Indolent NHLs are incurable, slow-growing lymphomas that cause few symptoms in early stages and feature a relapsing-remitting course.^1–2 Treatment is often deferred until disease progression occurs, when an oral alkylating agent such as chlorambucil is the standard primary treatment.³⁶ Patients with relapsed or advanced disease receive various combinations of chemotherapy, radiotherapy, immunotherapy, radioimmunotherapy, or high-dose chemoradiotherapy with stem-cell transplantation.^36–37 Patients with indolent NHL, particularly CLL, have well-characterized long-term immune dysfunction because of prediagnostic immune alterations, a relapsing-remitting disease course, and repeated exposure to immunosuppressive therapies over time.^38–39 In contrast, aggressive NHLs such as DLBCL are rapidly growing, and 30% to 50% of patients are cured with intensive initial chemotherapy with or without radiotherapy.^1–2 The standard chemotherapy regimen for DLBCL includes cyclophosphamide, doxorubicin, vincristine, and prednisone (CHOP), combined with rituximab (CHOP-R) in recent years.⁴⁰ Patients refractory to initial therapy receive additional chemotherapy and high-dose chemoradiotherapy with stem-cell transplantation, though prognosis is generally poor.⁴⁰

We observed excesses of lung cancer and melanoma after the more indolent lymphomas, CLL/SLL and FL, but not after the more aggressive DLBCL. Numerous studies of NHL and CLL have demonstrated associations for lung cancer^{10,13,14,17,18,22,25,27–29} and melanoma,^{11–13,15,22,29,30} but the mechanisms are unclear. Immunological alterations associated with lymphomas are likely to contribute, because lung cancer and melanoma occur in excess in immunosuppressed patients,^41–43 and inflammation and infections increase lung cancer risk.⁴⁴ However, it is noteworthy that these tumors were not associated with DLBCL, the subtype most strongly linked with immune dysfunction.⁶ Because the excesses of lung cancer and melanoma were limited to more indolent lymphomas, it is plausible that long-term immune dysfunction from underlying disease or repeated treatments is responsible. However, the excess risks immediately after CLL/SLL and FL diagnosis suggest an intrinsic mechanism, possibly involving prediagnostic immune alterations or defects in cell cycle and apoptosis, which are more characteristic of indolent than aggressive lymphomas.⁴⁵

Among Hodgkin's lymphoma survivors, epidemiologic studies have suggested that lung cancer risk is increased by radiotherapy and chemotherapy, alone or in combination with cigarette smoking.^46–47 These exposures may also contribute to excess lung cancer risk seen in some studies of NHL and CLL survivors.^{10,17,18,27,48} Although FL and CLL, but not DLBCL, may be weakly related to cigarette smoking,^49,50 the lack of association with small-cell carcinomas of the lung or other strongly tobacco-related cancers argue against smoking as a primary explanation for our findings. Additional research is warranted to confirm that excess lung cancer risks are confined to CLL/SLL and FL survivors and to evaluate the effects of treatment and shared environmental and genetic risk factors. Studies that include longer-term survivors will be necessary to consider effectively the role of radiotherapy.

In this study and others, the association between NHL and melanoma occurred at all latency periods and was not clearly related to initial treatment, which suggests the influence of shared risk factors. The main environmental risk factor for melanoma, sun exposure, appears to be protective for NHL across the major subtypes.⁵¹ We therefore hypothesize that underlying immunologic and genetic risk factors contribute to the association between melanoma, CLL/SLL, and FL.⁵² Although the association between NHL and melanoma may be influenced by heightened medical surveillance, the excess risks seen for melanomas ≥ 1 mm thick but not for early lesions reduces the likelihood of this explanation.

We observed elevated Hodgkin's lymphoma risks among all NHL survivors, which is consistent with previous studies.^11,13,17,22 Although immune dysfunction and shared risk factors may contribute to multiple, distinct lymphoid neoplasms,^42,53–54 diagnosis of Hodgkin's lymphoma after NHL may be complicated by overlapping morphologies that lead to diagnostic misclassification between recurrence and a second malignancy, under-reporting of a second lymphoma subtype, or histologic transformation.⁵⁵ CLL/SLL propensity for transformation (ie, Richter's syndrome)⁵⁶ may contribute to the higher risks seen after CLL than other subtypes. Additional research with serial clinical evaluation, detailed pathologic review, and evaluation of tumor clonality to identify transformation is necessary to clarify whether Hodgkin's lymphoma risks vary by NHL subtype.

The increased risk for acute nonlymphocytic leukemia after DLBCL and FL was primarily seen among patients receiving initial chemotherapy, especially in younger patients. This observation is consistent with previous studies that have implicated DNA-damaging agents, such as alkylating agents and topoisomerase II inhibitors, as well as nucleoside analogs in some studies.^{10,16–18,21,22,26} Acute nonlymphocytic leukemia risks after DLBCL and FL have not been compared previously. Although DLBCL patients receive more aggressive initial chemotherapy, our data raise the hypothesis that FL and DLBCL survivors may eventually receive similar amounts of chemotherapy because of the high risk of FL relapse and multiple courses of chemotherapy.² The increased risk after SLL but not CLL may be due to more frequent use of chemotherapy for SLL than CLL (44% v 25%).

Second malignancy risks were strikingly different in patients with HIV/AIDS–related lymphoma compared with patients without HIV/AIDS. The substantially elevated Kaposi's sarcoma risk after HIV/AIDS–related DLBCL is most likely due to the etiologic role of HIV in both malignancies,⁵⁷ whereas the excesses of buccal cavity/pharynx and anal cancers are likely related to coinfection with HIV and human papillomavirus.⁵⁸

The primary strength of our study was analysis of a large number of lymphoma survivors identified via population-based registry data, which eliminated the selection bias seen in hospital-based series. Additional strengths include the quantitative assessment of second malignancy risks by lymphoma subtype, distinction of patients with and without HIV/AIDS–related lymphoma, and the first comparison of malignancy risks after CLL and SLL. The similar patterns of second malignancy after CLL and SLL reinforces the notion that these diseases are different clinical manifestations of the same entity.¹ We limited our analysis to patients diagnosed with lymphoma in 1992 or later to utilize the available HIV/AIDS data and to minimize lymphoma subtype misclassification.³⁴ Although this restriction precluded comprehensive assessment of second malignancy risks among long-term lymphoma survivors, elevated risks were observed within the first 5 years after lymphoma diagnosis. Additional limitations included under-ascertainment of second malignancy risks as a result of patient migration outside SEER program areas and incomplete data on the initial course of chemotherapy, which may have attenuated our results.

Collection and assembly of data: Lynn A.G. Ries

Data analysis and interpretation: Lindsay M. Morton, Rochelle E. Curtis, Martha S. Linet, Elizabeth C. Bluhm, Margaret A. Tucker, Neil Caporaso, Lynn A.G. Ries, Joseph F. Fraumeni Jr

Manuscript writing: All authors