Improved Survival of Follicular Lymphoma Patients in the United States
http://www.100md.com
《临床肿瘤学》
the Division of Hematology, Oncology, and Blood and Marrow Transplantation, Department of Internal Medicine, College of Medicine, and Department of Epidemiology, College of Public Health, University of Iowa, Iowa City, IA
ABSTRACT
PURPOSE: Despite several new treatment options, single- and multi-institution analyses have not clarified whether survival patterns in follicular lymphoma (FL) patients have changed in recent decades. We undertook a study using a large population-based registry to analyze survival patterns among patients with FL.
PATIENTS AND METHODS: Surveillance, Epidemiology, and End Results morphology codes were used to identify 14,564 patients diagnosed with FL between 1978 and 1999. Observed median survival times, Kaplan-Meier survival curves, proportional death hazard ratios, and relative survival rates were calculated. Joinpoint regression analysis was used to identify trends in annual adjusted death hazard ratios.
RESULTS: An improvement in survival of all patients with FL was observed between each of three diagnosis eras (1978 to 1985, 1986 to 1992, and 1993 to 1999) by log-rank tests. Among patients with stage-specific data, the median survival time improved from 84 months (95% CI, 81 to 88 months) in the 1983 to 1989 era to 93 months (95% CI, 89 to 97 months) in the 1993 to 1999 era. Similar findings were identified across sex and age groups and for subsets including advanced-stage, large-cell FL and the combined subset of small cleaved- and mixed-cell FL. The inter-era survival advantage observed in white patients was not observed for black patients. The relative risk of death decreased by 1.8% per year over the 1983 to 1999 observation period.
CONCLUSION: The survival of patients with FL in the United States has improved over the last 25 years. The survival improvement may be a result of the sequential application of effective therapies and improved supportive care.
INTRODUCTION
Follicular lymphomas (FL) comprise 19% of all non-Hodgkin's lymphomas (NHL) in the United States and represent the majority of indolent NHL.1 Approximately half of patients with FL present with disseminated disease, and most exhibit a relatively indolent clinical course that is characterized by a cyclical pattern of induced remissions and relapses with occasional transformation to a more aggressive histology. Systemic therapies are not recognized as curative, and debate persists regarding the impact of treatment on survival.
Over the last 25 years, there has been a proliferation of management options in the treatment of FL, consisting of observation, alkylating agents, anthracyclines, purine nucleoside analogs, radiation therapy, combination chemotherapy, interferon, radiolabeled and unlabeled monoclonal antibodies, and autologous or allogeneic bone marrow or peripheral stem-cell transplantation. Although several of these treatment regimens induce significant tumor response, there is little evidence from controlled clinical trials that these therapies improve overall survival.2-4 Previous analyses examining long-term follow-up of indolent lymphoma patients, including follicular histologies, suggest that survival has remained stagnant over time.5-7
We examined the survival trends of patients with FL in a large population-based cancer surveillance program to assess for differences in survival by histologic subtype, age, race, and sex. We hypothesized that the sequential application of effective therapies during the course of the disease, coupled with improvements in supportive care, may have translated into a previously unrecognized survival benefit.
PATIENTS AND METHODS
We obtained survival data for individuals diagnosed with FL for the period of 1978 to 1999 from the 2002 Public-Use Database of the Surveillance, Epidemiology, and End Results (SEER) program of the National Cancer Institute.8 The analysis was restricted to nine population-based SEER cancer registries (SEER-9) comprising the states of Iowa, Connecticut, Hawaii, New Mexico, and Utah and the metropolitan areas of Detroit, Michigan; San Francisco-Oakland, California; Seattle-Puget Sound, Washington; and Atlanta, Georgia. The SEER-9 registries have a case ascertainment rate of 96% to 98%, and the nine registries represent approximately 10% of the US population.8,9 The morphologic codes in the SEER registry are based on the diagnoses of pathologists at site of care, using local standards for histology and immunohistochemistry. Because of the limited number of FL cases among minority racial groups, race information was coded as black, white, other, and unknown.
We excluded patients identified before 1978, as recommended by the SEER program for analyses of trends.10 Before 1978, the SEER program classified patients according to the Manual of Tumor Nomenclature and Coding system,11 which is less descriptive and provides less pathologic detail than the current system. In 1978, the SEER program began using the International Classification of Disease for Oncology (ICD-O)12 for tumor classification. The second edition was implemented in 1992 (ICD-O-2).13 To convert previously abstracted cases to new classification systems, the SEER program used a computerized algorithm based on published conversion tables. For our analysis, the pertinent conversion tables were reviewed.14
Analyses requiring stratification by stage of disease were restricted to patients diagnosed in 1983 and later because more detailed stage information was initially collected that year. Extent of disease criteria used by SEER registry abstractors were reviewed for FL, and the criteria are consistent with the Ann Arbor/American Joint Committee on Cancer staging system.15 The classification systems correlate as follows: limited, stage I; regional, stage II; and distant, stage III/IV.
Diagnostic Codes
Patients with FL were identified by ICD-O-2 morphology codes (9690-9698). Histologic FL subtypes were identified as follows: small cleaved-cell FL (FSC; 9693-9696), mixed FL (FM; small cleaved- and large-cell FL, 9691-9692), large-cell FL (FLC; 9697-9698), and FL not otherwise specified (9690).1
Study Cohort
We identified 14,637 patients in the SEER-9 database who were diagnosed with FL from 1978 through 1999. Patients were excluded if they did not have microscopic confirmation of disease (n = 46) and survival information (n = 27). For analyses requiring stratification by stage of disease, patients diagnosed before 1983 were excluded (n = 2,476). Of the remaining patients who met the study criteria and who were diagnosed when stage information was routinely collected, those patients for whom no stage information was provided were excluded (n = 890).
