|
Advertisement |
|
|
||
 |
|
|||||
|
|||||
|
 |
||||||
|
||||||
|
||||||
|
© 2003 American Society for Clinical Oncology Conditional Survival of Patients With the Four Major Histologic Subgroups of Lung Cancer in DenmarkFrom the Institute of Cancer Epidemiology, Danish Cancer Society, Copenhagen, Denmark. Address reprint requests to Halla Skuladottir, MD, Danish Cancer Society, Strandboulevarden 49, 2100 Copenhagen, Denmark; email: halla{at}cancer.dk.
Purpose: The survival probability of patients with lung cancer is usually based on the extent of disease as assessed at the time of diagnosis. The discouraging 5-year survival is often reported (< 10%) without taking into account changes in the survival probability as time advances from diagnosis. Patients and Methods: Conditional survival estimates by sex, age, extent of disease, and histology were estimated for patients diagnosed with lung cancer in Denmark from 1943 to 1997. Survival probabilities were calculated by the Kaplan-Meier method, and cumulative survival estimates were used to derive conditional survival estimates. Results: For every additional year survived, the probability of surviving the next 5 years increases from 33% (men) and 36% (women) after the first year, to 60% (men) and 67% (women) who have survived 5 years. The 5-year survival probability of patients younger than 49 years who had survived the first year was 33%, and increased to 81% after the fifth year. Corresponding estimates for 60- to 69-year-old patients were 23% and 52%. The conditional survival differed greatly among patients with localized and regional disease (29% and 10%, respectively) in the first year, but converged with time (52% and 47%, respectively) after 5 years. The conditional survival is similar in patients with squamous cell carcinoma, adenocarcinoma, and large-cell carcinoma, but is markedly lower in patients with small-cell carcinoma. Conclusion: For patients who have survived more than 1 year, the conditional survival probability provides a more accurate estimate of survival as compared with the conventional observed survival rates.
LUNG CANCER is the most common malignancy in Danish men and is still increasing in incidence in Danish women.1 The survival probability of patients with lung cancer is usually based on the extent of disease as assessed at the time of diagnosis, and often the 5-year survival is reported. Lung cancer has an unfavorable prognosis, with an overall 5-year survival of less than 10%.2 It is not as widely appreciated, however, that if patients survive for a specified interval of time (for example, 1 year or longer), their probability of survival changes. The survival probability that is calculated after a given length of survival includes only individuals who have survived to the predefined time of interest and is called conditional survival. Conditional survival, thus, is a survival probability of a cohort under the condition that the group has already survived a given amount of time. The Danish Cancer Registry and the Central Population Registry provide population-based, long-term, unbiased, follow-up data on all lung cancer patients in Denmark. These data provide the opportunity for us to calculate the conditional survival of Danish lung cancer patients, which reflects their average survival regardless of treatment and other prognostic factors. The survival of patients with lung cancer in Denmark has been reported previously, most recently in connection with the EUROCARE II Study, which reported on the survival of patients with the most common malignant diseases in Europe for the period from 1978 to 1989.2 A Nordic publication also reported on the survival of patients with the most common malignant diseases, but concentrated on the Scandinavian countries for the period from 1978 to 1991.3 This article reports the conditional survival of Danish lung cancer patients diagnosed from 1943 to 1997 by sex, age, extent of disease at diagnosis, and histology, and emphasizes the changes in survival probabilities during time after diagnosis by the use of conditional survival estimates. We also report the risk of dying from the disease. The most important prognostic factors are taken into account (sex, age, extent of disease at diagnosis, and histology) by use of a Cox proportional hazards regression analysis.
