Determinants of long term survival after surgery for cancer of the lung

A population-based study

Authors


Abstract

BACKGROUND

Even if some determinants of lung cancer (LC) prognosis have been established, their independent effect on long term survival remains to be seen. The objective of the current study was to identify the prognostic indicators of long term survival among LC patients treated by surgery.

METHODS

All patients with LC recorded at the Geneva Cancer Registry between 1977 and 1987 were analyzed by logistic regression, considering as cases (n = 98) those patients alive 10 years after their initial diagnosis and as controls (n = 330) all other patients, excluding those who did not undergo putative curative surgery. The effect of each prognostic factor was evaluated after accounting for age and gender (“crude” effect) and also for other a priori confounding factors (adjusted effect). Additional models considered two staging variables simultaneously to identify the strongest staging determinant. Results were presented as relative risk estimates of long term (≥10 years) survival.

RESULTS

Age, histology, and stage of disease significantly influenced prognosis regardless of the confounding factors considered. Gender also emerged as a discriminated factor in LC outcome, with a 2.1-fold increased chance (95% confidence interval, 1.6–3.5) of long term survival for women compared with men. Method of discovery, presence of symptoms, period of diagnosis, socioeconomic status, and tumor differentiation did not appear to be associated with long term survival. Extent and size of the tumor were found to be the most reliable prognostic staging factors, whereas adenopathy had no effect after accounting for extension.

CONCLUSIONS

The current population-based study quantifies the independent effect of the factors modifying the chances of curability in patients with LC. In particular, it provides additional evidence that gender strongly influences long term survival. Cancer 1999;86:2229–37. © 1999 American Cancer Society.

Lung cancer is a major health problem in Switzerland, where it represents the leading cause of cancer-related deaths among males. It is estimated that each year approximately 2500 men and 700 women are diagnosed with lung cancer and that 2300 males and 500 females died of the disease between 1989–1993.1 The age-standardized mortality rates (European standard population, per 100,000) (73.5 for males and 16.3 for females) are of European average for both genders.2 The mortality to incidence ratio of lung cancer is persistently one of the highest among cancers at different sites (i.e., >0.8 in both genders), thus making rate of incidence patterns very similar to mortality patterns.1 Despite the fact that Geneva has particularly high lung cancer survival rates,3 the overall prognosis for patients with lung cancer remains poor, with <10% of patients alive after 10 years.4 Among factors of possible prognostic significance in lung cancer survival, the age of patient and the extent of the disease at diagnosis exert the strongest influence in long term prognosis. Based on clinical studies, gender recently has emerged as a discriminating factor in nonsmall lung cancer outcome.5–7 Other factors, such as socioeconomic status and histologic differentiation, appear to influence survival but are not established prognostic factors. The aim of the current study was to identify the independent prognostic indicators of long term survival in cancer of the lung using data from a population-based cancer registry. This study is limited to patients treated by surgery because surgical resection to our knowledge to date is the only treatment that may offer a cure for patients with lung cancer.

MATERIALS AND METHODS

The data considered for the current analysis were derived from the Geneva Cancer Registry dataset, which included information regarding incident cases of malignant neoplasms occurring in the canton (approximately 400,000 inhabitants in 1996). The registration is based on several sources of information to be as exhaustive as possible, as attested by its very low percentage (<2%) of cases recorded from death certificates only.8 Notification is based on a voluntary agreement between the recording medical institutions of the canton and the registry. All hospitals, pathologic laboratories, and practitioners are required to report all current and past cases of cancer. Data are abstracted systematically from medical and laboratory records by trained tumor registrars. Inquiry forms are addressed regularly to private practitioners to secure missing clinical data. Death certificates are consulted systematically.

Recorded data include sociodemographic information on patients (age, gender, place of birth, marital status, address at time of diagnosis, and occupation), diagnostic circumstances (origin of diagnosis, presence of symptoms, and methods of assessment), tumor characteristics (primary site, histologic type, and differentiation coded according to the International Classification of Diseases for Oncology [ICD-O]),9 stage of disease at diagnosis, treatment during the first 6 months (surgery, radiotherapy, or chemotherapy), finality of treatment (curative, palliative, or not specified), survival status, and cause of death.