For survival analyses, we divided patients into eras by year of diagnosis. The diagnosis eras were designed to contain approximately equal numbers of patients. We defined the groups by the following diagnosis years: 1978 to 1985, 1986 to 1992, and 1993 to 1999. Because accurate staging data were not available until 1983, subsequent analyses were stratified by stage of disease, with patients divided into two groups defined by the diagnosis years of 1983 to 1989 and 1990 to 1999. Because patient follow-up was reported in the SEER database through December 2000 and because the diagnosis eras were constructed to provide adequate follow-up time to assess survival, the two diagnosis eras contained unequal numbers of patients.
For the stage-specific analyses, the FL histologic subtypes were combined into two histologic groups. FSC and FM were grouped together and collectively referred to as FSC + FM. FLC comprised the other histologic group.
Statistical Analysis
We obtained survival data from the SEER public-use data files.16 Observed median survival rates and adjusted death hazard ratios were calculated using SAS version 8.2 (SAS Institute, Cary, NC). Survival curves were constructed with the Kaplan-Meier method and compared with the log-rank test.17 Analyses requiring adjustments for potential confounding factors were conducted using the Cox proportional hazards method. The proportional hazards assumption was tested and satisfied for each mathematical model using Cox analysis. Death hazard ratios were adjusted for age at diagnosis, calendar year of diagnosis, sex, race, registry site, tumor histology, and stage of disease at diagnosis. A level of significance () of .05 was considered statistically significant.
The relative survival rate of patients with FL was calculated by taking into account the expected survival of a similar cohort of the general population without FL. The relative survival rate is the ratio of the observed survival divided by the expected survival of a cohort of the general population possessing similar characteristics with respect to age, race, sex, and era of diagnosis.18 In this analysis, the relative survival rate was calculated using SEER * Stat 5.0.20, which obtains expected survival rates for the general population from life tables obtained from the National Center for Health Statistics.19 The relative survival rates were calculated in 1-year increments for years 1 through 10 for patients in each diagnosis era.
A linear regression analysis of the adjusted death hazard ratios and year of diagnosis was conducted using SAS.20 The analysis was repeated using Joinpoint, which is a software program from the Statistical Research Applications Branch of the National Cancer Institute that fits trend data using joinpoint (or spline) models.21 The software calculates an estimated annual percent change with 95% CIs and identifies the best joinpoint model for the data.22 The analysis starts with the minimum number of joinpoints, and with each statistically significant joinpoint, a new estimated annual percent change is determined.
RESULTS
Characteristics of the Study Participants
We identified 14,564 patients who met the study criteria (Table 1). The median age at diagnosis was 63 years. The vast majority of patients in the study were white (92.6%). Black patients and other race patients each comprised 3.5% of the study cohort. Race was unknown for 0.4% of patients. The stage-adjusted survival analyses compared 12,088 patients distributed over two diagnosis eras (1983 to 1989 and 1990 to 1999). There was no significant stage migration over time because these eras had nearly identical proportions of limited- and regional-stage and advanced-stage disease patients.
Univariate Survival Analyses
Kaplan-Meier survival curves for all FL patients in the cohort of 14,564 patients demonstrated a temporal improvement in survival (Fig 1). Statistically significant improvements in survival were observed across consecutive diagnostic eras (1978 to 1985, 1986 to 1992, and 1993 to 1999) by log-rank tests. Median survival time for the 1978 to 1985 era was 82 months (95% CI, 78 to 86 months), and median survival time for the 1986 to 1992 era was 87 months (95% CI, 82 to 90 months). A median survival time was not reached for the 1993 to 1999 era.
For analyses requiring stratification by stage of disease, 12,088 patients were studied. Kaplan-Meier survival curves for two diagnosis eras (1983 to 1989 and 1990 to 1999) revealed a statistically significant 9-month increase (10.7%) in observed median survival time, which was confirmed by a log-rank test (P = .004; Table 2). Among patients with advanced-stage disease, we observed a significant survival improvement (observed median survival time, 63 months for 1983 to 1989 v 72 months for 1990 to 1999) between the eras by log-rank test (P < .001), whereas in patients with limited and regional disease, an observed median survival time of 114 months was observed in each era (P = .761). Survival improved across diagnosis eras for males, females, older patients ( 60 years), and younger patients (< 60 years), as determined by the log-rank test. Among FSC + FM patients, we identified an 8-month improvement in observed median survival time between diagnosis eras (P = .018). An improved observed median survival time from 62 to 77 months was identified among FLC patients (P = .015). Survival was similar for black and white patients in the early era and improved significantly across eras for white patients (P = .004) but not for black patients (P = .650).
Among patients with advanced-stage disease, a survival improvement across eras was seen in FSC + FM in both the younger and older subsets as well as in FLC in younger patients. (Fig 2). This difference was most pronounced among younger patients (< 60 years) with FLC. Among patients with limited- and regional-stage disease, young patients with FSC + FM were the only subset who demonstrated a significant improvement in survival across eras.
Multivariate Survival Probabilities
Adjusted death hazard ratios for selected subgroups were calculated comparing the risk of death for patients diagnosed in the 1990 to 1999 diagnosis era with the 1983 to 1989 era (Table 2). The adjusted death hazard ratio for all patients diagnosed in the 1990 to 1999 era was 0.878 (P < .001), indicating a 12% decreased risk of death for patients in the 1990 to 1999 diagnosis era. Significant decreases occurred for limited- and regional-stage and advanced-stage patients, both age and sex groups, and patients with FSC + FM and FLC. Although there was no difference for patients with limited- and regional-stage disease by log-rank test or by observed median survival, there was evidence of a decreased death hazard ratio between diagnosis eras. When the date from limited- and regional-stage patients was adjusted for age only, similar findings were demonstrated (death hazard ratio = 0.934, P = .013).
