NMSC was the most common SN (1,104 total episodes; Appendix Table A4, online only) and occurred as SN1 in 485 survivors (Fig 1A), 61 (12.6%) of whom subsequently developed an invasive SMN. Other important patterns were identified. Of 176 participants who had a breast neoplasm as SN1, 37 developed a new breast neoplasm as SN2 (Fig 1B). Multiple subsequent meningiomas were common (Fig 1C). Thyroid cancer, soft-tissue sarcomas, and CNS malignancies were frequently observed as SN1 and preceded a variety of additional SNs (Figs 1D to ​to11F).

Among survivors with SN1, the cumulative incidence of SN2 was 33.4% (95% CI, 30.3% to 36.5%) at 10 years, was 38.8% (95% CI, 35.1% to 42.5%) at 15 years, and was 46.9% (95% CI, 41.6% to 52.2%) at 20 years (Fig 2A). When SN1 was an SMN, the cumulative incidence of developing SMN2 after SMN1 was 12.4% (95% CI, 9.1% to 15.6%) at 15 years (Fig 2A). The cumulative incidence of SN2 was highest among survivors of a primary Hodgkin's lymphoma (50.3% at 15 years; 95% CI, 44.1% to 56.6%) or CNS malignancy (44.5% at 15 years; 95% CI, 33.4% to 55.6%; Figs 2B and ​and2C).2C). Among survivors exposed to radiation as therapy for their primary cancer, the cumulative incidence of SN2 was 41.3% (95% CI, 37.2% to 45.4%) at 15 years from SN1 compared with 25.7% (95% CI, 16.5% to 34.9%) for those not treated with radiation (Appendix Table A5, online only). Similarly, among patients with SMN1 who were not exposed to radiation, the cumulative incidence of SMN2 was 22.8% at 10 years.

Within this cohort, there were 252 breast lesions (n = 189 invasive, n = 61 in situ, and n = 2 benign). The cumulative incidence of developing a second breast SN was 20.7% (95% CI, 14.7% to 26.7%) at 10 years from the time of development of the initial breast SN (Fig 3A). Because of a significant rate of synchronous breast lesions, cumulative incidences of a second breast neoplasm conditioned on time from first breast SN of 6, 12, and 24 months were 12.6% (95% CI, 7.3% to 18.0%), 10.4% (95% CI, 5.2% to 15.7%), and 9.6% (95% CI, 4.4% t0 14.8%), respectively, at 10 years from initial breast SN. The cumulative incidence of developing a second NMSC at 10 years from the first NMSC was 49.0% (95% CI, 43.5% to 54.5%). Because multiple lesions at presentation were common, when conditioned on time from first NMSC (Fig 3B) of 6, 12, and 24 months, the cumulative incidences of a second NMSC were 40.8% (95% CI, 34.7% to 46.9%), 38.2% (95% CI, 32.0% to 44.4%), and 33.8% (95% CI, 27.3% to 40.3%), respectively, at 10 years from first NMSC.

Among survivors who received radiation and developed an SN, 414 had NMSC as SN1, whereas 570 had an invasive SMN as SN1. The cumulative incidence of an additional invasive malignancy (ie, SMN) was 20.3% (95% CI, 13.0% to 27.6%) at 15 years after NMSC compared with 10.7% (95% CI, 7.2% to 14.2%) after SMN1 (Fig 4) and was potentially influenced by the survival estimates for those with an invasive SMN as SN1 (45.4% at 15 years; 95% CI, 37.0% to 53.8%) that were lower than for those with NMSC as SN1 (79.0% at 15 years; 95% CI, 68.4% to 89.6%).

The limitation of incomplete information regarding treatment for SNs and SMNs is recognized. Table 2 provides multivariable models evaluating risk factors (ie, demographic, treatment exposure for primary cancer, health behaviors, family history) associated with the cumulative incidences of multiple SNs and multiple SMNs after SN1 and SMN1, respectively. Exposure to radiation for the primary cancer was associated with cumulative incidence of SN2 (subdistribution HR, 2.16; 95% CI, 1.32 to 3.55). However, radiation appeared protective from SMN2 (subdistribution HR, 0.48; 95% CI, 0.25 to 0.94; cause-specific HR, 0.41; 95% CI, 0.20 to 0.85). Note that the radiation-exposed survivors had a lower survival estimate at 10 years from SMN1 (56.1%; 95% CI, 49.8% to 62.4%) compared with those who had no radiation exposure (64.3%; 95% CI, 49.6% to 79.0%). Additionally, older age at SN1 (≥ 30 years) was associated with higher cumulative incidence of developing an SN2 compared with those younger than 30 years of age (HR, 1.9; 95% CI, 1.28 to 2.82). Women were more likely than men to develop SMN2 (HR, 2.53; 95% CI, 1.23 to 2.51). However, when breast neoplasia as SMN2 was removed from analysis, association with female sex did not reach statistical significance (HR, 1.57; 95% CI, 0.70 to 3.50). Having a first-degree relative with cancer and being a current smoker were associated with an increased cumulative incidence for multiple SMNs, but the associations were not statistically significant at the.05 level.

