Fax: (011) 45 6591 2943
Version of Record online: 3 APR 2006
Copyright © 2006 American Cancer Society
Volume 106, Issue 10, pages 2165–2170, 15 May 2006
How to Cite
Moller, M. B., Pedersen, N. T. and Christensen, B. E. (2006), Conditional survival of patients with diffuse large B-cell lymphoma. Cancer, 106: 2165–2170. doi: 10.1002/cncr.21877
Institutions (and physicians) cooperating in this study: Aalborg Sygehus (M.K. Jensen, P. Johansen); Aarhus Universitetshospital (A.M. Boesen, E.R. Hansen, K.B. Hansen); Sydvestjysk Sygehus (T. Mourits-Andersen); Herning Centralsygehus (M.R. Madsen); Odense Universitetshospital (B.E. Christensen, N.T. Pedersen, H.B. Rasmussen); Randers Centralsygehus (M. Brandsborg); Sonderborg Sygehus (M. Brons); Vejle Sygehus (O. Gadeberg); Viborg-Kjellerup Sygehus (B.B. Pedersen).
Informed consent was obtained from all patients entered in the LYFO registry.
- Issue online: 27 APR 2006
- Version of Record online: 3 APR 2006
- Manuscript Accepted: 3 JAN 2006
- Manuscript Revised: 22 DEC 2005
- Manuscript Received: 7 OCT 2005
- conditional survival;
- diffuse large B-cell lymphoma
Prognosis of lymphoma patients is usually estimated at the time of diagnosis and the estimates are guided by the International Prognostic Index (IPI). However, conditional survival estimates are more informative clinically, as they consider those patients only who have already survived a period of time after treatment. Conditional survival data have not been reported for lymphoma patients.
Conditional survival was estimated for 1209 patients with diffuse large B-cell lymphoma (DLBCL) from the population-based LYFO registry of the Danish Lymphoma Group. The Kaplan-Meier method was used to estimate conditional survival at 0-5 years after diagnosis.
The probability of surviving 5 years increases with each year survived for the first 3 years after diagnosis, after which the increase is minimal. The median survival increases from 38 months for all patients to between 108 and 120 months, conditioned on survival for at least 3-5 years. The prognostic capacity of the IPI and the age-adjusted IPI was high at diagnosis, but the significance gradually declined in the first years after diagnosis. Furthermore, the prognostic impact of the individual clinical variables of the IPI was also significant at diagnosis, but 2 years after diagnosis only age had prognostic impact. Multivariate analysis of patients who survived ≥3 years identified only age as a prognostic factor.
For patients with DLBCL who have survived more than 1 year after diagnosis, the conditional survival probability provides more accurate prognostic information than the conventional survival rate estimated from the time of diagnosis. Cancer 2006. © 2006 American Cancer Society.
Diffuse large B-cell lymphoma (DLBCL) is the most frequently occurring lymphoma in Western countries and is the lymphoma entity with the highest mortality.1 Approximately 40% to 50% of the patients with DLBCL can be cured with anthracycline-containing combination chemotherapy regimens. The cure rate depends on several pretreatment prognostic variables from which the widely used International Prognostic Index (IPI) is calculated.2 The projection of survival is traditionally made at the time of diagnosis using the IPI. However, these overall survival projections are often discouraging, and not necessarily pertinent, for patients who have survived the initial treatment period. Patients who have survived for a specified interval of time (for example, 2 years or longer) after diagnosis have a different probability of surviving the following 5 years than was estimated at the time of diagnosis. Survival probability calculated for patients who have already survived a given period of time is called “conditional survival.”3 Thus, conditional survival data provide physicians and their patients with a more meaningful measure of survival probability after an initial survival period.4
Conditional survival rates have been published for breast cancer,3, 5 colon cancer,4, 5 brain tumors,6 lung cancer,5, 7, 8 and prostate cancer,5 and is considered a particularly useful measure for patients with malignancies characterized by high mortality rates in the first 1 or 2 years after diagnosis followed by more stable survival.9 Thus, as most deaths of DLBCL patients occur within 2 years of diagnosis, and there still is a percentage of deaths occurring after 2 years,10 data on conditional survival of DLBCL patients are warranted. The objective of this investigation was to estimate survival probabilities for DLBCL patients conditioned on surviving for 1, 2, 3, 4, or 5 years.
