Division of Hematology and Oncology, Department of Medicine, Medical University of South Carolina, Charleston, South Carolina
Corresponding author: Luciano J. Costa, MD, PhD, Division of Hematology and Oncology, Department of Medicine, Medical University of South Carolina, 96 Jonathan Lucas St, 903 CSB MSC 635, Charleston, SC 29425-6350; Fax: (843) 792-0644; firstname.lastname@example.org
Presented at the 54th Annual Meeting of the American Society of Hematology; December 8-10, 2012; Atlanta, GA.
Survivors of classical Hodgkin lymphoma (cHL) are at an increased risk of developing secondary non-Hodgkin lymphomas (sNHLs). To the authors' knowledge, the outcome of patients with sNHL compared with their de novo counterparts (dnNHL) is unknown.
Data from 26,826 cases of HL from the Surveillance, Epidemiology, and End Results (SEER) program that were diagnosed between 1992 and 2009 were used to obtain the risk of further development of different subtypes of sNHL. The survival of patients with sNHL was compared with that of matched patients with dnNHL.
The estimated cumulative incidence of sNHL was 2.50% (95% confidence interval [95% CI], 2.10-2.89) at 15 years from the diagnosis of cHL. The standardized incidence ratio was 10.5 (95% CI, 8.9-12.4) for aggressive B-cell NHL, 4.0 (95% CI, 3.1-5.1) for indolent B-cell NHL, and 14.6 (95% CI, 10.3-20.1) for T-cell NHL. Patients with indolent B-cell sNHL had a worse overall survival compared with their dnNHL counterparts (hazards ratio [HR] of death, 2.7; 95% CI, 1.3-5.7). Survival was not significantly different between patients with sNHL and those with dnNHL with regard to aggressive B-cell NHL (HR, 1.3; 95% CI, 0.6-2.7) or T-cell NHL (HR, 0.8; 95% CI, 0.3-1.8).
Classical Hodgkin lymphoma (cHL) is a highly curable hematological malignancy. Successful frontline and salvage therapies have resulted in the cure of > 80% of patients with HL.[1, 2] However, cHL survivors are affected by the risk of second malignancies, which now represent one of the most important causes of late death.[2, 3]
Although the risk of myeloid neoplasms and epithelial solid tumors[5, 6] are better known, particularly with regard to their relation to cHL therapy, a higher-than-expected incidence of non-Hodgkin lymphoma (NHL) has also been reported after the completion of therapy for cHL.[2, 7, 8] In addition, the management of patients with secondary (after cHL) NHL (sNHL) may be challenging due to the cumulated toxicity from cHL treatment. Previous exposure to radiotherapy and systemic chemotherapy adversely alters the outcome of other secondary neoplasms in survivors of cHL by limiting treatment options.[9, 10] To the best of our knowledge, only limited information is available concerning the incidence and outcomes of sNHL after cHL therapy.
In the current study, we used a cohort of 26,826 patients who had cHL as a first malignant neoplasm and were reported to the National Cancer Institute (NCI)'s Surveillance, Epidemiology, and End Results (SEER) program to identify all patients who subsequently developed sNHL. Herein, we describe the patterns of incidence of sNHL after cHL and compare the outcomes of patients with sNHL with their matched counterparts with de novo NHL (dnNHL).
MATERIALS AND METHODS
Incidence of sNHL
We used the NCI's SEER-18 program to assess the incidence of sNHL in survivors of cHL. The SEER-18 registry includes Atlanta, Connecticut, Detroit, Hawaii, Iowa, New Mexico, San Francisco-Oakland, Seattle-Puget Sound, Utah, Los Angeles, San Jose-Monterey, Rural Georgia, Alaska Native Tumor Registry, Greater California, Kentucky, Louisiana, New Jersey, and Greater Georgia. In the “at-risk” population we included all registered cases of cHL (International Classification of Diseases-Oncology, third edition codes 9650/3 to 9667/3) diagnosed on or after 1992 as the first malignant neoplasm for a given individual and known sex and age. We excluded cases of nodular lymphocyte predominant HL due to its different biology, therapy, and natural history. We also excluded cases reported from death certificates or autopsy only.
