years, respectively; P < .001) and postmenopausal (79% and 68%, respectively; P < .001). As expected, tumors of women who received anti-estrogen therapy were more likely to express estrogen and progesterone receptors. Women with anti-estrogen therapy more often underwent radiotherapy and less often underwent surgery or chemotherapy than women without anti-estrogens.

Table 1. Characteristics of Breast Cancer Patients by Anti-estrogen Therapy Use. Geneva Cancer Registry, 1980-2003.

Patient characteristics	Anti-estrogen Therapy
	Yes (n = 3066)		No (n = 3589)		Pa
	No.	%	No.	%
Percentages are of nonmissing data. Numbers may not sum to 100% due to rounding and missing data. a Chi-square test for heterogeneity. b Recorded since 1995. c Recorded since 1990.
Period of cancer diagnosis					<.001
1980-1990	417	14	2061	57
1991-2003	2649	86	1528	43
Method of detection					<.001
Symptoms	777	25	2015	56
Clinical examination or screening	1059	35	626	17
Other	1230	40	948	26
Age, y, at diagnosis					<.001
<40	80	3	236	7
40-49	347	11	799	22
50-59	778	25	882	25
60-69	798	26	682	19
70-79	598	20	602	17
<80	465	15	388	11
Menopausal statusb					<.001
Pre- and peri-	435	21	293	32
Post	1675	79	614	68
Civil status					<.001
Single	420	14	490	14
Married	1474	48	1939	54
Widowed	714	23	652	18
Separated	458	15	508	14
Social class					.130
Low	425	17	395	15
Middle	1513	59	1553	59
High	616	24	685	26
Familial riskc					.077
None	1805	71	1064	74
Moderate	561	22	293	20
High	165	7	74	5
Tumor characteristics
Stage					.083
I	1057	36	1244	37
II	1351	46	1522	46
III	320	11	377	11
IV	221	7	199	6
Histological subtype					<.001
Ductal	2258	76	2787	80
Lobular	362	12	206	6
Other	370	12	505	14
Differentiation					<.001
Good	819	33	474	23
Moderate	1223	49	898	44
Poor	457	18	676	33
Diameter					.493
1 - 20 mm	1472	61	1733	60
21 - 40 mm	749	31	936	32
> 40 mm	196	8	220	8
Estrogen receptor statusb					<.001
Positive	1956	97	393	51
Negative	63	3	385	49
Progesterone receptor statusb					<.001
Positive	1592	79	325	42
Negative	430	21	454	58
Treatment
Radiotherapy					<.001
No	1005	33	1853	52
Yes	2061	67	1736	48
Surgery					<.001
No	482	16	369	10
Yes	2584	84	3220	90
Chemotherapy					0.28
No	2090	68	2355	66
Yes	976	32	1234	34

From medical records, we retrieved data on tobacco use for 3322 women, ie, 50% of the study population (Table 2). Approximately 57% of the missing data applied to women with breast cancer diagnosed during the first part of the study period. The proportion of never smokers was very similar among women who received anti-estrogen therapy (63%) and those who did not (65%), and no differences were found between the 2 groups of therapy, neither for smoking status (former or current) nor for the number of pack-years smoked.

Table 2. Smoking Status of Breast Cancer Patients According to Anti-Estrogen Use

	Anti-estrogen Therapy
	Yes		No
	n=2105		n=1217
	No.	%	No.	%	P
a Chi-square test for heterogeneity between never smokers and ever smokers. b Percentages are calculated among ever smokers. c Chi-square test for heterogeneity between former and current smokers. d Among current smokers.
Smoking status at time of cancer diagnosis
Never smoker	1325	63	785	65	.367a
Ever smoker	780	37	432	35
Former smoker	284	36b	143	33b	.249c
Current smoker	496	64b	289	67b
Pack-years of smokingd					1.00
1-20	181	42	108	42
21-40	155	36	92	36
>40	93	22	55	22

