Fax: (011) 33 231455097
Survival after Hodgkin lymphoma†
Causes of death and excess mortality in patients treated in 8 consecutive trials
Article first published online: 10 FEB 2009
Copyright © 2009 American Cancer Society
Volume 115, Issue 8, pages 1680–1691, 15 April 2009
How to Cite
Favier, O., Heutte, N., Stamatoullas-Bastard, A., Carde, P., van't Veer, M. B., Aleman, B. M. P., Noordijk, E. M., Thomas, J., Fermé, C., Henry-Amar, M. and for the European Organization for Research and Treatment of Cancer (EORTC) Lymphoma Group and the Groupe d'Études des Lymphomes de l'Adulte (GELA) (2009), Survival after Hodgkin lymphoma. Cancer, 115: 1680–1691. doi: 10.1002/cncr.24178
We acknowledge the following individuals for their participation in the study: Umberto Tirelli, MD, PhD, Division of Medical Oncology A, National Cancer Institute, Aviano, Italy; John M. M. Raemaekers, MD, PhD, St. Radboud University Medical Center, Nijmegen, the Netherlands; Houchingue Eghbali, MD, Bergonié Cancer Institute, Bordeaux, France; Pieternella J. Lugtenburg, MD, PhD, University Medical Center, Rotterdam, the Netherlands (European Organization for Research and Treatment of Cancer Lymphoma Group); and Pauline Brice, MD, PhD, Saint-Louis University Hospital, Paris, France; Catherine Sebban, MD, Léon-Bérard Cancer Center, Lyon, France; Laurent Voillat, MD, Besançon University Hospital, Besançon, France; René-Olivier Casasnovas, MD, PhD, Dijon University Hospital, Dijon, France (Adult Lymphoma Study Group).
- Issue published online: 6 APR 2009
- Article first published online: 10 FEB 2009
- Manuscript Accepted: 16 OCT 2008
- Manuscript Revised: 30 SEP 2008
- Manuscript Received: 7 AUG 2008
- Foundation for Medical Research
- Hodgkin lymphoma;
- long-term survival;
- causes of death;
- relative survival
The objective of this study was to analyze cause-specific excess mortality in adult patients with Hodgkin lymphoma (HL) with respect to treatment modality.
The study population consisted of 4401 Belgian, Dutch, and French patients aged 15 to 69, in all stages of disease, who were treated between 1964 and 2000. Excess mortality was expressed by using a standardized mortality ratio (SMR) and calculating the absolute excess risk (AER). Relative survival was calculated and analyzed using a previously described regression model.
At a median follow-up of 7.8 years, 725 of 4401 patients (16.5%) had died, 51% of HL, 10% of treatment-related toxicity, 18% of second cancer, 5% of cardiovascular diseases, 2% of infections, 8% of other causes, and 6% of an unspecified cause. Overall, the SMR was 7.4 (95% confidence limits [CL], 6.9-8.0), and the AER was 182.8 (95% CL, 167.7-198.8). These indicators were 3.8 (95% CL, 3.2-4.5) and 27.9, respectively, for deaths from a second cancer and 4.0 (95% CL, 2.3-6.7) and 3.3, respectively for deaths from infection. After 15 years, the observed survival rate was 75%, and the relative survival rate was 80%. In patients with early-stage disease, the overall excess mortality was associated with age ≥40 years (P = .007), men (P < .001), unfavorable prognosis features (P < .001), and 2 treatments: combined nonstandard nonalkylating chemotherapy plus involved-field radiotherapy (P = .002) and mantle-field irradiation alone (P = .003). With follow-up censored at the first recurrence, no treatment modalities were associated with excess mortality.
Progressive disease remained the primary cause of death in patients with HL in the first decades after treatment. Excess mortality in patients with early-stage disease was linked significantly to treatment modalities that were associated with poor treatment failure-free survival. Cancer 2009. © 2009 American Cancer Society.