All patients with cancer in the lung reported on during the period from 1943 to 1997 were identified in the Cancer Registry database. For each person, the following information is available: date of diagnosis, date of birth, sex, name, site of tumor, extent of disease, and, since 1968, a personal identification number. For registration of cancer patients in the period from 1943 to 1977, the Danish Cancer Registry used the topography codes of the International Classification of Diseases (7th revision; ICD-7), published by the World Health Organization.4 We extracted information from the registry on all patients who had the ICD-7 codes 162.0 to 162.9, exclusive of 162.2, which is the code for pleural neoplasm. Reliable information on the histology of the tumors is not available for this initial period. In 1978, the Danish Cancer Registry changed to the International Classification of Diseases for Oncology (ICD-O) codes for registration and these are still in use.5 We extracted information on all patients diagnosed from 1978 to 1997, with the same topographic ICD-7 codes as described earlier. The ICD-O classification allows a more detailed classification of the malignancies because of its histology code. After identification of patients with lung cancer diagnosed after 1978, patients were categorized into the following histologic groups: adenocarcinoma, squamous cell carcinoma, small-cell carcinoma, large-cell carcinoma, and all other histologic groups, according to the World Health Organization classification of lung tumors.6 The latter group was not included in the analysis because of the heterogeneity of the group. A system of multiple reports from different sources has secured a high degree of completeness of the registry; hospital and pathology departments, forensic medicine departments, and private medical clinics supply the registry with patient information. Physicians voluntarily provided notification to the registry until 1987, when notification of malignant diseases was legislated. Since 1987, the information in the Danish Cancer Registry is supplemented by an annual search for cancer patients in the files of the National Hospital Discharge Registry. When additional patients have been identified, the treating hospitals are requested to send a standard notification to the Danish Cancer Registry for verification. All information about unreported cancer patients listed on death certificates, after verification by the treating hospital, is added to the Danish Cancer Registry on an annual basis. The personal identification number also allows linkage of vital status between the Central Population Registry and the Danish Cancer Registry for subsequent follow-up of cancer patients. All patients were observed up to December 31, 1999, through a link with the Central Population Registry.
Statistical Analysis All statistical analyses were performed using STATA 7 software (STATA Corp, College Station, TX). Cox regression analysis was used to estimate hazard rate ratios, a measure of the relative risk of dying, using the Breslow method for ties, and a time scale of years from diagnosis (on the basis of recorded calendar years). Two-sided 95% confidence intervals (CIs) for the rate ratios were calculated on the basis of Wald’s test of the Cox regression parameter.
Observed survival probabilities were calculated by the Kaplan-Meier method. Cumulative survival estimates were used to derive conditional survival estimates using the multiplicative law of probability. Knowledge of the probability of event A and event B occurring (Pr A and B) and the probability of event A occurring (Pr A) allows estimation of the conditional probability of event B to occur given that event A has occurred: Pr(B A) = Pr(A and B)/Pr(A). To estimate the 5-year conditional survival of patients who have survived 2 years, the 7-year cumulative survival (event A and B) is divided by the 2-year cumulative survival (event A). Because age is an important predictor of survival of lung cancer, all estimates of the survivor function were adjusted for age (a continuous variable), and when appropriate, estimates were also adjusted for sex and extent of disease at diagnosis by estimating separate Cox regression models for each category of sex and extent of disease, as shown in Table 1
CIs were estimated around the conditional 5-year survival rates. The variance of the conditional survival rate was estimated by use of the following formula, which is a variation of Greenwood’s formula used for estimation of CIs for unconditional survival:
Given m time intervals, dk is number of deaths during interval k; rk is number at risk for interval k; S(t) is the probability of survival past time t; and S(tj ti) is the probability of survival past time tj, given survival past time ti. The 95% CIs are constructed assuming that the conditional survival rates follow a normal distribution.
A total of 90,511 patients diagnosed with primary lung cancer from 1943 to 1997 were included in the survival analyses (65,849 men and 24,617 women; Table 1  ). The patients were categorized into four age groups ( 49, 50 to 59, 60 to 69, and 70 to 89 years) and four categories of extent of disease (localized, regional, metastatic disease, and unknown extent of disease); patients diagnosed in the period from 1978 to 1997 were categorized into histologic subtypes (squamous cell cancer, small-cell cancer, adenocarcinoma, and large-cell carcinoma).
Table 2
Figure 1  shows the 5-year survival probabilities of lung cancer patients by sex, under the condition that the group has survived up to 5 years from diagnosis of lung cancer, adjusted for age and extent of disease. For every additional year survived, the probability of surviving the next 5 years increases from 33% (men) and 36% (women) in the first year to 60% (men) and 67% (women). Thus, 5-year conditional survival from lung cancer is slightly more favorable in women than in men.