The survival assessment performed by the registry was described previously.10 In brief, the index data for rate of incidence refer to the date of confirmation of diagnosis, or to the date of hospitalization if it preceded the diagnosis and was related to the disease. The registry performs two types of follow-up. The passive follow-up is comprised of routine examination of death certificates and of incidental findings in hospital records. Some deaths are discovered when the cancer registry files are updated according to the national mortality data files. Active follow-up is performed systematically, to validate passive follow-up and to record possible departures. The files of the Cantonal Population Office are consulted regarding the vital status of all living patients at successive 5-year intervals after the incidence data. For decreased patients the cause of death is established from clinical records and recorded systematically according to the World Health Organization classification (WHO).11

Selection of Patients

Between 1977 and 1987, 2027 patients with histologically proven primary cancer of the lung (ICD-O codes 162.2–162.9) were registered in the Swiss canton of Geneva. Only invasive cases (ICD-O five-digit behavior code 3) were included. Cases diagnosed at autopsy (n = 46; 2.3%), dead on the day of diagnosis (n = 2), or lost to follow-up (n = 41; 2.0%) were excluded. According to the aim of the study, which was to determine prognostic factors of long term survival, we included only patients with a priori curative surgery (i.e., surgery with total macroscopic or microscopic removal). Therefore we excluded all other patients: nonoperated patients (including nonoperable cases; n = 1405; 69.3%), patients with palliative surgery (i.e., surgery without total macroscopic or microscopic removal; n = 79; 3.9%), and patients with unspecified surgery (surgery with no specification of being curative n = 26; 1.3%). All patients excluded had hardly any chance of being cured, with only 9 survivors 10 years later (5 in the nonoperated, 2 in the palliative surgery, and 2 in the unspecified surgery groups). The series finally included 428 lung cancer patients treated by surgery presumed to be curative. Thus, such a curative approach unfortunately concerned a minority (approximately 20%) of all lung cancers arising in the general population.

Variables Considered

Four variables were used to characterize the stage of the tumor at diagnosis. The extent of the tumor and metastasis to lymph nodes (hereafter adenopathy) were coded according to the Surveillance, Epidemiology, and End Results (SEER) classification based on both postsurgical and clinical information.12 The extent of the primary tumor was studied in three categories: localized (localized with no other specification [NOS], single tumor or multiple masses confined to one lung and/or main stem bronchus), direct regional extension (pleura; visceral, NOS; pericardium; parietal, NOS; carina; trachea; esophagus; nerves; major blood vessels; and mediastinal extension NOS), and distant extension (brachial plexus from superior sulcus or Pancoast tumor; lung and/or main stem controlateral bronchus; pericardium, visceral; heart; pleura, parietal; and other direct distant extensions) or metastasis (distant lymph nodes and other distant involvement). The presence of adenopathy (clinical, surgical, or histologic invasion of regional lymph nodes) was coded into three categories (positive, negative, and unknown). The extent of the primary tumor also was studied using the size of tumor after resection (in cm) in 5 categories (≤2, ≤3, ≤4, >4, and unknown) and the T classification of the clinical TNM classification13, 14 into 4 groups: T1 (≤3 cm/no invasion), T2 (>3 cm/extension to the hilar region), T3 (macroscopic extension/effusion/atelectasis), and TX (unknown/unclassifiable/inapplicable).

Specific histology codes were grouped into cell types according to the WHO specifications for typing of lung tumors.15 Squamous cell carcinoma included tumors coded 8070–8074 (squamous cell, NOS; keratinizing type; nonkeratinizing type; and spindle cell). Small cell carcinoma referred to tumors coded 8041–8043 (oat cell, intermediate cell type, or combined oat cell). This group also included a few tumors (five cases in this series) coded 8021 (carcinoma, anaplastic type). Adenocarcinoma included all tumors coded 8140, 8230, 8250, 8260, 8480, 8481, and 8550 (adenocarcinoma NOS; acinar; papillary; bronchiolo-alveolar; solid carcinoma with mucus; and mucinous or mucin-producing adenocarcinoma). Large cell carcinoma included morphology codes 8012, 8031, and 8310 (large cell, NOS; giant cell; or clear cell). Carcinoid carcinoma coded 8240, 8241, 8242, 8243, and 8244 (malignant carcinoid, argentaffin, nonargentaffin, goblet cell, and composite) also were considered separately. The last group other histology referred to rare subtypes such as adenoid cystic carcinoma (ICD-O code 8200; three cases in the series), malignant fibrous histiocytoma (8330; one case), pulmonary blastoma (8981; one case), and hemangiopericytoma (9150; one case), as well as tumors that could not be classified in other histologic groups, such as carcinoma, NOS (8010; two cases), nonsmall cell undifferentiated carcinoma (8020; five cases), adenosquamous cell carcinoma (8560; seven cases), adenoid cystic carcinoma (8200; three cases), mucoepidermoid carcinoma (8430; five cases), giant and spindle cell carcinoma (8030; two cases), and adenoid squamous cell carcinoma (8075; one case).