Hodgkin lymphoma (HL) has become a highly curable cancer.1, 2 Patients have a relatively long survival, although a few studies have concluded that the death rate of patients with HL still is greater than that of the general population.3, 4 This may relate to various complications, such as second malignancies,5-9 cardiac toxicity,10-12 and infections.4 Twenty years after their treatment, more patients have died from other causes than from HL.4, 13 Because treatment efficiency has improved dramatically, 1 way to improve long-term survival is to reduce mortality from causes other than HL.
Several studies have examined excess mortality from major causes in large cohorts of patients with HL,3-5, 8, 13 but few have analyzed excess mortality by detailed treatment categories.12 However, Swerdlow et al failed to identify treatment categories (with detailed chemotherapy dose or dose and field of radiation used) linked to an excess risk of death from myocardial infarction.12 By using data from randomized clinical trials of the European Organization for Research and Treatment (EORTC) Lymphoma Group and the Adult Lymphoma Study Group (GELA), we conducted a retrospective study with the objectives of estimating excess mortality from causes other than HL and examining mortality with respect to patient and clinical characteristics and treatment modality.
MATERIALS AND METHODS
Patients and Data Collection
The cohort was composed of patients who were included in 8 consecutive clinical trials conducted from 1964 to 2000 in 10 European countries. The selection of patients and methods of data collection were published previously.14-17 The data used concerned nationality, date of birth, sex, clinical stage according to the Ann Arbor classification, date of randomization, treatment, date of first recurrence, date of last examination and vital status, date of death, and primary cause of death. Causes of death were reported as follows: progressive disease, treatment-related toxicity without evidence of active disease, second cancer, cardiovascular disease, infection, other causes, and death from cause unspecified. Overall, 5015 patients were included. Data were stored under the responsibility of the same researcher (M.H.-A.).
General Population Data
To estimate excess mortality, survival data from the studied population and mortality data from the general population were needed. Death rates were calculated using the numbers of deaths and individuals at risk obtained from the World Health Organization Mortality Database.18 Causes of deaths were specified using the International Classification for Diseases (revisions 7-10).19 Death rates were derived for each combination of 5-year age group, sex, country, and calendar year, by dividing the numbers of deaths from given causes (all causes, cancer, cardiovascular diseases, infections, and other causes) by the mid-year number of individuals at risk. Only patients who originated from countries where >5% of all patients in the database resided were selected (Fig. 1). Population data were available until 2004 for the Netherlands, until 2003 for France, and until 1997 for Belgium. Therefore, 23 Belgian patients who enrolled on trials in 1998 or later were excluded from the current analysis, and 4401 patients (88%) were available for analysis.
Time at risk was computed from the date of randomization to the date of death or the date of the last examination. Time at risk was limited to 15 years to prevent the possibility that older trials would overweight the last follow-up periods, and this lead to the censoring of 477 patients (ie, 1684 person-years), including 142 patients from the H1 trial, 83 patients from the H2 trial, 133 patients from the H5 trial, and 119 patients from the H6 trial. Time at risk also was censored when last vital status date exceeded the year for which general mortality data were available.
The standardized mortality ratio (SMR) was calculated as the ratio of observed deaths to expected deaths. The expected numbers of deaths, overall and for main causes, were calculated by multiplying the numbers of person-years by sex, 5-year age group, nationality, and calendar year by the corresponding mortality rates from general population data. The absolute excess risk (AER) of death was computed as the difference between observed and expected numbers of deaths divided by the number of person-years at risk and multiplied by 10,000. The SMR and AER 95% confidence limits (CL) were obtained assuming the Poisson distribution of the observed numbers.