MATERIALS AND METHODS
The population-based lymphoma registry of the Danish Lymphoma Group includes all newly diagnosed non-Hodgkin lymphoma patients in the region of Western Denmark since 1983, thereby prospectively collecting clinical and pathological data without selection bias. Three hematopathologists centrally review cases entered in the registry. Excluded from registration were cases of acute lymphoblastic leukemia, chronic lymphocytic leukemia, and multiple myeloma. Staging procedures included clinical history, physical examination including all lymphoid regions and ear-nose-throat examination, biopsy from involved tissue, bone marrow aspirate and biopsy, complete blood cell count, serum chemistry analysis, chest X-ray and/or thoracic CT scan, abdominal CT scan and/or lymphangiography, and abdominal ultrasonography. The Ann Arbor staging system was used. Complete response (CR) was defined as the resolution of all pretreatment clinical and laboratory evidence of disease for at least 1 month.
In all, 4786 patients age 15 or older were registered in the period 1983 to 1998. Of these, 1575 patients had DLBCL with centroblastic or immunoblastic morphology.11 Exclusion of patients diagnosed at autopsy, cases with incomplete staging or incomplete data for calculation of the IPI,2 and cases not treated with curative intent left 1209 patients in the study. The clinical data were reported previously.11 The median age at diagnosis was 65 years. Six hundred sixty-three (55%) patients presented with localized disease. IPI and age-adjusted IPI were calculated using the variables age, performance status, Ann Arbor stage, extranodal involvement, and serum lactate dehydrogenase (LDH) level as detailed by the International Non-Hodgkin's Lymphoma Prognostic Factors Project, but because of relatively few long-term survivors, the high-intermediate and high-risk groups were merged in the age-adjusted IPI.2
The patients were treated following common general LYFO guidelines.12 The majority of cases with localized disease were treated with involved-field radiotherapy supplemented with 6-9 courses of CHOP (cyclophosphamide, doxorubicin, vincristine, prednisone) or CHOP-like regimens. A minority were treated locoregionally only (radiotherapy and/or surgery). Standard treatment of patients with disseminated disease was 9 courses of CHOP or CHOP-like regimens. Additional radiotherapy was given for bulky tumors (>5 cm) or persistent lesions after chemotherapy.
CR was achieved in 811 (67%) patients, and 393 patients were alive at the end of the observation period, with a median follow-up time of 7.8 years (range, 3.0-18.8 years). The median survival was 38 months, with a 5-year overall survival rate of 44%. No patients were lost to follow-up.
Overall survival was defined as the duration between the date of histological diagnosis and the date of death of any cause or date last known alive. Conditional survival was defined as the duration between the date of histological diagnosis plus X years and the date of death of any cause or date last known alive for patients alive X years after the date of diagnosis. To illustrate, conditional survival for patients who survived 2 years or longer was estimated with date of diagnosis plus 2 years as the starting point. Overall and conditional survival rates were estimated using Kaplan-Meier estimators and were compared applying the log rank test. All analyses were performed on an intent-to-treat basis. The chi-square test was used for univariate comparison of categorical frequency data. Factors independently affecting survival were identified using a Cox proportional hazards regression model. All P values are 2-sided and considered significant if <.05. SPSS for Windows (Chicago, IL, v. 12.0.0) was used for all calculations.
Survival curves of all patients and survival curves conditioned on patients having already survived 1, 2, 3, 4, or 5 years after diagnosis are provided in Figure 1. The probability of surviving 5 years increases with each year survived for the first 3 years after diagnosis, after which the increase is minimal. The median survival increases from 38 months for all patients to between 108 and 120 months conditioned on survival 3-5 years (Fig. 2). Five years after diagnosis the average age of the surviving patients was 63 years, and these patients had a median conditional survival of 10 years. However, the average life-expectancy of a 63-year-old individual in Denmark is 17 years.13 Thus, having suffered from DLBCL but survived the critical first 5 years is associated with a reduction in life-expectancy of approximately 7 years.
Table 1 shows the distribution of patients according to the IPI and the individual clinical variables of the IPI at the time of diagnosis and at 1-10 years after diagnosis. The distribution of patients in the IPI risk stratification changes significantly over time. However, this effect is only seen in the first 5 years. At diagnosis, 28% of the patients were allocated to the high-intermediate and high-risk groups. Only 11% of patients who survived 5 years were from these risk groups (P <.01). In contrast, comparison of patients who survived to 5 years with those with 10 years survival showed that the rate was only reduced from 11% to 9% (P = .77). The same observation could be made by analyzing the individual clinical IPI variables. Stage (P < .01 vs. P = .89), performance score (P <.01 vs. P = .80), LDH (P <.01 vs. P = .90), and extranodal disease (P <.01 vs. P = .67) showed the same highly significant differences in the first 5-year span and no difference in the second 5-year span. Only for age was a difference identified in both time spans (P <.01 vs. P = .05).
