We thank Laura Brumbaugh for editorial assistance.
Lung cancer accounts for the largest absolute risk of second malignancies among Hodgkin lymphoma (HL) survivors. However, no population-based studies have compared overall survival (OS) between HL survivors who developed nonsmall cell lung cancer (HL-NSCLC) versus patients with first primary NSCLC (NSCLC-1).
The authors compared the OS of 178,431 patients who had NSCLC-1 and 187 patients who had HL-NSCLC (among 22,648 HL survivors), accounting for sex, race, sociodemographic status, calendar year, and age at NSCLC diagnosis, and NSCLC histology and stage. All patients were reported to the population-based Surveillance, Epidemiology, and End Results Program. Hazard ratios (HRs) were derived from a Cox proportional hazards model.
Although the NSCLC stage distribution was similar in both groups (20% localized, 30% regional, and 50% distant), HL survivors experienced significantly inferior stage-specific OS. For patients with localized, regional, and distant stage NSCLC, the HRs (95% confidence interval [CI]) for death among HL survivors were 1.60 (95% CI, 1.08-2.37; P < .0001), 1.67 (95% CI, 1.26-2.22; P = .0004), and 1.31 (95% CI, 1.06-1.61; P = .013), respectively. Among HL-NSCLC patients, significant associations were observed between more advanced NSCLC stage and the following variables: younger age at HL diagnosis (P = .003), younger age at NSCLC diagnosis (P = .048), and longer latency between HL and NSCLC diagnoses (P = .015).
Over the past half century, successful treatment regimens for Hodgkin lymphoma (HL) have rendered it a largely curable disease. Many studies have quantified the late effects of successful HL treatment, including subsequent malignancies,1-5 clearly documenting that lung cancer accounts for the largest absolute risk.1 Despite the substantially increased 3-fold to 20-fold relative risks of lung cancer after HL,1, 3-9 few reports have examined subsequent survival.6, 10, 11 These studies typically included <25 patients and did not account for the effects of important demographic and tumor-related parameters. In 1 study, actuarial survival analyses included patients with differing histologies, including nonsmall cell lung cancer (NSCLC), small cell lung cancer, and mesothelioma, and did not control for disease stage.10 In another study, neither lung cancer histology nor stage was specified.6
Established risk factors for lung cancer after HL include radiotherapy8, 12-14 and alkylating-agent chemotherapy8, 14, 15 (both produce a highly significant dose response) as well as smoking.8, 13, 14, 16 The largest study to date demonstrated that virtually all patients with HL who developed lung cancer were tobacco users.8 On the basis of that international analytic study, Travis et al estimated that, of the lung cancers that develop after HL, about 63% are caused by the combined effect of treatment and smoking, 24% are caused by smoking, 10% are caused by treatment alone (including radiation therapy and/or chemotherapy), and the remaining 3% are from causes unrelated to tobacco use or therapy.8
Within a cohort of 22,648 patients who had a first primary HL reported to the population-based cancer registries of the Surveillance, Epidemiology, and End Results (SEER) Program (http://seer.cancer.gov/), we identified all patients who developed NSCLC as a second primary cancer (the HL-NSCLC group). We hypothesized that the overall survival (OS) of patients in this HL-NSCLC group would be markedly inferior to the OS of patients who had a first primary cancer diagnosis of NSCLC (the NSCLC-1 group), even when we controlled the analyses for demographic, clinicopathologic, and treatment-associated variables.
MATERIALS AND METHODS
From the US population-based SEER 13 (1973-2006) database, patients were identified who developed NSCLC as a first primary cancer after HL diagnosis (HL-NSCLC group). A minimum latency of 2 months was required, which is the standard latency adopted by the SEER Program to exclude synchronous primary cancers.17 Because we made no a priori assumptions about NSCLC etiology (ie, tobacco use, HL therapy, host susceptibility), we did not otherwise require a specific latency period between HL and NSCLC.