Survival analysis was performed using the method of Kaplan and Meier. The relative survival method was used to correct observed survival from background mortality. Two approaches were used: the first compared observed survival with relative survival, and the second aimed at analyzing factors that may influence excess risk of death.18 In the latter approach, the relative excess risks of death from baseline characteristics were quantified based on the exponential of the estimates derived from the model described by Dickman et al.20 Baseline characteristics included age (ages 15-39 years, 40-49 years, and 50-69 years), sex, prognostic group, splenectomy, and treatment. Radiotherapy (RT) was subgrouped into involved-field (IF-RT), mantle-field RT and (sub)total lymph node irradiation. Chemotherapy was categorized into nonalkylating therapy, alkylating agent-containing therapy, 3 or 4 cycles, and 6 cycles. Nonalkylating chemotherapy included vinblastine (associated with mantle-field RT in the H1 trial and [sub]total lymph node irradiation in the H2 trial); combined doxorubicin, bleomycin, vinblastine, and dacarbazine (ABVD) (associated with mantle-field RT in the H6 trial); and combined epirubicin, bleomycin, vinblastine, and prednisone (EBVP) (associated with involved-field RT in the H7 trial). Alkylating chemotherapy consisted of procarbazine (associated with vinblastine in the H2 trial); combined mechlorethamine, vincristine, procarbazine, and prednisone (MOPP) (alone or combined with ABVD and iceberg RT in the H3B4 trial; MOPP alone and mantle-field RT in the H5 and H6 trials; and MOPP combined with doxorubicin, bleomycin, vinblastine [ABV] with or without RT in the H7, H8, and H34 trials). RT and chemotherapy categories were pooled into 1 variable (Table 1) in which the combination of IF-RT and 3 or 4 cycles of alkylating chemotherapy was considered the reference category. The definitions for prognostic groups were derived from those used in the design of the H7 and H8 trials.14, 17 An unfavorable prognosis was defined either as supradiaphragmatic clinical stage II disease with ≥4 lymph node areas involved, or as no B symptoms and an erythrocyte sedimentation rate (ESR) ≥50 mm in the first hour, or as B symptoms and an ESR ≥30 mm in the first hour, or a mediastinum/thorax ratio ≥0.35. Data were available for 2451 of the 2630 patients who were included in the H6, H7, and H8 trials (Table 2). To analyze the impact of first-line treatment, the same model was used with the time at risk censored at the date of first recurrence. Risk factor analyses were limited to patients who had early-stage disease, because the treatments used depended greatly on disease stage, and only 17% of all patients and 26% of all deaths involved advanced-stage disease (Table 3). A backward selection procedure was used. All statistical tests were 2-sided, and P < .01 indicated statistical significance. The SAS statistical software (release 9.1.3; SAS Institute, Cary, NC) was used.
|Early Stage Disease Trials|
|H1 (1964-1971)||H2 (1971-1977)||H5 (1977-1982)||H6 (1982-1988)||H7 (1988-1993)||H8 (1993-1999)||H1 to H8 Trials (1964-1999)|
|No. of patients accrued||267||264||490||550||710||1370||3651|