The differentiation of the tumor was coded according to the ICD-O code for histologic grading and differentiation as follows: Grade 1 (well differentiated, differentiated), Grade 2 (moderately differentiated, moderately well differentiated), Grade 3 (poorly differentiated), Grade 4 (undifferentiated, anaplastic), and unknown. This variable was available only for cases recorded after December 31, 1978.

The variable “occupation” (the patient's last occupation or, for unemployed women, that of the spouse) primarily coded in three digits, according to the Federal Office of Statistics' classification of vital statistics,16 was redefined as a socioeconomic indicator in four levels: low (manual employees and skilled and unskilled workers), middle (nonmanual employees and administrative staff), high (professionals, executives, and administrators), and unknown. The sector of health care was examined as follows: private (initial treatment and follow-up in private sector only), public (initial treatment and follow-up in public sector only), and mixed (both systems were involved).

The origin of diagnosis was recorded for all incoming cases: consultation after symptoms, screening, or check-up examination; fortuitous discovery during consultation; and unknown. For lung cancer screening the procedure mainly referred to a routine chest X-ray examination.

Statistical Analysis

All individuals (except 3 patients lost to follow-up) were followed for at least 10 years after their diagnosis. Because the curability of the tumor was the variable of interest, we chose to model the long term survival status instead of a more traditional Cox survival analysis.

All records on selected lung cancer patients were analyzed through unconditional multivariate logistic regression, considering patients remaining alive at 10 years as cases and patients with shorter survival as controls.17 The prognostic factor identified therefore concerns the probability of being alive at 10 years. However, it also would concern the risk of death during the 10-year period if the distribution of death over time is similar in the absence and in the presence of the prognostic factor. All models were log-linear, fitted using the generalized linear interaction modeling statistical package.18 Prognostic factors of interest were, alternatively, gender, age at diagnosis (<50 years, 50–59 years, 60–69 years, and ≥70 years), place of birth (Switzerland or other), socioeconomic status, period of diagnosis (1977–1981, 1982–1987), origin of diagnosis (screening or other methods), presence of symptoms at diagnosis (yes, no, or unknown), stages at diagnosis, histologic types, and differentiation. The models contained the factor of interest and the a priori-determined confounding factors: age (continuous) and gender for estimation of the “crude effect” ; and age, gender, socioeconomic status (low or other), period, method if discovery, extension (localized or other), and histology (squamous, adenocarcinoma, carcinoid, or other) for estimation of the adjusted effect. We also a priori defined additional models considering simultaneously two staging variables to identify the strongest staging determinant. The significance of each variable of interest was assessed by comparing the goodness of fit measure (deviance) of the model with and without the variables of interest.

Results are presented as relative risk estimates of long term survival versus < 10 years of survival. The analysis also was performed on patients considered long term survivors at 5 years to determine possible differences.

RESULTS

Among 428 lung cancer patients who underwent a priori curative surgical treatment, 98 still were alive 10 years after the diagnosis (the cases) and 330 had died within this period (the controls). The causes of death among the 282 male controls were lung cancer (n = 223, 79%), other tumors (n = 12, including 3 from upper aerodigestive tumors), cardiovascular (n = 20), cerebrovascular (n = 2), pulmonary (n = 5, including 3 from chronic obstructive diseases), digestive (n = 4, including 3 from cirrhosis) and infectious diseases (n = 4), suicide (n = 3), and ill-defined causes (n = 9). Among the 48 female controls, causes of death were lung cancer (n = 43, 90%), other tumors (n = 3), rheumatologic diseases (n = 1), and ill-defined causes (n = 1). Greater than 50% of the deaths occurred before 2 years in both males and females. The distribution of death during the 10-year period of follow-up was similar for both genders (data not shown).