Of 22,648 patients who were registered with a first primary HL in the SEER 13 data, 238 patients developed a second primary NSCLC. Patients were grouped into localized stage, regional stage, or distant stage NSCLC, as described in the SEER staging manuals (http://seer.cancer.gov/tools/codingmanuals/historical.html). Because SEER did not record lung cancer stage before 1988, analyses grouped by stage are restricted to those reported from 1988 forward. One-hundred eighty-seven of 238 patients with HL-NSCLC were assigned an NSCLC stage. For the NSCLC-1 comparison group, 278,081 patients with a first primary NSCLC were identified from the same SEER registries, and stage was designated for 178,431 of those patients. The following histologic types of NSCLC were included: squamous cell carcinoma, adenocarcinoma, bronchioloalveolar carcinoma, adenosquamous carcinoma, large cell carcinoma, and NSCLC. Patients with neuroendocrine, carcinoid, small cell carcinoma, oat cell carcinoma, sarcoma, and mesothelioma histologies were excluded, because the etiologies, natural history, treatment, and/or outcome of these histologic subtypes of cancer differ from those of NSCLC. Patients with unspecified lung neoplasms also were excluded.
The SEER database records sociodemographic parameters of the population residing in each patient's county of residence, as determined from US Census data. For each HL-NSCLC and NSCLC-1 study patient, the proportion of adults residing within the patient's county who were aged ≥25 years and had less than a high-school education was used as a surrogate for sociodemographic status, as described previously.18
Chi-square tests were used to compare frequency distributions of variables between patient subgroups. Actuarial OS was calculated using the Kaplan-Meier method. Survival was measured from the date of NSCLC diagnosis until date of death or last follow-up (through December 31, 2006). To estimate age-adjusted OS for the NSCLC-1 group, the HL-NSCLC and NSCLC-1 groups were each divided into age subgroups (<30 years, 31-35 years…81-85 years, >85 years). For each NSCLC-1 subgroup, a Kaplan-Meier survival table was obtained. The age-adjusted Kaplan-Meier OS represents the weighted average of the subgroups' survival, with weights based on the age distribution of the HL-NSCLC group (Fig. 1). The log-rank test and Cox regression analyses (controlling for covariates) were used to compare OS between the HL-NSCLC group and the NSCLC-1 group. All survival analyses were conducted using SAS software (version 9.1.3; SAS Institute Inc., Cary, NC). Kaplan-Meier curves were prepared using R statistical software (version 2.7.0; R Foundation for Statistical Computing, Vienna, Austria). All P values are 2-sided, and P < .05 indicates statistical significance.
Patient and Tumor Characteristics
Two-hundred thirty-eight HL survivors developed NSCLC at a median of 9.5 years (range, 0.2-30.7 years) after HL. Table 1 outlines the demographic and clinicopathologic characteristics of these patients at the time of HL diagnosis, including 187 patients who could be grouped according to NSCLC stage. Age at HL diagnosis differed significantly between the 3 stage groups (median 58 years, 51 years, and 42 years for localized, regional, and distant NSCLC, respectively; P = .003). Among the 3 stage groups, the distributions of race (P = .59), sex (P = .49), HL stage, and the presence of B symptoms were similar; whereas the distribution of HL histologic subtypes was significantly different (P = .005); the patients who developed localized NSCLC were more likely to have had mixed cellularity HL and were less likely to have had nodular sclerosis HL. Approximately 50% of the HL patients who developed localized NSCLC received radiation for HL, whereas 61% of patients who developed regional or distant NSCLC received radiation (P = .13).
Table 1. Patient and Tumor Characteristics at the Time of Hodgkin Lymphoma Diagnosis Among 238 Patients Who Developed Nonsmall Cell Lung Cancer
Abbreviations: NSCLC, nonsmall cell lung cancer; NOS, not otherwise specified.
P value was calculated using the chi-square test, and patients with unknown values were omitted from the calculation.
Year of diagnosis
Nodular lymphocyte predominant
Table 2 outlines demographic and clinicopathologic characteristics of patients in the HL-NSCLC and NSCLC-1 groups at the time of NSCLC diagnosis. Although the stage distribution of the HL-NSCLC population mirrored that of the NSCLC-1 population, for each designated stage group, patients in the HL-NSCLC group were significantly younger than patients in the NSCLC-1 group. For regional and distant NSCLC, the HL-NSCLC group tended to have a higher sociodemographic status than the NSCLC-1 group. For localized and regional NSCLC, the HL-NSCLC group had a greater percentage of patients with squamous cell carcinoma histologies.