A linear regression analysis of adjusted death hazard ratios and year of diagnosis revealed a statistically significant negative slope (P < .001). The analysis was repeated using Joinpoint regression software (Fig 3). For all patients in the cohort, the estimated percent change in risk of death decreased by 1.80% per year over the observation period (95% CI, –2.34% to –1.25%), and no significant changes in trend (joinpoints) were identified within the 1983 to 1999 observation period. The adjusted death hazard ratio decreased by approximately 25% across the observation period.
Relative Survival Rates
The observed survival rates, expected survival rates (among the general population), and relative survival rates in 1-year intervals were calculated using SEER * Stat and are displayed graphically in Figure 4 for each diagnosis era. The expected 10-year survival rates for the age-, race-, and sex-matched general population were 0.739 (1983 to 1989) and 0.721 (1990 to 1999). Among all patients in the staged FL cohort, the observed 10-year survival rates were 0.386 (1983 to 1989) and 0.438 (1990-1999). Relative 5-year survival rates were 0.698 (1983 to 1989) and 0.741 (1990 to 1999), and 10-year relative survival rates were 0.522 (1983 to 1989) and 0.608 (1990 to 1999). Among patients with advanced disease, 5-year relative survival rates were 0.591 (1983 to 1989) and 0.653 (1990 to 1999), and 10-year relative survival rates were 0.384 (1983 to 1989) and 0.508 (1990 to 1999).
DISCUSSION
This report is the first to describe a temporal improvement in the survival of patients with FL. Analyzing a large cohort of patients in the SEER registry demonstrated that observed median survival of FL patients diagnosed in the United States improved by 11% from the 1980s to the 1990s, which is reflective of an adjusted death hazard ratio that decreased by nearly 25% from 1983 to 1999. Using multiple analytic methods, we noted a consistent trend of improved survival since 1978 for patients with FL. The relative survival rates obtained by SEER * Stat analysis demonstrated that the survival improvement observed among FL patients has occurred independent of changes in the general population. The survival advantage was not demonstrated in all patient subsets, most notably the black population. The most striking improvement was demonstrated in patients with advanced (stage III/IV) disease, which was the group of patients most likely to receive sequential systemic treatment.
Several case series of FL and indolent lymphoma have been analyzed by single- and multi-institution investigators. In these analyses, patients were often participants in clinical trials or were treated at large referral centers and, thus, were subject to inherent selection bias. In contrast to previous studies, we used a larger, population-based surveillance program and restricted our analysis to patients with FLs. The large sample size also allowed for subset analyses with acceptable statistical power.
The SEER database is a useful tool to answer selected clinical and epidemiologic questions, yet it has inherent advantages and limitations.23 Whereas single-institution and clinical trial patient series are capable of producing uniform and reliable hematopathology-based cohorts, the identification of FL cohorts in the SEER database is often based on a single interpretation from a variety of pathologists and institutions. However, this apparent weakness may also be a strength of the study in that the cohort identified in the SEER database is more representative of the real-world population of FL patients, and for this reason, the outcomes are important.
The diagnostic concordance of FL between expert hematopathologists is among the highest of the distinct lymphoid malignancies24; however, subclassification is more difficult and has been described previously.25 Two reviews of the SEER data have quantified NHL diagnostic reliability. An expert review of the Iowa SEER cancer registry between 1981 and 1984 concluded that FLs may be used confidently in epidemiologic studies, whether separated or grouped.26 A recent review of patients from 1988 to 1994 in the Greater Bay Area Cancer Registry (San Francisco-Oakland SEER registry) also concluded that agreement was acceptable for most FL subtypes.27
An important limitation of the SEER database is the absence of clinical information, such as tumor bulk, performance status, serum lactate dehydrogenase, and hemoglobin concentrations, and hence, the inability to control for prognostic scores such as the International Prognostic Index or International Prognostic Index for Follicular Lymphoma. In our multivariate analyses, we controlled for several available clinical factors, including stage of disease, histologic subtype, patient age, race, and sex. Although these variables contribute to the clinical description of the disease, the SEER database lacks other important characteristics routinely used in clinical practice.
The temporal survival improvements in our study were not observed uniformly across histologic subtypes of FL. Among patients with advanced-stage FSC + FM, which is a group that was likely to receive sequential systemic chemotherapy, a modest improvement in survival over time was observed in both younger and older populations. In contrast, for patients with limited- and regional-stage FSC + FM, the survival advantage was confined to the subset of younger patients who were more likely to receive systemic chemotherapy than the older patients in this setting. The observed survival improvement is most dramatic in younger patients with advanced-stage FLC, which is a group that was highly likely to receive anthracycline-containing chemotherapy.
We found that black patients with FL have not experienced the nearly 1-year survival advantage over the last two decades that white patients experienced. This finding is consistent with an analysis of survival outcomes of the Detroit SEER registry.28 Varterasian et al28 recognized that, among NHL patients in the metropolitan Detroit registry, the 5-year relative survival of black patients decreased over the observation period, whereas survival for whites in the Detroit registry increased. Further investigation is needed to identify potential causes of this disparity.
In November 1997, rituximab was approved by the US Food and Drug Administration for treatment of relapsed low-grade and follicular CD20+ NHL. There has been speculation that the use of this monoclonal antibody therapy may alter the clinical course of FL. Therefore, we conducted an analysis in which the years of diagnosis and patient follow-up were truncated in 1996, which revealed similar survival outcomes to our present study (data not shown). This modified analysis, combined with the joinpoint analysis described here, leads us to conclude that the survival advantage described in this report was observed before the widespread use of this novel therapy in the treatment of FL.
The identification of improved survival among patients with FL has important clinical and public health implications. Several clinical trials have demonstrated enhancement of event-free or progression-free survival in this patient population with various treatment regimens, but the trials have been too small and follow-up too short to demonstrate improvement in survival. Similarly, supportive care has improved throughout the observation period and may have contributed to population-based survival improvements in patients with lymphoma, despite the fact that individual supportive care trials in patients with lymphoma have not demonstrated differences in survival. The clinical course of FL has been altered, which is coincident with the changes in availability of management options. We speculate that the sequential application of these effective therapies, coupled with improved supportive care, is responsible for the improvement in survival in patients with FL. This finding justifies the study and development of sequential therapies, in addition to new agents and regimens, to further enhance survival in patients with FL.