A search for “at-risk” cHL cases and sNHL was performed using the case list function of SEER*Stat Version 7.1.0 (seer.cancer.gov). Subsequent neoplasms on a given individual were visible by choosing “person selection” in the consult session. To avoid the inclusion of patients with a concomitant cHL and NHL diagnosis, cases with the 2 distinct histologies reported up to 3 months apart as well as patients with cHL and < 3 months of follow-up were excluded from the analysis. Follow-up was current up to the end of 2009.
Calculation of the standardized incidence ratio (SIR) was performed using the observed number of NHL cases in the cHL cohort and the expected number of cases as calculated using the standard rate(s) provided by SEER*Stat. The total number of patient-years of cHL survivors was stratified by age (5-year intervals up to age 84 years and age ≥ 85 years) and sex and, in each stratum, multiplied by the respective annual rate of incidence provided by SEER*Stat. The total expected number of cases was the sum of the expected number of cases in each stratum.
To compare outcomes of patients with sNHL with those of patients with dnNHL, we first listed and characterized all the cases of sNHL that were identified. We then retrieved case lists of dnNHL using the same histological subtypes as for cases of sNHL. Because the timespan of the study overlapped with different lymphoma classification systems (specifically the Working Formulation and the World Health Organization classification), the category used was the one entered in the database and no pathology review or reclassification was possible. The search for cases of dnNHL used the same parameters used to search for HL cases, but was limited to cases in which dnNHL was the first malignant neoplasm registered for a given individual. We subsequently matched dnNHL with sNHL cases in a 2:1 ratio based on the following parameters in decreasing priority: histological subtype (exact), stage (exact), calendar year of diagnosis (± 2 years), age (± 2 years), race (when possible), sex (when possible), and SEER registry (when possible). Case matching was performed without awareness of any other variable associated with the cases, including treatment, survival status, and cause of death. Matching for treatment was not possible because the SEER database does not capture information regarding pharmacological therapy and captures only limited information concerning radiotherapy. For both cHL and NHL, staging was captured using SEER summary staging, in which “localized” corresponds to Ann Arbor stage I; “regional” corresponds to Ann Arbor stages II and IIE; and “distant” corresponds to stages IIS, III, and IV. To compare survival outcomes, dnNHL and sNHL were grouped in 3 biological categories: T-cell NHL, indolent B-cell NHL, and aggressive B-cell NHL.
Comparisons between proportions were performed using the chi-square test. We described continuous numerical variables on the basis of the median and range and compared continuous variables between 2 groups using the Mann-Whitney U nonparametric test. We calculated the cumulative incidence of sNHL among HL survivors using a competing risk model, treating death from any cause in the absence of sNHL as the sole competing risk. Differences in incidence across strata were compared with the Gray test. The analysis was performed with the R v2.13.2 package using the “cmprsk” package. To compare overall survival between patients with sNHL and dnNHL, we used the Cox proportional hazards model with a normal frailty term to account for the 2:1 matched design. The model was adjusted for age, sex, race, year of diagnosis, stage, and radiation treatment. Hazards ratio (HR) estimates and 95% confidence intervals (95% CIs) were calculated for previous HL versus de novo conditions for each of the 3 biological subcategories. Only HRs for the overall period under study were constructed, disregarding potential changes in risk over time within subgroups. The modeling was performed with R statistical software using the “survival” package. In all inference analyses, 2-sided P values < .05 were considered to indicate statistical significance.