Table 2. Patient and Tumor Characteristics at the Time of Nonsmall Cell Lung Cancer Diagnosis
Table 2 also provides the latency periods between diagnoses of HL and NSCLC. The time to NSCLC diagnosis differed significantly between the 3 stage groups (P = .015), with median latencies of 6.9 years, 10.8 years, and 14.4 years, respectively. Thirty-six percent, 52%, and 66% of patients with localized, regional, and distant NSCLC, respectively were diagnosed ≥10 years after HL diagnosis. Even after omitting patients who were diagnosed with HL before 1988 (ie, when lung cancer stage was not recorded by the SEER Program), differences in latency remained significant (P = .017), with median latencies of 4.2 years, 6.4 years, and 7.8 years for patients with localized, regional, and distant NSCLC, respectively.
For patients who received radiation for HL, the latency from HL to NSCLC diagnosis exceeded that of patients who did not receive radiation. Differences in latency between the radiotherapy group and no radiotherapy group were significant for regional NSCLC (median latency, 15.3 years vs 5.4 years; P = .007) and distant NSCLC (median latency, 15.7 years vs 8.7 years; P = .044), but not for localized NSCLC (median latency, 9.0 years vs 6.0 years; P = .35).
Among patients with HL-NSCLC, age at NSCLC diagnosis differed significantly between the 3 stage groups (median age: 64 years, 60 years, and 58 years for patients with localized, regional, and distant NSCLC, respectively; P = .048). Although a similar trend was noted among patients in the NSCLC-1 group (median age: 70 years, 68 years, and 67 years, respectively; P < .0001), the median age discrepancy was less between stage groups.
Survival Analysis of HL-NSCLC Versus NSCLC-1
Table 3 outlines the OS of patients with HL-NSCLC and NSCLC-1 according to NSCLC stage. It is noteworthy that the OS for each HL-NSCLC group was inferior to that of the corresponding NSCLC-1 group, and the differences were significant for regional stage (P = .001) and distant stage (P = .040) disease.
Table 3. Comparison of Actuarial Survival Probabilities: Patients With Nonsmall Cell Lung Cancer After Hodgkin Lymphoma Compared With First Primary Nonsmall Cell Lung Cancer
Abbreviations: CI, confidence interval; CSS, cause-specific survival; HL, Hodgkin lymphoma; HL-NSCLC, nonsmall cell lung cancer after Hodgkin lymphoma; HR, hazard ratio; NA, not applicable; NSCLC, nonsmall cell lung cancer; NSCLC-1, first primary nonsmall cell lung cancer; OS, overall survival.
HRs and P values were derived from a Cox proportional hazards model that included history of HL, age at NSCLC diagnosis, calendar year of NSCLC diagnosis, radiotherapy for NSCLC, sociodemographic status at the time of NSCLC diagnosis, sex, race, and NSCLC histology group. Because this model primarily reflected risk factors among the NSCLC-1 population (which exceeded the number of patients with HL-NSCLC by >900-fold), the HRs and P values associated with the other included variables are not shown.
A 2-tailed log-rank test was used for comparison with the HL-NSCLC group.
Because of the marked difference in age at NSCLC diagnosis between the HL-NSCLC and NSCLC-1 groups, the influence of age on stage-specific survival was evaluated in more detail. Among NSCLC-1 patients, median survival decreased significantly with increasing age (P < .0001 for each stage group), a trend that was not observed among HL-NSCLC patients with regional or distant stage NSCLC (Fig. 2). For each NSCLC stage and in most age groups, the inferior survival of the HL-NSCLC group was apparent. The difference in survival between the NSCLC-1 and HL-NSCLC groups was most evident in patients aged <70 years.
Given the age discrepancy between the HL-NSCLC and NSCLC-1 groups, age-adjusted OS was calculated for each NSCLC-1 stage group. For each of the 3 NSCLC stages, the OS of patients in the HL-NSCLC group was significantly inferior (P ≤ .001) to that of the age-adjusted NSCLC-1 group (Fig. 1 and Table 3). Cox regression analyses that included age at NSCLC diagnosis, history of HL, and other variables consistently demonstrated a significantly worse OS for HL-NSCLC patients versus NSCLC-1 patients, with a hazard ratio and 95% confidence interval (CI) of 1.60 (95% CI, 1.08-2.37) for localized NSCLC, 1.67 (95% CI, 1.26-2.22) for regional NSCLC, and 1.31 (95% CI, 1.06-1.61) for distant NSCLC (Table 3).