Authors' Disclosures of Potential Conflicts of Interest
The authors indicated no potential conflicts of interest.
NOTES
Supported in part by National Cancer Institute grant Nos. T32 CA79445, HL 04117, and P50 CA097274.
Presented in part at the 40th Annual Meeting of the American Society of Clinical Oncology, New Orleans, LA, June 5-8, 2004 (abstract ID: 1090).
Authors' disclosures of potential conflicts of interest are found at the end of this article.
REFERENCES
Groves FD, Linet MS, Travis LB, et al: Cancer surveillance series: Non-Hodgkin's lymphoma incidence by histologic subtype in the United States from 1978 through 1995. J Natl Cancer Inst 92:1240-1251, 2000
Ardeshna KM, Smith P, Norton A, et al: Long-term effect of a watch and wait policy versus immediate systemic treatment for asymptomatic advanced-stage non-Hodgkin's lymphoma: A randomized controlled trial. Lancet 362:516-522, 2003
Brandt L, Kimby E, Nygren P, et al: A systematic overview of chemotherapy effects in indolent non-Hodgkin's lymphoma. Acta Oncol 40:213-223, 2001
Peterson BA: Current treatment of follicular low-grade lymphomas. Semin Oncol 26:2-11, 1999 (suppl 14)
Horning SJ: Follicular lymphoma: Have we made any progress? Ann Oncol 11:S23-S27, 2000 (suppl 1)
Horning SJ: Natural history of and therapy for the indolent non-Hodgkin's lymphomas. Semin Oncol 20:75-88, 1993 (suppl 5)
Johnson PW, Rohatiner AZ, Whelan JS, et al: Patterns of survival in patients with recurrent follicular lymphoma: A 20-year study from a single center. J Clin Oncol 13:140-147, 1995
National Cancer Institute: Surveillance, Epidemiology, and End Results (SEER) Program public-use data (1973-1999) program Web site: Data quality. http://seer.cancer.gov/about/quality.html
Zippin C, Lum D, Hankey BF: Completeness of hospital cancer case reporting from the SEER Program of the National Cancer Institute. Cancer 76:2343-2350, 1995
National Cancer Institute: Surveillance, Epidemiology, and End Results (SEER) Program Web site: Recoded SEER public access record description, cases diagnosed in 1973-2000, user file. http://seer.cancer.gov/manuals/CD2.SEERDic.pdf
American Cancer Society: Manual of Tumor Nomenclature and Coding. New York, NY, American Cancer Society, 1968
WHO: International Classification of Diseases for Oncology (ed 1). Geneva, Switzerland, WHO, 1976
Percy C, Van Holten V, Muir C (eds): International Classification of Diseases for Oncology (ed 2). Geneva, Switzerland, WHO, 1990
Van Holten V, Percy C: Conversion of neoplasms by topography and morphology from the International Classification of Diseases for Oncology (ICD-O) (First Edition and Field Trials) to International Classification of Diseases for Oncology, Second Edition (ICD-O-2). Bethesda, MD, Surveillance Program, Division of Cancer Prevention and Control, National Cancer Institute, NIH publication 98-3331, 1991
Greene FL, Page DL, Fleming ID, et al: AJCC Cancer Staging Manual (ed 6). New York, NY, Springer-Verlag, 2002
National Cancer Institute: Surveillance, Epidemiology, and End Results (SEER) 1973-2000 public-use data (1973-2000). http://seer.cancer.gov/publicdata
Allison PD: Survival Analysis Using SAS: A Practical Guide. Cary, NC, SAS Institute Inc, 1995
Henson DE, Ries LA: The relative survival rate. Cancer 76:1678-1688, 1995
National Cancer Institute: SEER * Stat Version 5.0.20. http://seer.cancer.gov/seerstat/software
Allison PD: Logistic Regression Using SAS System: Theory and Application. Cary, NC, SAS Institute Inc, 1999
National Cancer Institute: Joinpoint Regression Program, Statistical Research Applications Branch, Division of Cancer Control and Population Sciences. http://srab.cancer.gov/joinpoint
Kim H, Fay MP, Feuer EJ, et al: Permutation tests for joinpoint regression with applications to cancer rates. Stat Med 19:335-351, 2000
Harlan LC, Hankey BF: The Surveillance, Epidemiology, and End Results program database as a resource for conducting descriptive epidemiologic and clinical studies. J Clin Oncol 21:2232-2233, 2003
The Non-Hodgkin's Lymphoma Classification Project: A clinical evaluation of the International Lymphoma Study Group classification of non-Hodgkin's lymphoma: The Non-Hodgkin's Lymphoma Classification Project. Blood 89:3909-3918, 1997
Metter GE, Nathwani BN, Burke JS, et al: Morphologic subclassification of follicular lymphoma: Variability of diagnoses among hematopathologists, a collaborative study between the Repository Center and Pathology Panel for Lymphoma Clinical Studies. J Clin Oncol 3:25-28, 1985
Dick F, VanLier S, Banks P, et al: Use of the working formulation for non-Hodgkin's lymphoma in epidemiologic studies: Agreement between reported diagnoses and a panel of experienced pathologists. J Natl Cancer Inst 78:1137-1144, 1987
Clarke CA, Glaser SL, Dorfman RF, et al: Expert review of non-Hodgkin's lymphomas in a population-based cancer registry: Reliability of diagnosis and subtype classifications. Cancer Epidemiol Biomarkers Prev 13:138-143, 2004
Varterasian ML, Graff JJ, Severson RK, et al: Non-Hodgkin's lymphoma: An analysis of the Metropolitan Detroit SEER database. Cancer Invest 18:303-308, 2000(Wade T. Swenson, James E.)