The distribution of lung cancer patients by long term survival was presented according to sociodemographic indicators (Table 1), method of discovery and health care sector (Table 2), and tumor characteristics (Table 3). The long term survivors were younger (mean age at diagnosis, 56 years; standard deviation [SD]: 10 years) than other lung cancer patients (mean age, 60 years; SD: 9 years). Overall, 82% of patients were men. The proportion of males was lower among long term survivors (72%) than among the control population (85%). The distribution of the other sociodemographic characteristics considered was similar regardless of long term survival status (Table 1). The relatively high proportion of patients born outside Switzerland (30%) reflected the cosmopolitan population of the canton. Only 14% of lung tumors were diagnosed by a screening procedure and the tumor-related symptoms were specified as present among 40% of patients (Table 2). Approximately 14% of lung cancer patients were diagnosed and treated in private institutions only (Table 2). Similar proportions were found among long term survivors and other patients. In the current series of lung cancer patients treated by surgery, squamous cell carcinoma was the overall predominant histologic type (57%) (Table 3) and also was predominant among males (63%). Among females, adenocarcinoma (46%) surpassed all other histologic types. The proportion of both squamous cell and adenocarcinoma was similar for long term survivors (51% and 29%, respectively) and other lung cancer patients (59% and 25%, respectively). The same observation could be made for small cell carcinoma and large cell carcinoma, representing 5% and 1%, respectively, of tumors among long term survivors and 2% and 4%, respectively, of tumors among other patients. Approximately 50% of the 18 malignant carcinoids were low grade. Carcinoid cases accounted for higher percentages among long term survivors (12%) than among other patients (2%). The opposite was observed for unclassifiable or rare types, accounting for 2% of cases among long term survivors and 8% among other patients.

Table 1. Distribution of Lung Cancer Patients by Survival Status and Estimation of the Effect of Sociodemographic Characteristics on Long Term Survival
 Survival status after 10 yearsCrude effectAdjusted effect
Alive (cases) n = 98Dead (controls) n = 330ORa(95% CI)ORb(95% CI)
  • OR: odds ratio; 95% CI: 95% confidence interval.

  • a

    Odds ratio adjusted for age (continuous) and gender.

  • b

    Odds ratio adjusted for age (continuous), gender, period (1977–1981/1982–1987), socioeconomic status (low/other), method of discovery (screening/other), extent of tumor (localized/other), and histologic groups (squamous cell/adenocarcinoma/carcinoid/other).

  • c

    Odds ratio not adjusted for age.

  • d

    Reference category.

  • e

    P < 0.05.

  • f

    P < 0.01.

  • g

    Odds ratio not adjusted for gender.

  • h

    Odds ratio not adjusted for period.

  • i

    Odds ratio not adjusted for socioeconomic status.

Age group (yrs)c
 < 5025411.0d1.0d
 50–59311080.5e(0.3–1.0)0.5(0.3–1.1)
 60–69321320.4f(0.2–0.8)0.5e(0.2–0.9)
 ≥7010490.4e(0.2–0.9)0.4e(0.2–1.0)
Genderg
 Male712821.0d1.0d
 Female27482.0e(1.6–3.5)2.1e(1.1–3.7)
Period of diagnosis
 1977–1981381571.0d1.0dh
 1982–1987601731.4(0.9–2.2)1.3(0.8–2.1)
Civil status
 Ever married883101.0d1.0d
 Other10201.4(0.6–3.2)1.8(0.8–3.9)
Country of birth
 Switzerland592401.0d1.0d
 Other39901.5(0.9–2.5)1.3(0.8–2.3)
Socioeconomic status
 Low481471.0d1.0di
 Medium401240.9(0.5–1.5)0.9(0.5–1.6)
 High8480.5(0.2–1.0)0.5(0.2–1.1)
 Unknown2110.4(0.1–2.2)0.3(0.1–1.8)
Table 2. Distribution of Lung Cancer Patients by Survival Status and Estimation of the Effect of Method of Discovery, Presence of Symptoms, and Health Care Sector on Long Term Survival
 Survival status after 10 yearsCrude effectAdjusted effect
Alive (cases) n = 98Dead (controls) n = 330ORa(95% CI)ORb(95% CI)
  • OR: odds ratio; 95% CI: 95% confidence interval.