Incidence of sNHL
A total of 26,826 cHL cases were included in the analysis with a median follow-up of 62 months (range, 3 months-215 months), yielding a total follow-up of 163,979 patient-years. The median age of patients was 34 years (range, 2 years-99 years); 8324 patients (31.0%) had advanced stage disease, 14,619 (54.5%) had limited stage disease, and the stage of disease was unknown for 3883 patients (14.5%). The characteristics of the entire “at-risk” population and of patients with HL who subsequently did or did not develop sNHL are shown in Table 1.
Table 1. Characteristics of Patients in the HL Cohort
Radiation use in localized and regional stage disease
Radiation use in advanced stage disease
Overall, 259 cases of sNHL were diagnosed. The median interval between the diagnosis of HL and sNHL was 48 months (range, 3 months-202 months). Patients who later developed sNHL were older, more likely to be male, and more likely to have an advanced stage of disease at the time of diagnosis of cHL than patients who did not develop sNHL. There were no differences noted between patients with cHL who later developed sNHL and those who did not in terms of race and the use of radiotherapy for the management of cHL (Table 1).
The median age at the time of diagnosis of sNHL was 56 years (range, 10 years-92 years). Thirty-eight cases (14.7%) of sNHL were T-cell lymphomas, 144 (55.6%) were aggressive B-cell NHL, 64 (24.7%) were indolent B-cell NHL, and 13 (5%) were not classified further at the time of registration.
The cumulative incidence of sNHL was 0.41% (95% CI, 0.32%-0.50%), 1.34% (95% CI, 1.41%-1.55%), and 2.50% (95% CI, 2.10%-2.89%), respectively, at 5 years, 10 years, and 15 years after the diagnosis of cHL. The incidence of sNHL was found to be closely linked to the patient's age at the time of diagnosis of HL. The cumulative incidence at 15 years after cHL diagnosis was 0.67% (95% CI, 0.04%-1.29%) for age < 20 years, 2.23% (95% CI, 1.77%-2.67%) for age 20 years to 59 years, and 5.59% (95% CI, 4.21%-6.97%) for age ≥ 60 (P < .001) (Fig. 1).
The SIR of NHL for the entire cohort was 7.5 (95% CI, 6.6-8.5), being higher for T-cell lymphomas and for aggressive B-cell lymphomas and lower for indolent B-cell lymphomas (Table 2). The SIR for all NHL cases was also influenced by patient age at the time of diagnosis of HL, and was 16.2 (95% CI, 7.0-32.0) for patients aged < 20 years, 8.8 (95% CI, 7.5-10.3) for patients aged 20 years to 59 years, and 5.8 (95% CI, 4.7-7.1) for patients aged ≥ 60 years
Table 2. Standardized Incidence Ratio for Different Histological Groups of sNHL in Survivors of HL
Characteristics of patients with sNHL and matching dnNHL are shown in Table 3. Patients with indolent B-cell sNHL demonstrated a higher risk than their dnNHL counterparts (HR of death, 2.7; 95% CI, 1.3-5.7 [P = .008]). There was no difference noted in the risk of aggressive B-cell sNHL (HR, 1.3; 95% CI, 0.6-2.7 [P = .19]) or T-cell sNHL (HR, 0.8; 95% CI, 0.3-1.8 [P = .42]) when compared with their de novo counterparts for the overall period under study. Risk was variable over time in the subgroups of aggressive B-cell and T-cell lymphoma, as evidenced by the crossing of survival curves for these 2 groups (Fig. 2). In all the histological categories, NHL was the most frequent cause of death for patients with both sNHL and dnNHL. However, death from cHL remained an important component of overall mortality among patients with sNHL, particularly those with indolent B-cell NHL (Table 4).
Table 3. Characteristics of Patients With sNHL and Matched Patients With dnNHL
Abbreviations: +, positive; CLL/SLL, chronic lymphocytic leukemia/small lymphocytic lymphoma; dnNHL, de novo non-Hodgkin lymphoma; ICD-O-3, International Classification of Diseases for Oncology, third edition; NHL, non-Hodgkin lymphoma; NOS, not otherwise specified; sNHL, secondary non-Hodgkin lymphoma.