|MOPP-ABV×3 or ×4||0||0||0||0||0||0||0||0||0||0||528||39||528||14|
|Mantle-field irradiation and|
|VLB and PCZ||0||0||44||17||0||0||0||0||0||0||0||0||44||1|
|No. of Patients (%)|
|Characteristic||Early-Stage Disease||Advanced-Stage Disease||All Patients|
|No. of patients accrued||3651||750||4401|
|Belgium||378 (10)||131 (17)||509 (12)|
|France||1749 (48)||233 (31)||1982 (45)|
|Netherlands||1524 (42)||386 (52)||1910 (43)|
|Mean age [range], y||33 [15-70]||36 [15-70]||34 [15-70]|
|15-39||2723 (75)||490 (65)||3213 (73)|
|40-49||531 (14)||118 (16)||649 (15)|
|50-69||397 (11)||142 (19)||539 (12)|
|I||1171 (32)||2 (<1)||1173 (27)|
|II||2480 (68)||5 (1)||2485 (56)|
|III||0 (0)||409 (55)||409 (9)|
|IV||0 (0)||334 (44)||334 (8)|
|PY, uncensored data||30,052||4282||34,334|
|PY, censored data at recurrence*||26,119||3828||29,947|
|Median FU for uncensored data [range], y||8.1 [0-25]||6.2 [0-15]||7.8 [0-25]|
|Median FU for censored data at recurrence [range], y*||7.2 [0-25]||5.3 [0-15]||6.9 [0-25]|
|Cause of Death||H1||H2||H5||H6||H7||H8||H3B4||H34||All Patients|
|No. of patients at risk||267||264||490||550||710||1370||190||560||4401|
|All causes||14/101* [5/38]||42/71 [16/27]||50/106 [10/22]||27/67 [5/12]||30/83 [4/12]||49/110 [4/8]||31/71 [16/37]||65/116 [12/21]||308/725 [7/16.5]|
|Hodgkin lymphoma||0/75 (0/74)||3/27 (7/38)||3/31 (6/29)||1/35 (4/53)||2/40 (7/48)||5/55 (10/50)||18/51 (58/72)||16/52 (25/45)||48/366 (16/51)|
|Treatment-related toxicity||0/1 (0/1)||3/3 (7/4)||6/10 (12/9)||5/6 (18/9)||2/6 (7/7)||13/17 (27/16)||5/8 (16/11)||17/23 (26/20)||51/74 (16/10)|
|Second cancer||2/10 (14/10)||11/14 (26/20)||16/31 (32/29)||7/10 (26/15)||12/17 (40/21)||17/18 (35/16)||4/5 (13/7)||19/25 (29/21)||88/130 (29/18)|
|Cardiovascular diseases||4/6 (29/6)||10/10 (24/14)||7/8 (14/8)||5/7 (19/10)||1/1 (3/1)||0/0 (0/0)||0/1 (0/2)||0/0 (0/0)||27/33 (9/5)|
|Infections||0/0 (0/0)||0/1 (0/1)||4/9 (8/9)||0/0 (0/0)||0/0 (0/0)||0/0 (0/0)||3/5 (10/7)||0/0 (0/0)||7/15 (2/2)|
|Other causes||3/4 (21/4)||7/7 (17/10)||4/5 (8/5)||7/7 (26/10)||6/10 (20/12)||11/13 (22/12)||1/1 (3/1)||11/14 (17/12)||50/61 (16/8)|
|Unknown causes||5/5 (36/5)||8/9 (19/13)||10/12 (20/11)||2/2 (7/3)||7/9 (23/11)||3/7 (6/6)||0/0 (0/0)||2/2 (3/2)||37/46 (12/6)|
Overall, 725 of 4401 patients (16.5%) who were treated for HL died (Table 3). Among the patients with early-stage disease, the 10-year survival rate increased from 70% in the H1 trial to 90% in the H8 trial. Among the patients with advanced-stage disease, the 10-year survival rate was 62% in the H3B4 trial and 73% in the H34 trial. Of the 725 deaths, 366 deaths (51%) were from disease progression, 74 deaths (10%) were from treatment-related toxicity, 130 deaths (18%) were from second cancer, 33 deaths (5%) were from cardiovascular disease, 15 deaths (2%) were caused by infection, 61 deaths (8%) were from other causes, and 46 deaths (6%) were from an unspecified cause. Of the 366 disease progression-related deaths, 82% occurred from 0 to 4 years after randomization, 15% occurred from 5 to 9 years after randomization, and 3% occurred from 10 to 14 years after randomization. The rates were 37%, 42%, and 21%, for deaths from second cancers, respectively; and 60%, 28%, and 12% for deaths from other causes, respectively.