ABSTRACT
PURPOSE: Despite several new treatment options, single- and multi-institution analyses have not clarified whether survival patterns in follicular lymphoma (FL) patients have changed in recent decades. We undertook a study using a large population-based registry to analyze survival patterns among patients with FL.
PATIENTS AND METHODS: Surveillance, Epidemiology, and End Results morphology codes were used to identify 14,564 patients diagnosed with FL between 1978 and 1999. Observed median survival times, Kaplan-Meier survival curves, proportional death hazard ratios, and relative survival rates were calculated. Joinpoint regression analysis was used to identify trends in annual adjusted death hazard ratios.
RESULTS: An improvement in survival of all patients with FL was observed between each of three diagnosis eras (1978 to 1985, 1986 to 1992, and 1993 to 1999) by log-rank tests. Among patients with stage-specific data, the median survival time improved from 84 months (95% CI, 81 to 88 months) in the 1983 to 1989 era to 93 months (95% CI, 89 to 97 months) in the 1993 to 1999 era. Similar findings were identified across sex and age groups and for subsets including advanced-stage, large-cell FL and the combined subset of small cleaved- and mixed-cell FL. The inter-era survival advantage observed in white patients was not observed for black patients. The relative risk of death decreased by 1.8% per year over the 1983 to 1999 observation period.
CONCLUSION: The survival of patients with FL in the United States has improved over the last 25 years. The survival improvement may be a result of the sequential application of effective therapies and improved supportive care.
INTRODUCTION
Follicular lymphomas (FL) comprise 19% of all non-Hodgkin's lymphomas (NHL) in the United States and represent the majority of indolent NHL.1 Approximately half of patients with FL present with disseminated disease, and most exhibit a relatively indolent clinical course that is characterized by a cyclical pattern of induced remissions and relapses with occasional transformation to a more aggressive histology. Systemic therapies are not recognized as curative, and debate persists regarding the impact of treatment on survival.
Over the last 25 years, there has been a proliferation of management options in the treatment of FL, consisting of observation, alkylating agents, anthracyclines, purine nucleoside analogs, radiation therapy, combination chemotherapy, interferon, radiolabeled and unlabeled monoclonal antibodies, and autologous or allogeneic bone marrow or peripheral stem-cell transplantation. Although several of these treatment regimens induce significant tumor response, there is little evidence from controlled clinical trials that these therapies improve overall survival.2-4 Previous analyses examining long-term follow-up of indolent lymphoma patients, including follicular histologies, suggest that survival has remained stagnant over time.5-7
We examined the survival trends of patients with FL in a large population-based cancer surveillance program to assess for differences in survival by histologic subtype, age, race, and sex. We hypothesized that the sequential application of effective therapies during the course of the disease, coupled with improvements in supportive care, may have translated into a previously unrecognized survival benefit.
PATIENTS AND METHODS
We obtained survival data for individuals diagnosed with FL for the period of 1978 to 1999 from the 2002 Public-Use Database of the Surveillance, Epidemiology, and End Results (SEER) program of the National Cancer Institute.8 The analysis was restricted to nine population-based SEER cancer registries (SEER-9) comprising the states of Iowa, Connecticut, Hawaii, New Mexico, and Utah and the metropolitan areas of Detroit, Michigan; San Francisco-Oakland, California; Seattle-Puget Sound, Washington; and Atlanta, Georgia. The SEER-9 registries have a case ascertainment rate of 96% to 98%, and the nine registries represent approximately 10% of the US population.8,9 The morphologic codes in the SEER registry are based on the diagnoses of pathologists at site of care, using local standards for histology and immunohistochemistry. Because of the limited number of FL cases among minority racial groups, race information was coded as black, white, other, and unknown.
We excluded patients identified before 1978, as recommended by the SEER program for analyses of trends.10 Before 1978, the SEER program classified patients according to the Manual of Tumor Nomenclature and Coding system,11 which is less descriptive and provides less pathologic detail than the current system. In 1978, the SEER program began using the International Classification of Disease for Oncology (ICD-O)12 for tumor classification. The second edition was implemented in 1992 (ICD-O-2).13 To convert previously abstracted cases to new classification systems, the SEER program used a computerized algorithm based on published conversion tables. For our analysis, the pertinent conversion tables were reviewed.14
Analyses requiring stratification by stage of disease were restricted to patients diagnosed in 1983 and later because more detailed stage information was initially collected that year. Extent of disease criteria used by SEER registry abstractors were reviewed for FL, and the criteria are consistent with the Ann Arbor/American Joint Committee on Cancer staging system.15 The classification systems correlate as follows: limited, stage I; regional, stage II; and distant, stage III/IV.
Diagnostic Codes
Patients with FL were identified by ICD-O-2 morphology codes (9690-9698). Histologic FL subtypes were identified as follows: small cleaved-cell FL (FSC; 9693-9696), mixed FL (FM; small cleaved- and large-cell FL, 9691-9692), large-cell FL (FLC; 9697-9698), and FL not otherwise specified (9690).1
Study Cohort
We identified 14,637 patients in the SEER-9 database who were diagnosed with FL from 1978 through 1999. Patients were excluded if they did not have microscopic confirmation of disease (n = 46) and survival information (n = 27). For analyses requiring stratification by stage of disease, patients diagnosed before 1983 were excluded (n = 2,476). Of the remaining patients who met the study criteria and who were diagnosed when stage information was routinely collected, those patients for whom no stage information was provided were excluded (n = 890).