  • a

    Odds ratio adjusted for age (continuous) and gender.

  • b

    Odds ratio adjusted for age (continuous), gender, period (1977–1981/1982–1987), socioeconomic status (low/other), method of discovery (screening/other), extent of tumor (localized/other), and histologic groups (squamous cell/adenocarcinoma/carcinoid/other).

  • c

    Reference category.

  • d

    Odds ratio not adjusted for method of discovery.

Method of discovery
 Other872831.0c1.0cd
 Screening11470.7(0.3–1.4)0.6(0.3–1.3)
Presence of symptoms
 Yes381321.0c1.0cd
 No/unknown601981.0(0.6–1.7)1.1(0.7–1.7)
Sector of health care
 Public12461.0c1.0c
 Mixed852750.8(0.4–1.7)0.7(0.4–1.5)
 Private19
Table 3. Distribution of Lung Cancer Patients by Survival Status and Estimation of the Effect of Tumor Characteristics on Long Term Survival
 Survival status after 10 yearsCrude effectAdjusted effect
Alive (cases) n = 98Dead (controls) n = 330ORa(95% CI)ORb(95% CI)
  • OR: odds ratio; 95% CI: 95% confidence interval.

  • a

    Odds ratio adjusted for age (continuous) and gender.

  • b

    Odds ratio adjusted for age (continuous), gender, period (1977–1981/1982–1987), socioeconomic status (low/other), method of discovery (screening/other), extent of tumor (localized/other), and histologic groups (squamous cell/adenocarcinoma/carcinoid/other).

  • c

    Reference category.

  • d

    Odds ratio not adjusted for histology.

  • e

    P < 0.01.

  • f

    P < 0.05.

  • g

    Odd ratio not adjusted for the extent of tumor.

  • h

    P < 0.001.

Histologic group
 Squamous cell501951.0c1.0cd
 Adenocarcinoma28851.1(0.6–1.8)1.0(0.6–1.8)
 Carcinoid126 ]4.8e(1.6–14.1)4.0f(1.3–12.4)
 Large cell56
 Small cell1126 ]0.6(0.3–1.4)0.6(0.3–1.5)
 Other226 ]
Tumor differentiation
 Well18691.0c1.0c
 Moderately13580.9(0.4–2.0)0.9(0.4–1.9)
 Poorly14730.7(0.3–1.5)0.7(0.3–1.6)
 Undifferentiated7171.3(0.5–3.7)1.6(0.6–4.4)
 Unknown30681.5(0.7–2.9)1.7(0.9–3.3)
Extent of tumor
 Localized681401.0c1.0cg
 Direct regional29173 ]0.3h(0.2–0.5)0.4h(0.2–0.6)
 Distant or Metastasis117 ]
Lymph nodes
 Positive221371.0c1.0cg
 Negative641682.6h(1.5–4.4)2.4e(1.4–4.0)
 Unknown12252.6f(1.1–6.1)3.0e(1.3–6.8)
Size of tumor (cm)
 ≤233571.0c1.0cg
 ≤327730.6(0.3–1.2)0.6(0.3–1.2)
 ≤415680.4f(0.2–0.8)0.4e(0.2–0.8)
 >412900.2h(0.1–0.5)0.2h(0.1–0.5)
 Unknown11420.4f(0.2–0.8)0.5f(0.2–1.0)
TNM classification
 T147761.0c1.0cg
 T2462060.4h(0.2–0.7)0.5e(0.3–0.8)
 T3442 ]0.2h(0.1–0.4)0.2e(0.1–0.7)
 Unknown16 ]

The distribution of lung cancer patients according to the stage of the primary tumor at diagnosis is presented in Table 3. Approximately half of the patients (49%) presented with a localized stage of tumor, whereas 4% presented with a tumor extending distantly. The mean size of the operated tumors was 3.8 cm and adenopathy was present among 37% of patients. Compared with other patients, long term survivors presented with higher proportions of localized tumors (69% vs. 42%), of tumors measuring <2 cm (34% vs. 23%), and of T1 tumors (48% vs. 23%), and a lower proportion of positive lymph nodes (22% vs. 42%).