Excess Mortality by All Causes
Overall, the SMR was 7.4 (95% CL, 6.9-8.0), corresponding to 182.8 excess deaths (95% CL, 167.7-198.8 excess deaths) for 10,000 person-years (Table 4). The SMR decreased by 2-fold between 0 to 4 years after randomization and 5 to 9 years after randomization and remained stable thereafter. Overall, the SMR was higher in women than in men (9.1 vs 6.8), whereas the AER was 2 times higher in men (225.3 vs 133). SMR changes with time also differed according to sex: Among women, the SMR decreased by 3-fold and then increased; whereas, among men, it constantly decreased. This also was true for the AER. However, the AER was similar in the last 5 years after randomization in both men and women. The SMR and AER also varied with age. The SMR decreased with increasing age: It was 14.0 (95% CL, 12.7-15.5) in patients ages 15 to 39 years, 7.1 (95% CL, 6.0-8.4) in patients ages 40 to 49 years, and 3.6 (95% CL, 3.1-4.2) in patients ages 50 to 69 years. In contrast, the AER increased with age: It was 145.8 (95% CL, 131.0-161.8) in patients ages 15 to 39 years, 250.6 (95% CL, 204.3-303.0) in patients ages 40 to 49 years, and 370.2 (95% CL, 297.1-451.9) in patients ages 50 to 69 years.
|All Causes of Death||Second Cancer||Infection|
|Age, y||No. of PY at Risk||Obs/Exp||SMR (95% CL)||AER (95% CL)||Obs/Exp||SMR (95% CL)||AER* (95% CL)||Obs/Exp||SMR (95% CL)||AER* (95% CL)|
|0-14||34,334||725/97.3||7.4 (6.9-8.0)||182.8 (167.7-198.8)||130/34.3||3.8 (3.2-4.5)||27.9 (21.7-35)||15/3.7||4 (2.3-6.7)||3.3 (1.4-6.1)|
|0-4||19,531||485/50||9.7 (8.9-10.6)||222.7 (201.1-245.8)||48/17.2||2.8 (2.1-3.7)||15.8 (9.3-23.8)||12/1.9||6.3 (3.3-11.1)||5.2 (2.2-9.8)|
|5-9||11,325||174/35.1||5 (4.2-5.8)||122.6 (100.7-147.2)||55/12.7||4.3 (3.3-5.7)||37.4 (25.4-52)||1/1.4||0.7 (0-4.1)||−0.3 (−1.2-3.7)|
|10-14||3478||66/12.2||5.4 (4.2-6.9)||154.7 (111.7-206.4)||27/4.4||6.1 (4-8.9)||65 (38.5-100.3)||2/0.4||4.5 (0.5-16.1)||4.5 (−0.6-19.5)|
|0-14||15,798||236/25.9||9.1 (8-10.4)||133 (114.6-153.3)||43/10.6||4.1 (2.9-5.5)||20.5 (13-30)||7/0.9||7.5 (3-15.4)||3.8 (1.2-8.5)|
|0-4||8908||167/12||13.5 (11.5-15.7)||173.6 (146.2-204.2)||21/5||4.2 (2.6-6.4)||18 (9-30.4)||5/0.4||11.1 (3.6-25.8)||5.1 (1.3-12.6)|
|5-9||5318||43/10.2||4.2 (3.1-5.7)||61.8 (39.4-89.8)||12/4.2||2.9 (1.5-5)||14.7 (3.8-31.5)||1/0.4||2.6 (0.1-14.7)||1.2 (−0.7-9.8)|
|10-14||1572||26/3.3||7.8 (5.1-11.5)||144.3 (86.9-221.2)||10/1.4||7.2 (3.4-13.2)||54.7 (21.6-108.1)||1/0.1||9.4 (0.2-52.3)||5.7 (−0.5-34.8)|
|0-14||18,536||489/71.4||6.8 (6.3-7.5)||225.3 (202.4-249.7)||87/23.7||3.7 (2.9-4.5)||34.2 (24.8-45.1)||8/2.8||2.9 (1.2-5.7)||2.8 (0.4-7)|
|0-4||10,623||318/37.6||8.4 (7.5-9.4)||263.9 (231.9-298.7)||27/12.2||2.2 (1.5-3.2)||13.9 (5.3-25.5)||7/1.5||4.8 (1.9-10)||5.2 (1.3-12.2)|
|5-9||6007||131/24.9||5.3 (4.4-6.2)||176.5 (140.8-217.2)||43/8.5||5.1 (3.7-6.8)||57.5 (37.7-82.3)||0/1||0 (0-3)||−1.6 (−1.6-3.3)|