The median follow-up period for the whole cohort was 7.3 years. The cohort yielded 57,257 person-years (21,677 among anti-estrogen users and 35,580 in nonusers). From July 1980 until December 2007, we observed a total of 40 lung cancer cases occurring at least 6 months after breast cancer diagnosis. The incidence rate of lung cancer was 55.4 per 100,000 person-years for women with anti- estrogens versus 78.8 per 100,000 for women without (P = .39; Table 3). Compared with the general population, the risk (SIR) of developing lung cancer among women who received anti-estrogens was 0.63 (95% CI, 0.33-1.10) and among women without anti-estrogens 1.12 (95% CI, 0.74-1.62; Table 3). Lung cancer mortality rates were 9.2 of 100,000 for women with anti-estrogens and 45.0 of 100,000 for women without anti-estrogens (P = .026; Table 4). Compared with the general population, lung cancer mortality risk (SMR) was 0.13 (95% CI, 0.02-0.47) among women who received anti-estrogens (P < .001) and 0.76 (95% CI, 0.43-1.23) among women who did not. The SIR predicted from the multivariate Poisson regression model, adjusted for age and period at diagnosis, estrogen-receptor status, and radiotherapy use, was 1.60 (95% CI, 0.21-12.1) for women with anti-estrogen and 2.47 (95% CI, 0.48-12.7) for women without. The SMRs adjusted for the same variables, except estrogen receptor status, did not change markedly from those in the univariate analysis, indicating that these results are likely not due to confounding variables. In particular, the SMR for women with anti-estrogens was 0.14 (95% CI, 0.02-1.23), and the SMR for women without anti-estrogens was 0.88 (0.23-3.41). All the estimates were no longer significant.

Table 3. Risk of Lung Cancer After Breast Cancer Among Women With and Without Anti-Estrogens. Geneva Cancer Registry 1980-2003

Anti-estrogen Therapy	Incidence Ratesa	Observed Cases	Expected Cases	SIR	95% CI	P
SIR indicates age-standardized incidence ratio; CI, confidence interval. a Age-standardized (Geneva population distribution) incidence rates per 100,000 person-years.
Yes	55.4	12	19.1	0.63	0.33-1.10	.116
No	78.8	28	25.0	1.12	0.74-1.62	.600

Table 4. Risk of Mortality From Lung Cancer After Breast Cancer Among Women With and Without Anti-Estrogen Use. Geneva Cancer Registry 1980-2003

Anti-estrogen Therapy	Mortality Ratesa	Observed Cases	Expected Cases	SMR	95% CI	P
SMR indicates age-standardized mortality ratio; CI, confidence interval. a Age-standardized (Geneva population distribution) mortality rates per 100,000 person-years.
Yes	9.2	2	15.3	0.13	0.02-0.47	<.001
No	45.0	16	21.1	0.76	0.43-1.23	.316

DISCUSSION

In analyses comparing tumor registry to population results from standardized mortality ratios, we found that anti-estrogen treatment for breast cancer was associated with a reduced risk of death from lung cancer, providing new evidence on the role of estrogen in lung cancer progression. These findings are unlikely attributable to smoking differences, as the exposure to tobacco among women who received anti-estrogen therapy and those who did not was almost identical. In addition, patterns of tobacco use in our study population (never smokers 57%, former smokers 14%, and current smokers 29%) were very comparable to that of the general female population of Switzerland in 2003 (55%, 16%, and 29%, respectively).30 However, we retrieved information on smoking history for approximately only half of the study population. Also, we have no information on changes in smoking habits after the breast cancer diagnosis. Missing information mainly applied to older women diagnosed in the years 1980-1990. Because smoking was uncommon in the Swiss female population 2 to 3 decades ago, it is likely that most of the women with missing data on smoking were never smokers. Indeed, when stratifying by tobacco use, we found that women with unknown smoking status presented with lung cancer incidence rates similar to never smokers. For women with anti-estrogens, the SIRs among never smokers and women with unknown smoking status were 0.35 (95% CI, 0.07-1.02) and 0.33 (95% CI, 0.04-1.19), respectively. For women without anti-estrogens, the correspondent SIRs were 0.55 (95% CI, 0.11-1.61) and 0.47 (95% CI, 0.20-0.93), respectively. These findings, plus the results of our sensitivity analysis only considering women diagnosed after 1990, confirm that differences in smoking exposure cannot explain our results. A limitation of our study comes from the type of analytical approach used to compare Cancer Registry data with results expected in the general population, which can also be subject to potential, random variations and confounding. Additional limitations of this study include the few number of events, possible biases and misclassifications of data, incomplete data, and lacking type, dosage, and duration of anti-estrogen. Strengths of the study are its prospective design in a population-based cohort and the accuracy of death certificates because trained tumor registrars establish both the cause of death and the presence of the tumor at death for all diseased patients by systematically consulting clinical records and/or interpreting questionnaires filled in by the patient's physician.