The 5-year conditional survival of lung cancer patients in four age groups is shown in Figure 2 , adjusted for sex. The conditional survival is steeper, and thus more favorable in the younger age groups when compared with the older age groups; however, in all age groups the survival probability increases with every succeeding year survived. The 5-year survival probability of patients in the youngest age group ( 49 years old) under the condition that the group has survived the first year is 33%; if they survive 1 additional year, their 5-year survival probability increases to 58%; if they survive yet another year, their 5-year survival probability increases to 71%. Similarly, patients in the older age groups (eg, patients who were 60 to 69 years old at diagnosis) have a 5-year survival probability of 23% under the condition that the group has survived the first year, and that probability increases to 39% and 47% if they survive the second and third year, respectively. Four years after the diagnosis, the conditional survival seems to be level in patients  60 years old.
Extent of disease at diagnosis is a strong determinant of prognosis. We have estimated the 5-year conditional survival of patients with localized disease and with regional spread of lung cancer, adjusted for age and sex (Fig 3 ). The survival probability increases with every additional year survived from 29% in patients with localized disease who have survived the first year to 52% for patients who have survived 4 and 5 years from diagnosis. In patients with regional disease, the increase in conditional survival is even steeper than for patients with localized disease: 5-year conditional survival after having survived the first year is 10% and increases to 47% in patients having survived 5 years from diagnosis. The 5-year conditional survival also increased with every succeeding year from diagnosis in patients with metastatic disease, but because of the relatively few numbers of patients in that category, their estimates are not shown on the graph.
The 5-year conditional survival of lung cancer by the four major histologic subtypes of lung cancer is shown in Figure 4 , adjusted for age, sex, and extent of disease. The survival increases with years already survived in all histologic groups. The conditional survival is similar in patients with squamous cell carcinoma, adenocarcinoma, and large-cell carcinoma, but is consistently lower in patients with small-cell carcinoma. Patients with squamous cell carcinoma who have survived the first year have a 5-year survival probability of 36%, which increases to 50%, 57%, and 62% for those who survive 2, 3, and 4 years, respectively, after diagnosis. After the fourth year, the conditional survival probability does not increase much more in patients with squamous cell carcinoma, adenocarcinoma, and large-cell carcinoma, and decreases in patients with squamous cell carcinoma after the fourth year. Patients with small-cell carcinoma have a less favorable prognosis; 5-year survival after patients have survived the first year is 12%, and increases to 32%, 43%, and 50% after patients have survived 2, 3, and 4 years, respectively. The rate increases to 57% after patients have survived 5 years.
To date, the effects of the major determinants of survival in lung cancer patients have been shown separately. It also is possible to use knowledge of all three major determinants of prognosis at once (age, extent of disease at diagnosis, and histology) to predict survival probabilities for different groups of patients. Figure 5 shows the 5-year conditional survival for lung cancer patients with localized disease by histology and age (adjusted for sex), with age regrouped to include two groups to increase the strength of the estimate ( 59 and 60 to 89 years old). Generally, the younger age group has a better prognosis than the older age group; the 5-year conditional survival after patients have survived 5 years regardless of histology is 69% to 73% in the younger age group, but is only 33% to 50% in the older age group.
In our study of 90,511 patients diagnosed with primary lung cancer from 1943 to 1997, we found a strong dependence of survival probability on the time that had elapsed since diagnosis, after we adjusted for sex, age, and extent of disease. Lung cancer has a poor prognosis, with less than 10% overall 5-year survival, but for patients who have passed the initial critical period, the probability of long-term survival increases with every additional year survived. Age has a marked influence on the conditional survival of patients with lung cancer; conditional survival is more favorable in the younger age groups as compared with the older age groups, independent of extent of disease. The 5-year conditional survival increases throughout the first 5 years, indicating that the underlying hazard still is leveling off in this group of patients, whereas in the older age groups, the conditional survival remains unchanged after the fourth year survived, indicating that the underlying hazard remains the same after the fourth year. To the authors’ knowledge, only one study has previously reported the conditional survival of lung cancer patients in general.7 They used the unadjusted observed cumulative survival probabilities to derive the conditional survival values