Number of primary cancers registered in the Surveillance, Epidemiology, and End Results (SEER) program for individuals in each of the cohorts.
“Localized” corresponds to Ann Arbor stage I; “regional” corresponds to Ann Arbor stages II and IIE; and “distant” corresponds to stages IIS, III, and IV.
To our knowledge, this is the largest population-based analysis (163,979 patient-years of follow-up) to investigate not only the incidence but also outcomes of sNHL among survivors of HL.
Other studies have described the risk of NHL among survivors of cHL. In a large British study involving 5798 patients who were treated for HL between 1963 and 2001, 82 cases of sNHL were observed. In this study, the SIR was calculated at 11.5 for patients treated with chemotherapy only and 17.1 for patients also treated with radiotherapy. The risks encountered in that study were somewhat higher than reported in the current study, perhaps reflecting its longer follow-up and the effect of older and possibly more toxic therapeutic strategies. Another larger study combined data on 35,591 patients with HL from 16 population-based registries in North American and Europe and found 162 cases of sNHL. It corresponded to 5.5 times the number of expected cases from a non-HL population, an increase in risk similar to that reported herein. Yet another study with 1380 patients aged < 17 years who were treated for HL and reported by the Late Effects Study Group found a SIR for NHL of 11.7, which is comparable to our finding for the strata of similar age (SIR of 16.2 for patients aged < 20 years).
To the best of our knowledge, the current study is the first to detail the incidence of different subtypes of NHL, which is a very large category encompassing entities with very distinct prognoses. It is of interest that an increased risk was noted for patients with both indolent and aggressive B-cell histologies and also for those with T-cell lymphomas. The latter group had the highest relative increase in incidence risk. The mechanisms for an increased incidence risk are beyond the scope of this article and are likely to include late effects of therapies as well as an underlying predisposition toward lymphoid malignancies that may or may not be tangible (eg, linked to infection with the human immunodeficiency virus [HIV]). However, it does not appear that the increased incidence of NHL after a diagnosis of cHL is fully explained by a genetic and/or environmental predisposition toward lymphoid malignancies. If that were the case, cHL would be expected to be frequent after the diagnosis and treatment of NHL. Using inclusion criteria that were equivalent to the ones described herein and focusing on the most frequent histology of NHL, diffuse large B-cell lymphoma (DLBCL), we identified in an exploratory analysis 63,884 “at-risk” patients with DLBCL as their first malignant neoplasm but only 67 cases of cHL were found to occur after a median of 54 months from the time of the DLBCL diagnosis. This is in sharp contrast with the “reverse” numbers described in the current study results: 26,826 “at-risk” cases of cHL and 259 cases of sNHL (128 cases of DLBCL and its variants). This exploratory analysis suggests that the increased incidence of DLBCL after a diagnosis of cHL is only partially explained by a genetic or environmental predisposition toward both diseases.
Not surprisingly, the absolute incidence of sNHL increased with follow-up and was found to be higher among older patients, possibly because most NHLs have an increasing incidence with age. Conversely, younger age was linked to the highest SIRs, reflecting the low expected incidence of NHL in the general population of that age. We observed no difference in the risk of sNHL according to sex.
One important limitation of the current study is the lack of a centralized pathology review, particularly because the diagnostic workup of lymphomas is complex, specialized, and constantly evolving. For example, CD30-positive anaplastic T-cell lymphoma and primary mediastinal B-cell lymphoma are occasionally mistaken for cHL. It is possible that some cases of sNHL are indeed recurrences of the same condition previously misdiagnosed as cHL. Without a pathology review, it is impossible to quantify the impact of misdiagnosis in the results of the current study. However, we believe that if patients with recurrent misdiagnosed NHL accounted for the majority of the cases of apparent sNHL occurring after cHL, we would observe a pattern of “recurrence” that is much closer to the initial “misdiagnosis” of cHL than the median of 48 months observed in the current study. We would also expect that misdiagnosed patients, after being partially treated with therapy for cHL and developing a disease recurrence as an apparent sNHL, would have a much worse prognosis than the dnNHL matches, something that was not noted in the current study except for patients with indolent B-cell histologies.