|10-14||1906||40/8.9||4.5 (3.2-6.1)||163.3 (103.3-239.2)||17/3||5.7 (3.3-9.1)||73.4 (36.2-127)||1/0.3||2.9 (0.1-16.3)||3.5 (−1.7-27.4)|
Excess Mortality by Cause of Death
Although deaths from second cancer represented 18% of all deaths (Table 3), they accounted for 15% of excess deaths overall (AER, 27.9 vs 182.8) (Table 4). This proportion increased with time from 7% in the period 0 to 4 years after randomization, to 31% in the period 5 to 9 years after randomization, and to 42% in the period 10 to 14 years after randomization. The increase was more pronounced in men than in women, ie, 5%, 33%, and 45% versus 10%, 24% and 38%, respectively. The SMR decreased with increasing age: It was 7.5 (95% CL, 5.5-9.9) in patients ages 15 to 39 years, 3.6 (95% CL, 2.4-5.1) in patients ages 40 to 49 years, and 2.7 (95% CL, 2.0-3.5) in patients ages 50 to 69 years; whereas the AER increased with age: It was 15.9 (95% CL, 11.1-21.9) in patients ages 15 to 39 years, 41.9 (95% CL, 22.3-67.9) in patients ages 40 to 49 years, and 99.6 (95% CL, 60.3-147.7) in patients ages 50 to 69 years.
Cardiovascular diseases represented the third most common cause of death but were not associated with excess mortality. Infections were responsible for excess mortality both overall and by sex. However, significant excess mortality was observed only in the period from 0 to 4 years after randomization.
Among the 15 deaths from infection, 3 deaths occurred in splenectomized patients (SMR, 9.1; 95% CL, 1.9-26.6), and 12 deaths occurred in nonsplenectomized patients (SMR, 3.6; 95% CL, 1.8-6.2). Of these 12 deaths, 10 were observed in patients who received ≥4 cycles of MOPP with or without RT, corresponding to an SMR of 27.1 (95% CL, 13.0-49.9).
Overall, the observed 15-year survival estimate was 74.5% (95% CL, 72.3%-76.6%) compared with an expected rate of 94.1% and leading to a relative survival estimate of 79.7% (Fig. 2A). For patients with early-stage disease, these rates were 76.6% (95% CL, 74.4%-78.8%), 94.4%, and 81.8%, respectively; and for patients with advanced-stage disease, these rates were 69.3% (95% CL, 65.1%-73.6%), 95.6%, and 73.8%, respectively.
The impact of age on survival is illustrated in Figure 2B-D. The differences between relative and observed 15-year survival estimates were 1.8%, 5.9%, and 17.6% in patients ages 15 to 39 years, 40 to 49 years, and 50 to 69 years, respectively. These differences were 2.2% (83.8% vs 81.6%) in women and 5.9% (75.9% vs 70%) in men. In Figure 3, the improvement in treatment efficacy between trials is illustrated for patients with early-stage disease (Fig. 3A,B) and patients with advanced-stage disease (Fig. 3C,D). In patients with early-stage disease, the differences between relative and observed 10-year survival estimates were 2.8%, 3.4%, 3.6%, 2%, 3.3%, and 2.6% for the H1, H2, H5, H6, H7, and H8 trials, respectively. In patients with advanced-stage disease, the estimates were 1.1% for the H3B4 trial and 5.6% for the H34 trial.