For survival analyses, we divided patients into eras by year of diagnosis. The diagnosis eras were designed to contain approximately equal numbers of patients. We defined the groups by the following diagnosis years: 1978 to 1985, 1986 to 1992, and 1993 to 1999. Because accurate staging data were not available until 1983, subsequent analyses were stratified by stage of disease, with patients divided into two groups defined by the diagnosis years of 1983 to 1989 and 1990 to 1999. Because patient follow-up was reported in the SEER database through December 2000 and because the diagnosis eras were constructed to provide adequate follow-up time to assess survival, the two diagnosis eras contained unequal numbers of patients.
For the stage-specific analyses, the FL histologic subtypes were combined into two histologic groups. FSC and FM were grouped together and collectively referred to as FSC + FM. FLC comprised the other histologic group.
Statistical Analysis
We obtained survival data from the SEER public-use data files.16 Observed median survival rates and adjusted death hazard ratios were calculated using SAS version 8.2 (SAS Institute, Cary, NC). Survival curves were constructed with the Kaplan-Meier method and compared with the log-rank test.17 Analyses requiring adjustments for potential confounding factors were conducted using the Cox proportional hazards method. The proportional hazards assumption was tested and satisfied for each mathematical model using Cox analysis. Death hazard ratios were adjusted for age at diagnosis, calendar year of diagnosis, sex, race, registry site, tumor histology, and stage of disease at diagnosis. A level of significance () of .05 was considered statistically significant.
The relative survival rate of patients with FL was calculated by taking into account the expected survival of a similar cohort of the general population without FL. The relative survival rate is the ratio of the observed survival divided by the expected survival of a cohort of the general population possessing similar characteristics with respect to age, race, sex, and era of diagnosis.18 In this analysis, the relative survival rate was calculated using SEER * Stat 5.0.20, which obtains expected survival rates for the general population from life tables obtained from the National Center for Health Statistics.19 The relative survival rates were calculated in 1-year increments for years 1 through 10 for patients in each diagnosis era.
A linear regression analysis of the adjusted death hazard ratios and year of diagnosis was conducted using SAS.20 The analysis was repeated using Joinpoint, which is a software program from the Statistical Research Applications Branch of the National Cancer Institute that fits trend data using joinpoint (or spline) models.21 The software calculates an estimated annual percent change with 95% CIs and identifies the best joinpoint model for the data.22 The analysis starts with the minimum number of joinpoints, and with each statistically significant joinpoint, a new estimated annual percent change is determined.
RESULTS
Characteristics of the Study Participants
We identified 14,564 patients who met the study criteria (Table 1). The median age at diagnosis was 63 years. The vast majority of patients in the study were white (92.6%). Black patients and other race patients each comprised 3.5% of the study cohort. Race was unknown for 0.4% of patients. The stage-adjusted survival analyses compared 12,088 patients distributed over two diagnosis eras (1983 to 1989 and 1990 to 1999). There was no significant stage migration over time because these eras had nearly identical proportions of limited- and regional-stage and advanced-stage disease patients.
Univariate Survival Analyses
Kaplan-Meier survival curves for all FL patients in the cohort of 14,564 patients demonstrated a temporal improvement in survival (Fig 1). Statistically significant improvements in survival were observed across consecutive diagnostic eras (1978 to 1985, 1986 to 1992, and 1993 to 1999) by log-rank tests. Median survival time for the 1978 to 1985 era was 82 months (95% CI, 78 to 86 months), and median survival time for the 1986 to 1992 era was 87 months (95% CI, 82 to 90 months). A median survival time was not reached for the 1993 to 1999 era.
For analyses requiring stratification by stage of disease, 12,088 patients were studied. Kaplan-Meier survival curves for two diagnosis eras (1983 to 1989 and 1990 to 1999) revealed a statistically significant 9-month increase (10.7%) in observed median survival time, which was confirmed by a log-rank test (P = .004; Table 2). Among patients with advanced-stage disease, we observed a significant survival improvement (observed median survival time, 63 months for 1983 to 1989 v 72 months for 1990 to 1999) between the eras by log-rank test (P < .001), whereas in patients with limited and regional disease, an observed median survival time of 114 months was observed in each era (P = .761). Survival improved across diagnosis eras for males, females, older patients ( 60 years), and younger patients (< 60 years), as determined by the log-rank test. Among FSC + FM patients, we identified an 8-month improvement in observed median survival time between diagnosis eras (P = .018). An improved observed median survival time from 62 to 77 months was identified among FLC patients (P = .015). Survival was similar for black and white patients in the early era and improved significantly across eras for white patients (P = .004) but not for black patients (P = .650).
Among patients with advanced-stage disease, a survival improvement across eras was seen in FSC + FM in both the younger and older subsets as well as in FLC in younger patients. (Fig 2). This difference was most pronounced among younger patients (< 60 years) with FLC. Among patients with limited- and regional-stage disease, young patients with FSC + FM were the only subset who demonstrated a significant improvement in survival across eras.
Multivariate Survival Probabilities
Adjusted death hazard ratios for selected subgroups were calculated comparing the risk of death for patients diagnosed in the 1990 to 1999 diagnosis era with the 1983 to 1989 era (Table 2). The adjusted death hazard ratio for all patients diagnosed in the 1990 to 1999 era was 0.878 (P < .001), indicating a 12% decreased risk of death for patients in the 1990 to 1999 diagnosis era. Significant decreases occurred for limited- and regional-stage and advanced-stage patients, both age and sex groups, and patients with FSC + FM and FLC. Although there was no difference for patients with limited- and regional-stage disease by log-rank test or by observed median survival, there was evidence of a decreased death hazard ratio between diagnosis eras. When the date from limited- and regional-stage patients was adjusted for age only, similar findings were demonstrated (death hazard ratio = 0.934, P = .013).
A linear regression analysis of adjusted death hazard ratios and year of diagnosis revealed a statistically significant negative slope (P < .001). The analysis was repeated using Joinpoint regression software (Fig 3). For all patients in the cohort, the estimated percent change in risk of death decreased by 1.80% per year over the observation period (95% CI, –2.34% to –1.25%), and no significant changes in trend (joinpoints) were identified within the 1983 to 1999 observation period. The adjusted death hazard ratio decreased by approximately 25% across the observation period.