Cause of Death
Table 4 summarizes vital status at last follow-up and cause of death for the HL-NSCLC and NSCLC-1 groups. For patients with localized NSCLC, deaths from heart disease were twice as common in the HL-NSCLC group versus the NSCLC-1 group (20% vs 11%, respectively; P = .15); cardiac deaths occurred in patients who were diagnosed at ages 61 to 85 years with localized NSCLC and, thus, do not account for the relatively poorer survival of patients with HL-NSCLC aged ≤59 years (Fig. 2). Sixteen percent of patients in the HL-NSCLC group who had localized NSCLC died from other cancers, compared with 5% of patients in the NSCLC-1 group (P = .015). For regional and distant NSCLC, the distribution of major causes of death was comparable, with 81% to 85% and 85% to 88% of patients dying from NSCLC, respectively. It is noteworthy that, overall, only 4% of all patients with HL-NSCLC died of HL.
Table 4. Vital Status and Cause of Death Among Patients With Nonsmall Cell Lung Cancer After Hodgkin Lymphoma and First Primary Nonsmall Cell Lung Cancer
No. of Patients (%)
Abbreviations: HL, Hodgkin lymphoma; HL-NSCLC, nonsmall cell lung cancer after Hodgkin lymphoma; NSCLC, nonsmall cell lung cancer; NSCLC-1, first primary nonsmall cell lung cancer.
Percentage estimates for cause of mortality are based on the total number of deaths.
An important new finding in our study, based on 187 HL-NSCLC patients and 178,431 NSCLC-1 patients, was the significantly inferior OS among HL survivors who developed NSCLC compared with patients who had de novo NSCLC. We are not aware of other investigations that compare the survival of these 2 groups, simultaneously accounting for demographic, clinicopathologic, and treatment-associated variables. Moreover, our study was conducted within the large, population-based registries that comprise the US SEER Program. By using similar analytic methods, we previously demonstrated that breast cancer after HL also is characterized by worse survival outcomes versus patients with de novo breast cancer.19
Possible explanations for the inferior OS of HL-NSCLC patients include host factors inherent to the development of HL, more aggressive tumor biology of therapy-associated cancers, and the limited treatment options for NSCLC after HL. Other factors that contribute to the inferior survival of patients with localized HL-NSCLC include increased deaths from heart disease and other cancers (Table 4), causes of mortality that are known to increase among HL survivors,2, 20, 21 although our study is the first to explicitly demonstrate this in a comparison against patients with de novo NSCLC.
NSCLC developing in patients with prior HL may have a different molecular biology than de novo NSCLC by virtue of differences in inherent tumor biology and/or host genetic susceptibility factors. A history of prior radiotherapy may introduce genetic alterations that affect not only risk of NSCLC induction but also biologic behavior of the NSCLC. In an analytic population-based study of lung cancer after radiation for HL,22 archived paraffin-embedded tissues were evaluated from 15 patients who developed NSCLC and were compared with tissues from patients who had de novo NSCLC. In that study, NSCLC after HL was characterized by a 5.9-fold increase (P = .0002) in microsatellite alterations.22 Of 20 microsatellite alterations, at least 1 was present in >74% of NSCLCs. The authors of that report concluded that NSCLC developing in irradiated HL patients demonstrates widespread genomic instability, as manifested by increased numbers of microsatellite alterations, occurring by a yet unknown mechanism that warrants further research.
Patient Age and Latency Between HL and NSCLC
To our knowledge, this is the first study demonstrating that patients with HL who are diagnosed with NSCLC at an earlier stage (vs a more advanced stage) tend to be significantly older at the time of HL and NSCLC diagnoses and tend to develop NSCLC after a significantly shorter latency. The decreased latency among older patients with HL may reflect in part the strong influence of tobacco in these second primary lung tumors.8 Thus, as hypothesized previously,8 a greater lifetime of tobacco use may have resulted in a near-critical number of mutations superimposed on the carcinogenic effects of treatment. The shorter latency period in older patients also simply may reflect the relatively shorter remaining lifespan of these patients. However, the discrepancy in median age at HL diagnosis between NSCLC stage groups exceeds the discrepancy in median latency, accounting for the association of older age at NSCLC diagnosis with earlier stage NSCLC. Several explanations may underlie this latter observation. For example, older patients may be more likely to undergo diagnostic workups for pertinent symptoms than younger patients, resulting in the diagnosis of NSCLC at an earlier stage. Possible age-related differences in tumor biology also may be relevant.