We believe one of the most important contributions of the current study is that it provides for the first time (to the best of our knowledge) a comparison between the outcomes of patients with sNHL and matched patients with dnNHL. We suspected that sNHL developing in survivors of cHL would have worse prognosis than dnNHL diagnosed in individuals matched for histological subtype, stage, calendar year of diagnosis, age, race, sex, and SEER registry but without a history of prior cancer. The detrimental effect of prior treatment of HL has been previously verified for patients who developed lung cancer as well as breast cancer. Many factors in patients who survive cHL may affect the ability to successfully treat a second cancer, including a worse tolerance to chemotherapy drugs (eg, limitation in doses of anthracyclines), limitations on the dose and field of radiotherapy, and impaired physiologic reserve (eg, suboptimal pulmonary function after bleomycin and/or radiotherapy). More intriguing and difficult to demonstrate is the possibility that secondary cancers have an intrinsically worse biology that is not captured by conventional staging and prognostic systems.
Because there was great heterogeneity in the histologies of sNHL found in the current study, we determined that it would be more appropriate to examine the outcome by cohorts of patients with NHL of a similar natural history (indolent B-cell, aggressive B-cell, and T-cell lymphomas). We were surprised by the lack of a difference in survival among patients with aggressive B-cell lymphomas and T-cell lymphomas with regard to a prior diagnosis of cHL. For the most part, patients with these histologies are treated with combined chemotherapy or chemoimmunotherapy using many drugs belonging to the same classes as those used in the management of cHL and with the potential for cumulative toxicity.[18, 19] In many instances, radiation is also used. Possible explanations are the already unfavorable short-term survival in the dnNHL cohorts of aggressive B-cell and T-cell lymphomas and the success of alternative drug combinations avoiding cross-toxicity with HL therapies, making cases of sNHL still curable. Another possibility is that patients who developed dnNHL may have been more likely than those with sNHL to be positive for HIV, which is a known risk factor for aggressive B-cell lymphomas,[22, 23] leading to an excessive risk of fatal infectious complications. Unfortunately, HIV status is not captured in the SEER database. It is interesting to note that there were more deaths from infection noted among patients with dnNHL than among those with sNHL (Table 4).
Patients with B-cell indolent sNHL were found to have a higher mortality than their dnNHL counterparts. The reasons for this difference are unknown. Indolent NHLs are typically characterized by an asymptomatic phase, often allowing for delayed therapy, multiple treatments, and disease recurrence and late mortality.[24-26] The frequency of causes of death in these cohorts (Table 4) suggests that although death from NHL may not have been different between patients with sNHL and those with dnNHL, the ongoing risk of fatal recurrence of cHL was a major contributor to the mortality in the former group, possibly accounting for some of the differences noted in overall survival. However, cause-of-death attribution is very prone to error and should be interpreted cautiously.
The current study has the caveat of lacking detailed information regarding disease stage and treatments, thereby limiting the establishment of linkage between cHL treatments and the development of sNHL. The characteristics of the database also hinder any analysis of the relative success and caveats of different therapeutic modalities that could be used to treat patients with sNHL. However, the results of the current study demonstrated that sNHL is a significant complication among survivors of cHL and relative prognosis depends on histological subtype, among other things. The most common category, aggressive B-cell sNHL, does not appear to have a worse outcome than its dnNHL counterpart and should be treated aggressively with curative intent.
The research presented in this article was supported in part by the Biostatistics Shared Resource as part of the Hollings Cancer Center at the Medical University of South Carolina, which is funded by a Cancer Center Support Grant P30 CA138313.