The effect of each factor is reported as relative risk estimates of long term survival. For estimation of the crude effect, relative risks were adjusted only for age and gender (when pertinent). With respect to sociodemographic factors, only age and gender were found to be significant prognostic factors (Table 1). Older age at diagnosis was associated with a decreasing chance of long term survival (chi-square for trend = 11.4; P < 0.001). Gender emerged as a discriminating factor in lung cancer outcome, with a 2.1-fold increased chance (95% confidence interval [95% CI], 1.1–3.7) of being a long term survival for women compared with men, after accounting for potential confounding factors (Table 1). The crude effects of period of diagnosis, civil status, country of birth, and socioeconomic status were under the threshold of statistical significance and in general were minimized when adjusting for other variables (Table 1). Method of discovery, presence of symptoms, and health care sector were not significantly related to long term survival status (Table 2).

As expected, all staging variables were significantly linked to the long term prognosis (Table 3). In particular, the probability of long term survival when adenopathy or regional or distant extension were absent was approximately threefold higher. The chance of long term survival decreased as the size of the tumor increased (chi-square for trend = 13.7; P < 0.001). When considering simultaneously several staging-related variables, the extent and the size of the tumor were found to be the most reliable prognostic staging factors, whereas adenopathy effects disappeared after accounting for the extent of the tumor as defined. The extension-adjusted odds ratio associated with the presence of involved lymph nodes therefore was equal to 1.0 (95% CI, 0.4–2.4) compared with the absence of involved lymph nodes. The effect of T classification, even if slightly lowered, persisted when accounting for the extent of the tumor.

Histology also was associated strongly with the long term survival status. Compared with individuals with squamous cell carcinoma, patients with malignant carcinoid had a 4.8-fold (95% CI, 1.6–14.0) increased chance of being survivors after 10 years. The differentiation of the tumor was not associated with long term survival status.

Results were consistent regardless of adjustment considered. In particular, additional adjustments for adjuvant chemotherapy (yes/no) or radiotherapy (yes/no) did not modify the results presented in the current study. The analyses performed on the same series (including 3 more patients lost to follow-up at 10 years and not at 5 years) but considering survivors at 5 years (instead of 10 years) as cases (n = 159) and patients with a shorter survival as controls (n = 272) provided similar results. For example, in a multivariate approach, the odds ratio associated with female gender (39 cases and 36 controls) was 2.2 (95% CI, 1.2–3.7) compared with males (120 cases and 236 controls); the odds ratio associated with regional or distant extension (55 cases and 166 controls) was 0.3 (95% CI, 0.1–0.5) compared with localized tumors (104 cases and 106 controls); the odds ratio associated with carcinoid histology (16 cases and 3 controls) was 4.6 (95% CI, 1.2–17.5) compared with squamous cell histology (91 cases and 156 controls).

DISCUSSION

This study identifies factors modifying the chance of lung cancer curability. We arbitrarily used the 10-year survival status as criterion of curability because only a few patients died of their lung cancer afterward. As presented in Figure 1, lung cancer survival (specific and crude) decreased constantly before reaching a plateau after 8 years, and remained constant after this period of time. We decided to study determinants acting on long term survival status rather than those acting on risk of death along the period. To this end the case–control approach was appropriate because all patients (except three) were followed during the entire study period.

Figure 1.

Survival after a priori curative surgery for cancer of the lung according to the Geneva Cancer Registry.

The main limitation of the current study is its low power. Even for patients who undergo putative curative resection (i.e., total macroscopic or microscopic removal), lung cancer rates of cure remain low. Only 23% of patients were survivors 10 years after diagnosis, providing fewer than 100 cases in the current series. Therefore we cannot adequately study factors of low frequency or with little effect. This study is based on lung cancer curability of patients diagnosed from the end of the 1970s until the end of the 1980s; therefore, it does not reflect the potential higher efficiency of more recent therapeutic approaches. However, chances of curability over time, if any, still are believed to be low. In particular, this study shows that long term survival expectancy is found to be fairly similar for patients diagnosed between 1982–1987 compared with patients diagnosed between 1977–1981 (Table 1).