Relative Survival Rates
The observed survival rates, expected survival rates (among the general population), and relative survival rates in 1-year intervals were calculated using SEER * Stat and are displayed graphically in Figure 4 for each diagnosis era. The expected 10-year survival rates for the age-, race-, and sex-matched general population were 0.739 (1983 to 1989) and 0.721 (1990 to 1999). Among all patients in the staged FL cohort, the observed 10-year survival rates were 0.386 (1983 to 1989) and 0.438 (1990-1999). Relative 5-year survival rates were 0.698 (1983 to 1989) and 0.741 (1990 to 1999), and 10-year relative survival rates were 0.522 (1983 to 1989) and 0.608 (1990 to 1999). Among patients with advanced disease, 5-year relative survival rates were 0.591 (1983 to 1989) and 0.653 (1990 to 1999), and 10-year relative survival rates were 0.384 (1983 to 1989) and 0.508 (1990 to 1999).
DISCUSSION
This report is the first to describe a temporal improvement in the survival of patients with FL. Analyzing a large cohort of patients in the SEER registry demonstrated that observed median survival of FL patients diagnosed in the United States improved by 11% from the 1980s to the 1990s, which is reflective of an adjusted death hazard ratio that decreased by nearly 25% from 1983 to 1999. Using multiple analytic methods, we noted a consistent trend of improved survival since 1978 for patients with FL. The relative survival rates obtained by SEER * Stat analysis demonstrated that the survival improvement observed among FL patients has occurred independent of changes in the general population. The survival advantage was not demonstrated in all patient subsets, most notably the black population. The most striking improvement was demonstrated in patients with advanced (stage III/IV) disease, which was the group of patients most likely to receive sequential systemic treatment.
Several case series of FL and indolent lymphoma have been analyzed by single- and multi-institution investigators. In these analyses, patients were often participants in clinical trials or were treated at large referral centers and, thus, were subject to inherent selection bias. In contrast to previous studies, we used a larger, population-based surveillance program and restricted our analysis to patients with FLs. The large sample size also allowed for subset analyses with acceptable statistical power.
The SEER database is a useful tool to answer selected clinical and epidemiologic questions, yet it has inherent advantages and limitations.23 Whereas single-institution and clinical trial patient series are capable of producing uniform and reliable hematopathology-based cohorts, the identification of FL cohorts in the SEER database is often based on a single interpretation from a variety of pathologists and institutions. However, this apparent weakness may also be a strength of the study in that the cohort identified in the SEER database is more representative of the real-world population of FL patients, and for this reason, the outcomes are important.
The diagnostic concordance of FL between expert hematopathologists is among the highest of the distinct lymphoid malignancies24; however, subclassification is more difficult and has been described previously.25 Two reviews of the SEER data have quantified NHL diagnostic reliability. An expert review of the Iowa SEER cancer registry between 1981 and 1984 concluded that FLs may be used confidently in epidemiologic studies, whether separated or grouped.26 A recent review of patients from 1988 to 1994 in the Greater Bay Area Cancer Registry (San Francisco-Oakland SEER registry) also concluded that agreement was acceptable for most FL subtypes.27
An important limitation of the SEER database is the absence of clinical information, such as tumor bulk, performance status, serum lactate dehydrogenase, and hemoglobin concentrations, and hence, the inability to control for prognostic scores such as the International Prognostic Index or International Prognostic Index for Follicular Lymphoma. In our multivariate analyses, we controlled for several available clinical factors, including stage of disease, histologic subtype, patient age, race, and sex. Although these variables contribute to the clinical description of the disease, the SEER database lacks other important characteristics routinely used in clinical practice.
The temporal survival improvements in our study were not observed uniformly across histologic subtypes of FL. Among patients with advanced-stage FSC + FM, which is a group that was likely to receive sequential systemic chemotherapy, a modest improvement in survival over time was observed in both younger and older populations. In contrast, for patients with limited- and regional-stage FSC + FM, the survival advantage was confined to the subset of younger patients who were more likely to receive systemic chemotherapy than the older patients in this setting. The observed survival improvement is most dramatic in younger patients with advanced-stage FLC, which is a group that was highly likely to receive anthracycline-containing chemotherapy.
We found that black patients with FL have not experienced the nearly 1-year survival advantage over the last two decades that white patients experienced. This finding is consistent with an analysis of survival outcomes of the Detroit SEER registry.28 Varterasian et al28 recognized that, among NHL patients in the metropolitan Detroit registry, the 5-year relative survival of black patients decreased over the observation period, whereas survival for whites in the Detroit registry increased. Further investigation is needed to identify potential causes of this disparity.
In November 1997, rituximab was approved by the US Food and Drug Administration for treatment of relapsed low-grade and follicular CD20+ NHL. There has been speculation that the use of this monoclonal antibody therapy may alter the clinical course of FL. Therefore, we conducted an analysis in which the years of diagnosis and patient follow-up were truncated in 1996, which revealed similar survival outcomes to our present study (data not shown). This modified analysis, combined with the joinpoint analysis described here, leads us to conclude that the survival advantage described in this report was observed before the widespread use of this novel therapy in the treatment of FL.
The identification of improved survival among patients with FL has important clinical and public health implications. Several clinical trials have demonstrated enhancement of event-free or progression-free survival in this patient population with various treatment regimens, but the trials have been too small and follow-up too short to demonstrate improvement in survival. Similarly, supportive care has improved throughout the observation period and may have contributed to population-based survival improvements in patients with lymphoma, despite the fact that individual supportive care trials in patients with lymphoma have not demonstrated differences in survival. The clinical course of FL has been altered, which is coincident with the changes in availability of management options. We speculate that the sequential application of these effective therapies, coupled with improved supportive care, is responsible for the improvement in survival in patients with FL. This finding justifies the study and development of sequential therapies, in addition to new agents and regimens, to further enhance survival in patients with FL.