Risk Factor Analysis in Patients With Early-Stage Disease
Characteristics that were included in the relative survival model were age group (ages 15-39 years vs 40-49 years vs 50-69 years), sex, prognosis (unfavorable; no vs yes), and treatment. Treatment included splenectomy and an 8-item list of various combinations of chemotherapy and RT (see Table 5). After backward selection, the age group 40 to 69 years, men, an unfavorable prognosis, and 2 treatment modalities (combined EBVP chemotherapy plus IF-RT and mantle-field RT alone) were associated with a significant excess risk of death (P < .01) (see Table 5, Final Model). When censoring the time at risk at first recurrence to analyze the potential impact of initial treatment on excess mortality, no treatment types were associated with a significant excess risk.
|Model With All Treatment Classes||Final Model|
|Variable||No. of Patients at Risk||Median FU, y||No. of Deaths Observed||RER||95% CL||P||RER||95% CL||P||10-y RS Estimate, %|
|MOPP-ABV×3 or ×4||503||7.03||28||1.0||1.0†||92.2|
|Mantle-field irradiation and|
Almost 66% of deaths that occur in the first 2 decades after treatment for HL are related to disease progression or early recurrence (51%) or to acute treatment-related toxicity (10%), which mainly occurs during the first 5 years after diagnosis. Other frequent causes of death that occur later are second cancers, cardiovascular diseases, and infections; and, of these, only deaths from second cancers and infections are significantly in excess. The data also indicate that the AER is higher in men than in women, although the SMR is systematically lower in men because of higher background mortality. Excess mortality from infections is significant only in patients who have undergone splenectomy or who received MOPP chemotherapy.
In a series of 1080 patients with early-stage HL aged ≤50 years, Ng et al reported an SMR (all causes) of 6.4 and an AER of 104.2.3 In our series, the SMR was comparable although slightly higher (7.4) (Table 4). According to time period, Ng et al reported SMRs of 9.9, 6.2, and 4.8 in the follow-up periods at 0 to 4 years, 5 to 9 years, and 10 to 14 years, respectively, which were very similar to our findings. In contrast, their AERs were much lower (116.7, 89.3, and 87.3, respectively) compared with the AERs in our series (222.7, 122.6, and 154.7, respectively). These differences probably account for age, because our patients were almost 9 years older. Aleman et al reported on 1261 patients aged ≤40 years in all disease stages, including 534 patients who died.4 In their series, the AER decreased by 5-year periods from >350 to 150 until Year 15 and then increased to 330 until Year 30 after diagnosis. In both series, deaths from HL represented the major cause of death during the period from 0 to 10 years after diagnosis. Among deaths that were not related to HL, significant excess risk was observed for second cancers and cardiovascular diseases, mainly ≥15 years after diagnosis and treatment initiation. In our series, no significant excess in deaths from cardiovascular disease was observed despite chest RT, probably because the time at risk was limited to 15 years and because our patients were older.21 Nevertheless, deaths from myocardial infarction can occur in excess even a short time after treatment initiation, as demonstrated recently12: In a study of 7033 patients who were treated between 1979 and 1999, a significant excess risk of myocardial infarction was observed from the first year after diagnosis and for 25 years. The SMR ranged between 1.7 and 4.2, and the AER ranged between 4.6 and 28.9. The risk associated with supradiaphragmatic irradiation is always significant (SMR, >2.0; P < .01) as well as the risk associated with anthracyclines alone or in combination with supradiaphragmatic irradiation. However, both the SMR and the AER decreased with increasing calendar year of first treatment, suggesting that modern therapy is associated with a lower risk.