Like other studies, we found no significant association between lung cancer incidence and anti-estrogen therapy.31 However, we cannot exclude a protective role of anti-estrogens on lung cancer risk, as women with anti-estrogens presented a nonsignificant lower risk of developing lung cancer, despite that another well-known lung cancer risk factor, radiotherapy, was more often given to women with anti-estrogen therapy than to women without (67% vs 48%). The small sample size and the short follow-up could be partly responsible for this lack of results.

The decreased risk of death from lung cancer among women who received anti-estrogens is compatible with the findings of higher lung cancer mortality rates among menopausal hormone users reported by recent studies.9, 20, 21

The association between hormone replacement therapy (HRT) and higher lung cancer mortality rates has not always been observed. A lack of association was found in 2 studies,22, 31, 32 whereas a protective effect of HRT on lung cancer mortality was reported in another 2 studies.16, 23 This apparent lack of coherence may be explained by the use of diverse study designs and important differences in smoking levels among the examined populations. In fact, HRT seems to act as a risk factor or prognostic factor of lung cancer only among smokers,12, 14, 16, 21 or more importantly, type of HRT used. Indeed, it has recently been determined that the effect of HRT is different according to its formulation, with estrogen plus progesterone combination being associated with increased lung cancer incidence9 and mortality,20 whereas no association has been found with estrogen alone.22

From a biological perspective, the observations that estrogen intake is associated with increased lung cancer mortality21, 24, 25 and that anti-estrogen treatment is associated with a decreased lung cancer mortality, as demonstrated in this study, strongly suggest that estrogens are involved in lung cancer progression. Previous studies have reported that a great majority of NSCLC, which represents the majority of lung cancer cases, expressed estrogen receptors (ER) alpha and beta.6–8, 32 Experimental studies have also shown that estrogens stimulate NSCLC cells through ER-mediated signaling⁵,7, 8 and that anti-estrogens, such as tamoxifen or raloxifene, inhibit tumor growth induced by estrogen.6, 33 A study examining the level of protein expression of aromatase in 422 patients with NSCLC reported an association between a high level of aromatase expression and worse lung cancer prognosis.34 Finally, a recent study found several associations between single-nucleotide polymorphisms in a set of genes involved in the pathways of sex steroid synthesis and development of NSCLC in women.35

Our a priori hypothesis was that if estrogens have a real detrimental effect on lung cancer outcome, then anti-estrogens would affect lung cancer prognosis. Albeit comparisons across 2 separate populations should require cautious interpretation, our results show that breast cancer patients who received anti-estrogens have a decrease in lung cancer mortality, reinforcing the evidence that estrogen plays a key role in lung cancer progression. Existing large randomized clinical trials for adjuvant breast cancer therapy and for primary prevention with tamoxifen are the ideal settings for more definitive assessment regarding the role of anti estrogens on lung cancer outcome.

Acknowledgements

We thank Stina Blagojevic, for technical and editorial assistance, and the Cancer Registry team for providing data and support.

CONFLICT OF INTEREST DISCLOSURES

The authors made no disclosures.