|Years After Diagnosis|
|Year 0||Year 1||Year 2||Year 3||Year 4||Year 5||Year 6||Year 7||Year 8||Year 9||Year 10|
|n = 1209||n = 830||n = 694||n = 614||n = 520||n = 439||n = 363||n = 314||n = 257||n = 217||n = 178|
|No. (%)||No. (%)||No. (%)||No. (%)||No. (%)||No. (%)||No. (%)||No. (%)||No. (%)||No. (%)||No. (%)|
|IPI risk groups|
|Low||556 (46)||473 (57)||421 (61)||389 (63)||339 (65)||291 (66)||242 (67)||214 (68)||178 (69)||153 (71)||117 (66)|
|Low-intermediate||316 (26)||215 (26)||179 (26)||148 (24)||121 (23)||101 (23)||84 (23)||68 (22)||56 (22)||47 (22)||45 (25)|
|High-intermediate||190 (16)||103 (12)||72 (10)||58 (9)||45 (9)||34 (8)||28 (8)||24 (8)||18 (7)||13 (6)||13 (7)|
|High||147 (12)||39 (5)||23 (3)||19 (3)||15 (3)||13 (3)||9 (3)||8 (3)||5 (2)||4 (2)||3 (2)|
|≤60||485 (40)||376 (45)||327 (47)||302 (49)||263 (51)||232 (53)||199 (55)||176 (56)||150 (58)||131 (60)||110 (62)|
|>60||724 (60)||454 (55)||367 (53)||312 (51)||257 (49)||207 (47)||164 (45)||138 (44)||107 (42)||86 (40)||68 (38)|
|Ann Arbor stage|
|I-II||663 (55)||517 (62)||460 (66)||416 (68)||361 (69)||305 (70)||250 (69)||218 (69)||181 (70)||155 (71)||122 (69)|
|III-IV||546 (45)||313 (38)||234 (34)||198 (32)||159 (31)||134 (31)||113 (31)||96 (31)||76 (30)||62 (29)||56 (32)|
|Extranodal sites, no.|
|0-1||1029 (85)||739 (89)||622 (90)||552 (90)||473 (91)||401 (91)||333 (92)||287 (91)||237 (92)||198 (91)||160 (90)|
|≥2||180 (15)||91 (11)||72 (10)||62 (10)||47 (9)||38 (9)||30 (8)||27 (9)||20 (8)||19 (9)||18 (10)|
|Normal||821 (68)||637 (77)||556 (80)||495 (81)||431 (83)||361 (82)||301 (82)||258 (82)||213 (83)||180 (83)||145 (82)|
|High||388 (32)||193 (23)||138 (20)||119 (19)||89 (17)||78 (18)||62 (17)||56 (18)||44 (17)||37 (17)||33 (19)|
|WHO performance score|
|0-1||851 (70)||676 (81)||580 (84)||524 (85)||443 (85)||377 (86)||313 (86)||274 (87)||226 (88)||191 (88)||155 (87)|
|≥2||358 (30)||154 (19)||114 (16)||90 (15)||77 (15)||62 (14)||50 (14)||40 (13)||31 (12)||26 (12)||23 (13)|
Five-year conditional survival estimates for the IPI risk groups are presented in Figure 3. The estimated 5-year survival probabilities increase and then level off when conditioned on successive years already survived. The conditional survival is steeper, and thus more favorable, in the higher-risk groups. The 5-year survival probability under the condition of having survived 5 years for each of the IPI risk groups is between 65% and 78%. Compared with the 5-year survival probabilities at the time of diagnosis of 11% to 62%, the differences between the risk groups significantly level off when conditioned on survival. The prognostic capacity of IPI is thus very high at diagnosis (P<.01; log rank, 320.26), but is only of borderline significance for patients with at least 5 years survival (P = .06; log rank, 7.46).
Patients with Stage III/IV disease at presentation had a poor prognosis, with a 5-year overall survival of 31% compared with 54% for patients with Stage I/II disease (P <.01). However, for patients who survived at least 2 years the difference in survival was eliminated, with 5-year conditional survival rates of 67% and 64%, respectively (P = .70).