Multivariate analysis allowed for the assessment of the effect on curability of each prognostic factor after accounting for others. We observed that age, stage, and histology independently determined lung cancer long term survival. In particular, lung cancer cure is 2.5-fold lower among patients age ≥ 60 years compared with patients age < 50 years. Patients with carcinoid carcinoma have a better prognosis with an approximately fourfold increased chance to be long term survivors in comparison with patients with epidermoid carcinoma. On the contrary, the curability of adenocarcinoma and epidermoid carcinoma is similar. The staging variables show an independent effect on prognosis, except for the protective effect linked to the absence of local adenopathy, which no longer is significant after accounting for the extent of the primary tumor. However, the assessment of multiadjusted effects is hampered by the small number of cases and should be reevaluated in larger series.

In addition to these well known prognostic factors, this study strongly suggests that gender is an important discriminating factor in lung cancer outcome, with women presenting a twofold likelihood of long term survival. This difference of gender in lung cancer survival, although often observed in large statistical series3, 19 has been poorly investigated. To our knowledge only a few studies from hospital series previously analyzed the gender-associated effects on survival from lung cancer,5, 6, 7, 20 and reasons for better prognosis in women remain obscure. In the current cancer registry-based study, possible differences between males and females regarding sociodemographic profile, tumor characteristics, and places of treatment as well as adjuvant therapy do not explain survival differences related to gender.

In Switzerland, as well as in other well developed countries, the general mortality rate is higher among males than among females. This gender differential in life expectancy does not explain the current results. As observed in Figure 1, gender differences are observed for both crude and specific survival. Additional analyses performed after excluding 64 patients (59 males and 5 females) dead of a cause other than lung cancer provide very similar results. Compared with males, women maintain a 2.1-fold increased chance (95% CI, 1.0–3.9) of being long term survivors after accounting for potential confounding factors.

Weight loss after diagnosis recently has been found to play a role in mediating gender-related differences in nonsmall lung cancer survival.7 Coexisting medical disorders and other clinical parameters such as general condition and cardiovascular and pulmonary functions are not available from the registry files. However, these factors are believed to determine postoperative mortality and short term survival rather than long term prognosis. The study also is limited by the absence of records regarding potential prognostic factors related to life-style habits such as smoking. Even if smoking status was found not to be linked to lung cancer survival in one previous study,7 the possibility exists that smoking may shorten survival. However, we do not believe that gender differences in smoking exposure explains the entire gender differential observed in the current study.

The fact that lung cancer, as well as other tumors (such as melanoma and colon cancer), has a better prognosis among females can be correlated with gender-dependent modifiers of carcinogenesis. Although the lung usually is not considered a major “target organ” of sex hormones, epidemiologic observations and studies of pulmonary neoplasms in laboratory animals suggest that sex hormones may play a role in both the pathogenesis and prognosis of bronchogenic carcinoma.21–26 It also has been observed that the formation of new blood vessels (angiogenesis), which is critical for the growth of tumors, can be inhibited by an endogenous estrogen metabolite.27 If sex hormone-related factors both modify lung cancer prognosis and differ between genders (by its presence or by its action), it may explain gender differences in lung cancer outcome.

In addition to age, extent of tumor, histologic type, and gender, no additional prediction variables were identified in the multivariate analysis. In particular, the curability of tumors discovered by screening appears similar to that of tumors discovered with another method. This indirectly confirms that although lung cancer screening can increase the length of survival by advancing the date of diagnosis it does not modify the chance of curability.28, 29

This population-based study further demonstrates that gender strongly influences the long term survival expectancy of patients with lung cancer, independently of other well established prognostic factors. Reasons for the better prognosis in women remain speculative. Further studies examining possible environmental and biologic prognostic factors are needed to explain this gender difference.

Acknowledgements

The authors thank Mr. Luc Raymond for his helpful comments regarding the article and Mrs. Stina Blagojevic for her expert technical help and editorial assistance.

Ancillary