In our study, we used the relative survival model because it estimates net survival (ie, survival corrected for background mortality). This model avoids the problem of inaccuracy in death certificates. The 5-year relative survival estimate (90%) (Fig. 2A) is similar to that published recently by the French Cancer Registry Network (88%) and the American Cancer Society (86% for the period 1996-2003),2, 22 although lower rates also have been reported.1, 23 The 10-year relative survival estimate (85.4%) also was higher than the observed survival in the United States during 2000 to 2004 (80.1%).24 These studies indicate that no significant improvement has been made in survival rates since the 1980s. With longer follow-up, the British National Lymphoma Investigation has reported an estimated 15-year relative survival rate of 74.2% for patients with early-stage disease, an estimate that is lower than ours.25
In our current series, the 10-year expected survival is almost 97%, leading to a difference between relative and crude survival estimates ranging from 2% to 3.5% in patients with early-stage HL and from 1.1% to 5.6% in patients with advanced-stage HL without heterogeneity between trials. These limited differences suggest that deaths unrelated to HL and its treatment do not account for much in patient life expectancy at least during the 15 years after diagnosis. However, the difference is more pronounced with increasing age (Fig. 2B-D) probably because comorbidities are present more often in older patients.25, 26
Prognostic analysis applied to relative survival indicates that being a man, age ≥40 years, unfavorable features at diagnosis, and 2 treatment modalities are associated with a significant (P < .01) excess mortality from all causes (Table 5). Variables that are used to determine patient prognosis, including age, were used in the design of the H6, H7 and H8 trials to predict the risk of early progression/recurrence. However, the current results indicate that the same model also may be useful for predicting excess mortality. We also confirmed in part the results reported by Roy et al, who used the same model from British National Lymphoma Investigation data to demonstrate that being a man and age ≥25 years (with risk increasing with increasing age) both are associated with excess mortality from all causes, like what was reported by Allemani et al.25, 27 Similar results also were reported by Janssen-Heijnen et al based on cancer registry data.28 Of the 2 treatment modalities that were associated with an excess risk of death, 1 concerned the nonstandard nonalkylating EBVP chemotherapy regimen followed by IF-RT, which was used in the H7 trial.17 Therefore, in addition to the excess risk of mortality overall, this treatment was associated with a low survival rate in patients who had early-stage HL with unfavorable features. In contrast, no excess risk of death was associated with the combination of ABVD and IF-RT. The second treatment modality that was associated with a significant excess mortality was mantle-field irradiation alone. This treatment led to very low event-free survival in patients who had a very favorable prognosis29: Now, it has been abandoned even for these patients. The same model applied to patients who were continuously disease-free (ie, with time at risk censored at first recurrence) was unable to identify a treatment modality that was associated with an increased risk of death, suggesting that the excess risk of death associated with the 2 treatments described above probably was caused by disease progression or recurrence and its treatment. Finally, no impact of treatment on excess mortality from second cancer was observed, whereas alkylating chemotherapy and extensive RT were included in the treatment administered to these patients. However, most deaths from second cancers (with the exception of deaths from leukemia) occur ≥10 years after treatment,3-5 and the limited time at risk as well as the limited number of events probably jeopardize the statistical power of the analysis.
Despite the high survival rates, even after modern treatment strategies, long-term complications can be responsible for excess mortality decades after patients have been cured of the disease. In patients with early-stage HL, excess mortality is linked significantly to treatments associated with poor treatment failure-free survival. However, it is probable that not all risk factors of fatal complications have been brought to light; the possible roles of the host and HL per se outline the need for prospective studies and careful long-term follow-up, including a meticulous search for treatment-related complications. Patients also should be offered regular screening programs for relevant second cancers and cardiovascular diseases.
Conflict of Interest Disclosures
Supported by research grants from the Foundation for Medical Research.
- 5Second cancers after treatment of Hodgkin lymphoma. In: HoppeRT,MauchPT,ArmitageJO,DiehlV,WeissLM, eds. Hodgkin Lymphoma,2nd ed. Philadelphia, Pa: Lippincott Williams & Wilkins; 2007: 347-370., , .
- 18World Health Organization. WHO Statistical Information System (WHOSIS), 17 November 2006 update. Available at: http://www.who.int/whosis/mort/download/en/index.html. Accessed January 10, 2007.
- 19World Health Organization. International Statistical Classification of Diseases and Health Related Problems. (The) ICD-10, 2nd ed. Geneva, Switzerland: World Health Organization; 2005.
- 29Combination of radiotherapy and chemotherapy is advisable in all patients with clinical stage I-II Hodgkin's disease. Six-year results of the EORTC-GPMC controlled clinical trials H7-VF, H7-Fand H7-U [abstract]. Int J Radiat Oncol Biol Phys. 1997; 39( suppl 2): 173., , , et al.