Age is a strong determinant of long-term prognosis. We have estimated the conditional survival probabilities of patients in 4 age groups (Fig. 4). In all 4 age groups the conditional survival curves are steep in the first 3 years, after which they level off. The increase in conditional survival is steepest in the youngest age group (≤49 years old), where the 5-year survival probability increases from 62% at the time of diagnosis to 93% conditioned on having survived 3 years. In the oldest age group (≥70 years old) the increase is less dramatic (from 29% to 46%). However, age is related to both overall survival (P <.01; log rank, 133.68) and survival conditioned on 1, 2, 3, 4, and 5 years survival or longer (P <.01; log rank, 73.7-116.69).
The 5-year conditional survival of patients <60 years old was significantly influenced by the age-adjusted IPI 1 and 2 years after diagnosis (P <.01), but after 3 years there were no differences between the risk groups (P = .11; Fig. 5). In the low-risk group the survival probability is constant for every year survived, whereas there is a steep increase in the conditional survival the first 3 years after diagnosis for patients in the other risk groups. For patients ≥60 years, the conditional survival steadily increased for the first 3 years and thereafter leveled off in all 4 IPI risk groups (Fig. 6). The conditional survival differences between the risk groups are only significant 1 year after diagnosis (P <.01); thereafter, they become nonsignificant (data not shown).
For most of the variables analyzed, the 5-year conditional probability of survival approximates a constant rate 3 years after diagnosis. To identify potential prognostic factors for patients who survived ≥3 years, we analyzed the prognostic significance of the IPI and the individual IPI variables of these patients. Univariate analysis identified age ≥60 years (P <.01) and IPI (P <.01) as related to survival, but in a multivariate analysis only age retained significance (relative risk 4.1; P <.01).
Long-term follow-up is required to calculate the clinically very informative measure of conditional survival. To calculate the conditional probability of surviving 5 years after having survived 5 years already requires a 10-year follow-up. In the LYFO registry of the Danish Lymphoma Group we prospectively collected population-based clinical and outcome data since 1983. Thus, long-term follow-up is available through the LYFO registry. For the present study, we chose DLBCL patients diagnosed between 1983 and 1998 having a median follow-up time of 7.8 years. Thus, the patients are from a period with a rather stable treatment approach, with CHOP as the gold standard, and before the introduction of rituximab,14 thereby making the overall survival rates comparable.
Our study demonstrates that survival probability is strongly dependent on time lapsed since diagnosis. The median survival of all patients from the time of diagnosis was 3.2 years, whereas it increased to 9-10 years 2-5 years after diagnosis.
The IPI is currently the standard prognostic index for DLBCL. As expected, the IPI had a strong predictive capacity at the time of diagnosis. However, the impact of the IPI decreased with every year survived and was nonsignificant after 5 years. These results are supported by the original IPI publication, where the death rates within the IPI risk groups were similar 2-3 years after diagnosis.2 Thus, the IPI is not informative for individual patients who have survived several years after diagnosis.
Of the individual IPI variables, stage, performance score, LDH, or number of extranodal sites were not prognostically significant 3 years after diagnosis. Age was the only IPI factor of prognostic importance for conditional survival 3 years after diagnosis. A study of the time-dependent effect of the age-adjusted IPI variables (thus not including age) on patients younger than 60 years with aggressive non-Hodgkin lymphoma (NHL) identified only stage as a predictor of long-term survival.10 This is at variance with our results, where stage did not influence conditional survival 3 years after diagnosis, neither overall nor for patients younger than 60 years. Even among the patients younger than 60 years, age was the only predictive clinical pretreatment variable when comparing patients older or younger than 50 years (data not shown). The effect of age on long-term survival may at least partially reflect the older patients reduced ability to tolerate intensive salvage therapy, age-related propensity for late doxorubicin-induced cardiomyopathy,15 and increased mortality from other medical conditions.
Being diagnosed with DLBCL is associated with reduced life expectancy. We show that even among the patients with the most favorable outcome, i.e., those who survived the first 5 years after diagnosis, life expectancy was still reduced by 7 years. This may in part be related to late recurrences of lymphoma, but late, mainly cardiovascular, effects of chemotherapy and radiotherapy may also contribute significantly.15, 16
The conditional survival data presented in the present report provide more meaningful survival projections than conventional estimates. This information should be useful to both patients who wish to be informed of their prognosis after an initial survival period as well as to their healthcare professionals. It should be noted that the data are from the prerituximab era, and conditional survival statistics should be reevaluated when long-term follow-up of rituximab-treated DLBCL patients becomes available.
- 11World Health Organization classification of tumours. Pathology and genetics of tumours of haematopoietic and lymphoid tissues. Lyon: IARC Press, 2001., , , .
- 13StatBank Denmark. Available from URL: http://www.statbank.dk [accessed Sept. 14, 2005].