Authors' Disclosures of Potential Conflicts of Interest
The authors indicated no potential conflicts of interest.
NOTES
Supported in part by National Cancer Institute grant Nos. T32 CA79445, HL 04117, and P50 CA097274.
Presented in part at the 40th Annual Meeting of the American Society of Clinical Oncology, New Orleans, LA, June 5-8, 2004 (abstract ID: 1090).
Authors' disclosures of potential conflicts of interest are found at the end of this article.
REFERENCES
Groves FD, Linet MS, Travis LB, et al: Cancer surveillance series: Non-Hodgkin's lymphoma incidence by histologic subtype in the United States from 1978 through 1995. J Natl Cancer Inst 92:1240-1251, 2000
Ardeshna KM, Smith P, Norton A, et al: Long-term effect of a watch and wait policy versus immediate systemic treatment for asymptomatic advanced-stage non-Hodgkin's lymphoma: A randomized controlled trial. Lancet 362:516-522, 2003
Brandt L, Kimby E, Nygren P, et al: A systematic overview of chemotherapy effects in indolent non-Hodgkin's lymphoma. Acta Oncol 40:213-223, 2001
Peterson BA: Current treatment of follicular low-grade lymphomas. Semin Oncol 26:2-11, 1999 (suppl 14)
Horning SJ: Follicular lymphoma: Have we made any progress? Ann Oncol 11:S23-S27, 2000 (suppl 1)
Horning SJ: Natural history of and therapy for the indolent non-Hodgkin's lymphomas. Semin Oncol 20:75-88, 1993 (suppl 5)
Johnson PW, Rohatiner AZ, Whelan JS, et al: Patterns of survival in patients with recurrent follicular lymphoma: A 20-year study from a single center. J Clin Oncol 13:140-147, 1995
National Cancer Institute: Surveillance, Epidemiology, and End Results (SEER) Program public-use data (1973-1999) program Web site: Data quality. http://seer.cancer.gov/about/quality.html
Zippin C, Lum D, Hankey BF: Completeness of hospital cancer case reporting from the SEER Program of the National Cancer Institute. Cancer 76:2343-2350, 1995
National Cancer Institute: Surveillance, Epidemiology, and End Results (SEER) Program Web site: Recoded SEER public access record description, cases diagnosed in 1973-2000, user file. http://seer.cancer.gov/manuals/CD2.SEERDic.pdf
American Cancer Society: Manual of Tumor Nomenclature and Coding. New York, NY, American Cancer Society, 1968
WHO: International Classification of Diseases for Oncology (ed 1). Geneva, Switzerland, WHO, 1976
Percy C, Van Holten V, Muir C (eds): International Classification of Diseases for Oncology (ed 2). Geneva, Switzerland, WHO, 1990
Van Holten V, Percy C: Conversion of neoplasms by topography and morphology from the International Classification of Diseases for Oncology (ICD-O) (First Edition and Field Trials) to International Classification of Diseases for Oncology, Second Edition (ICD-O-2). Bethesda, MD, Surveillance Program, Division of Cancer Prevention and Control, National Cancer Institute, NIH publication 98-3331, 1991
Greene FL, Page DL, Fleming ID, et al: AJCC Cancer Staging Manual (ed 6). New York, NY, Springer-Verlag, 2002
National Cancer Institute: Surveillance, Epidemiology, and End Results (SEER) 1973-2000 public-use data (1973-2000). http://seer.cancer.gov/publicdata
Allison PD: Survival Analysis Using SAS: A Practical Guide. Cary, NC, SAS Institute Inc, 1995
Henson DE, Ries LA: The relative survival rate. Cancer 76:1678-1688, 1995
National Cancer Institute: SEER * Stat Version 5.0.20. http://seer.cancer.gov/seerstat/software
Allison PD: Logistic Regression Using SAS System: Theory and Application. Cary, NC, SAS Institute Inc, 1999
National Cancer Institute: Joinpoint Regression Program, Statistical Research Applications Branch, Division of Cancer Control and Population Sciences. http://srab.cancer.gov/joinpoint
Kim H, Fay MP, Feuer EJ, et al: Permutation tests for joinpoint regression with applications to cancer rates. Stat Med 19:335-351, 2000
Harlan LC, Hankey BF: The Surveillance, Epidemiology, and End Results program database as a resource for conducting descriptive epidemiologic and clinical studies. J Clin Oncol 21:2232-2233, 2003
The Non-Hodgkin's Lymphoma Classification Project: A clinical evaluation of the International Lymphoma Study Group classification of non-Hodgkin's lymphoma: The Non-Hodgkin's Lymphoma Classification Project. Blood 89:3909-3918, 1997
Metter GE, Nathwani BN, Burke JS, et al: Morphologic subclassification of follicular lymphoma: Variability of diagnoses among hematopathologists, a collaborative study between the Repository Center and Pathology Panel for Lymphoma Clinical Studies. J Clin Oncol 3:25-28, 1985
Dick F, VanLier S, Banks P, et al: Use of the working formulation for non-Hodgkin's lymphoma in epidemiologic studies: Agreement between reported diagnoses and a panel of experienced pathologists. J Natl Cancer Inst 78:1137-1144, 1987
Clarke CA, Glaser SL, Dorfman RF, et al: Expert review of non-Hodgkin's lymphomas in a population-based cancer registry: Reliability of diagnosis and subtype classifications. Cancer Epidemiol Biomarkers Prev 13:138-143, 2004
Varterasian ML, Graff JJ, Severson RK, et al: Non-Hodgkin's lymphoma: An analysis of the Metropolitan Detroit SEER database. Cancer Invest 18:303-308, 2000(Wade T. Swenson, James E.)