Patients who received radiation for HL experienced a longer latency between HL and NSCLC diagnoses versus patients with HL who did not receive radiation. Although the lack of chemotherapy information in the SEER database precludes a detailed analysis of the impact of chemotherapy on latency, it is conceivable that a greater percentage of patients (perhaps most) who did not receive radiation for HL received chemotherapy with alkylating agents. The shorter latency observed in the “no radiotherapy” group is consistent with the analytic study by Travis et al, who observed that the lung cancer risk already was increased in the interval 1 to 4 years after chemotherapy but not until 5 to 9 years after radiotherapy.8 Those investigators also reported a similar relative risk of stage I/II lung cancer versus stage III/IV lung cancer subsequent to the receipt of alkylating agent chemotherapy (risks of 3.2 [95% CI, 1.5-7.3] vs 4.7 [95% CI, 2.3-10.2], respectively). It is noteworthy that the patients with HL-NSCLC in our series who developed localized NSCLC were more likely to have had mixed cellularity HL (Table 1), a subtype for which patients historically were more apt to have received chemotherapy.23, 24 This also may account for the shorter latency time in patients who developed localized NSCLC. However, we observed no apparent association between HL subtype and the receipt of radiation (P = .84, chi-square test; data not shown).
The NSCLC stage distribution of HL survivors mirrors that of the general population, perhaps reflecting the common etiology of tobacco in most lung cancers, even among HL survivors, for whom Travis et al8 estimated that approximately 87% are caused by smoking (similar to what is reported for the general population25) either alone or in combination with treatment. Compared with patients who have de novo NSCLC, HL survivors who have similarly staged NSCLC are relatively younger at the time of NSCLC diagnosis and are diagnosed with NSCLC in relatively later decades, in part reflecting the typical latency period of ≥10 years required for most radiation-associated solid tumors. Previous studies have documented a higher sociodemographic status among younger patients with HL who have a nodular sclerosis subtype,26, 27 whereas older age and a mixed cellularity subtype have been correlated with lower sociodemographic status,26, 28, 29 although recent data are not consistent with the latter findings.27 We observed a correlation of localized NSCLC with older age, mixed cellularity HL subtype, and poorer sociodemographic status (vs regional/distant NSCLC cohorts). Although we observed no correlation between sociodemographic status and HL subtype among our HL-NSCLC cohort, increasing age was associated with worse sociodemographic status (P = .004; data not shown).
Strengths of the current study include the sizable number of patients (178,431 patients with de novo lung cancer and 187 patients with HL-NSCLC) identified in a large population-based setting. Substantial patient numbers allowed for analyses of outcomes according to NSCLC stage, age, and other patient-related and tumor-related variables. Known limitations of SEER data include a lack of detailed information about radiotherapy and an absence of chemotherapy data and tobacco history, factors that have a known association with NSCLC development and that may affect outcomes.8, 12, 14-16, 30 In addition, radiotherapy is under-reported in the SEER Program,17 and SEER registries do not collect data on known prognostic factors associated with NSCLC survival, such as weight loss and performance status.31-33
Given the mortality and morbidity associated with NSCLC, the adaptation of smoking-cessation programs in HL survivors should be strongly encouraged. Furthermore, future studies should investigate the possible role of lung cancer screening in selected HL survivors.34-36 Notably, the National Lung Screening Trial randomized >53,000 high-risk individuals from the general US population (individuals ages 55 to 74 years who were former smokers [within 15 years since cessation] or current smokers with a ≥30 pack-year smoking history) to screening with either chest x-ray or low-dose computerized tomography (CT).37, 38 Preliminary findings suggest that CT screening may reduce lung cancer deaths by 20%.39 These compelling data naturally raise the question of whether similar CT screening would benefit selected HL survivors, who can experience an increased risk of NSCLC up to 20-fold1, 3-9; moreover, we now know that these HL survivors have significantly increased 1.3-fold to 1.7-fold risk of death after an NSCLC diagnosis. Any benefit of CT screening should be weighed against potential risks of radiation exposure from CT imaging,40 particularly for those patients who are scanned at a younger age. To our knowledge, currently, there is no evidence-based research to support routine thoracic CT screening in HL survivors. To fill this gap, Ng and colleagues41 have initiated a prospective study of low-dose chest CT screening in high-risk HL survivors. Future research that addresses the clinical-biologic underpinnings of HL-NSCLC also may provide an opportunity to further understanding of lung cancer carcinogenesis. Any such inroad also could provide an opportunity to develop newer, more effective NSCLC therapy.
This work was supported by the University of Rochester Medical Center (L.B.T.).