and found that the survival probabilities of the patients increased with years survived, in agreement with our findings. We have adjusted our analysis for the effects of sex, age at diagnosis, and extent of disease to ensure that these variables could not by themselves explain our findings. Another study reported the conditional median survival of patients with different types of advanced cancers; among them were lung cancer patients with metastatic disease.8 The median survival of the patients increased with every succeeding year survived, which is consistent with our findings. The strongest single predictor of survival in lung cancer patients is extent of disease at diagnosis, but other factors such as age and histology are still important predictors of survival within a given level of extent of disease. Five-year conditional survival after patients have survived 5 years can vary from 33% in patients with adenocarcinoma in the age group 60 to 89 years to 50% in patients with squamous cell carcinoma. Up to 63% of all patients who have undergone radical surgery for non–small-cell lung cancer and are considered to have a disease confined to the lung (stage I) by a conventional assessment of tumor slides stained with hematoxylin and eosin have occult micrometastasis detected by immunohistochemical methods in the regional lymph nodes. Studies including patients with mixed nodal status have also reported a high rate of occult metastases to the bone marrow (reviewed in9). Although the possibility of false-positive findings is always present in such studies, the findings indicate an early metastatic spread, even in stage I tumors, which explains the unfavorable prognosis of these patients despite seemingly successful radical eradication of the tumors. As time evolves from diagnosis, patients considered to have localized disease, but with occult metastases, will eventually succumb to the disease. These are probably the patients that might benefit from adjuvant therapy. Long-term survivors diagnosed with regional disease might, conversely, be overdiagnosed with mediastinal lymphadenopathy that is assumed to be malignant but does not contain metastases. A meta-analysis has demonstrated that up to 29% of computed tomography scans showing nodal enlargement were false-positive.10 Thus, the long-term survivors will probably be those who truly had a localized disease. Histology also has influence on the survival of patients with lung cancer. The conditional survival curves for patients with squamous cell carcinoma, adenocarcinoma, and large-cell carcinoma have a similar course, indicating a similar underlying hazard for patients with these three histologic groups, whereas the conditional survival for small-cell lung cancer seems to have a different and less favorable course. The survival probability of patients with small-cell lung cancer increases with every succeeding year survived, but to a lesser degree than in patients diagnosed with the three other histologies. For patients who have survived more than 1 year, the conditional survival probability provides a more accurate estimate of survival as compared with the conventional observed survival rates. This important information should be communicated to patients when they wish to be informed of their prognosis as time passes from diagnosis.
1. Skuladottir H, Olsen JH, Hirsch FR: Incidence of lung cancer in Denmark: Historical and actual status. Lung Cancer 27:107–118, 2000[Medline] 2. Janssen-Heijnen MLG, Gatta G, Forman D, et al: Variation in survival of patients with lung cancer in Europe, 1985–1989. Eur J Cancer 34:2191–2196, 1998[CrossRef][Medline] 3. Engeland A, Haldorsen T, Tretli S, et al: Prediction of cancer mortality in the Nordic countries up to the years 2000 and 2010, on the basis of relative survival analysis: A collaborative study of the five Nordic Cancer Registries. APMIS 103 S1–S163, 1995 (suppl 49) 4. World Health Organization: International Statistical Classification of Diseases and Related Health Problems (ed 7). Geneva, Switzerland, World Health Organization, 1978 5. World Health Organization: International Classification of Diseases for Oncology (ed 1). Geneva, Switzerland, World Health Organization, 1976 6. Travis WD, Colby TV, Corrin B, et al: Histological Typing of Lung and Pleural Tumours (ed 3). Geneva, Switzerland, World Health Organization, 1999 7. Merrill RM, Henson DE, Barnes M: Conditional survival among patients with carcinoma of the lung. Chest 166:697–703, 1999 8. Kato I, Severson RK, Schwartz AG: Conditional median survival of patients with advanced carcinoma: Surveillance, epidemiology, and end results data. Cancer 92:2211–2219, 2001[Medline]
9. Jiao X, Krasna MJ: Clinical significance of micrometastasis in lung and esophageal cancer: A new paradigm in thoracic oncology. Ann Thorac Surg 74:278–284, 2002 10. Dales RE, Stark RM, Raman S: Computed tomography to stage lung cancer: Approaching a controversy using meta-analysis. Am Rev Respir Dis 141:1096–1101, 1990[Medline] Submitted April 2, 2003; accepted May 23, 2003.
This article has been cited by other articles:
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|||||||||||
|
|||||||||||
|
|
Copyright © 2003 by the American Society of Clinical Oncology, Online ISSN: 1527-7755. Print ISSN: 0732-183X
|
TOP
ABSTRACT
INTRODUCTION