Certain patterns of SNs raise concern. Of 176 survivors who experienced a breast neoplasm as SN1, 37 (21%) experienced one or more subsequent primary breast neoplasms. In all, 42 women experienced multiple breast neoplasms, with a cumulative incidence of 20.7% at 10 years from diagnosis of the initial lesion. In many of these occurrences, women had bilateral (synchronous) lesions at presentation not thought to be metastatic, or occurrence of a tumor in the contralateral breast shortly after the primary diagnosis. Nonetheless, among those who were tumor free at 1 year from the initial diagnosis, 10% will still develop a subsequent breast neoplasm 10 years from the initial diagnosis. Although increased risk for breast cancer after treatment for childhood cancer has been well documented,¹⁷ this pattern for multiple cancers raises new concerns. Similarly, of 142 participants who had a nonmalignant meningioma, 126 occurred as SN1, and 12 survivors (10%) experienced multiple meningiomas. Therefore, not only is the incidence of meningioma increasing with time from treatment exposure¹⁸ but also, with time, survivors may be at increasing risk for multiple meningiomas as well.

Among radiation-exposed individuals who developed NMSC as SN1 in this analysis, one in five developed an invasive neoplasm within 15 years, an incidence almost twice that of those also exposed to radiation but who had an invasive neoplasm (SMN) as SN1. Therefore, NMSC may represent a clinical marker for early identification of a population at high risk for a future malignant neoplasm. Occurrence of NMSC as a first SN may identify survivors with genetic susceptibility to radiation injury and/or deficient DNA repair. In the general population, there is evidence for an association between a NMSC diagnosis and increased risk for subsequent cancer.^19,20 Future studies regarding genetic susceptibility should include evaluation of polymorphisms in genes associated with DNA repair.²² In addition, a surprising number of patients with SMN1 who were not exposed to radiation developed SMN2 (cumulative incidence, 22.8% at 10 years). This may identify a population of patients with an underlying cancer predisposition syndrome. We note that SMN1 as SN1 may also identify a genetically susceptible population, but, because of the higher rate of mortality after SMN1 (compared with NMSC as SN1), these patients don't have the same opportunity to develop SMN2. Additionally, though many survivors had multiple NMSCs at presentation, conditional cumulative incidence analysis identifies the high risk that remains in these survivors for future additional NMSCs.

Repeated investigations have identified that female sex, young age at primary cancer diagnosis, radiation exposure, family history of cancer, as well as a primary diagnosis of sarcoma or Hodgkin's lymphoma increase risk for SMN.^5,6,23 In this analysis, radiation exposure and older age at SN1 (≥ 30 years) appear to be the most important risk factors for development of multiple SNs. Although the association with radiation is plausible given the established association between radiation and SMN1, we hypothesize that older age at SN1 identifies a population of survivors who are reaching ages at which, in the general population, cancer risks increase. The appearance of a protective effect of radiation for development of SMN2 after SMN1 is likely artifactual as a result of death as a competing risk for SMN2 development. Survivors with SMN1 who had received radiation were more likely to die and, thus, less likely to have SMN2. Furthermore, because both the subdistribution HR (0.48) and cause-specific HR (0.41) for developing SMN2 after SMN1 indicated protective effects of having had radiation for the original childhood cancer, we infer that risk of death and risk of SMN2 were positively correlated, perhaps highly, after SMN1. That is, deaths after SMN1 removed survivors who were at high risk for SMN2 from risk sets disproportionately more in the radiation-exposed subgroup. Had this not been the case, the cause-specific HR would not have been in the protective direction. Although females were at increased risk for SMN2 after SMN1, patients with first-degree relatives with cancer and patients with a smoking history may also be at increased risk. However, these findings were of borderline statistical significance. Additionally, having a sarcoma or a p53 (Li-Fraumeni) –associated SMN1 was not found to be associated with developing multiple malignancies.

Limitations of this study should be considered. The majority of SNs are initially ascertained by self report, which may result in underestimation of true incidence rates. Additionally, although CCSS has collected detailed chemotherapy exposures and radiation dosimetry and volume measures relating to the primary cancer, only limited information was available regarding treatment of SNs, restricting risk factor analyses to primary cancer exposures only. Clearly, exposures resulting from treatment of subsequent neoplasms could impact subsequent cancer risk. The limitation of not having the SN treatment information to include in the risk factor analysis does not detract from the findings and clinical implications of the high cumulative incidence of multiple subsequent neoplasms. Finally, systematic and precise determination of SN occurrence as in field versus out of the radiation field was not possible in many of the occurrences, because exact information regarding location of the SN was not always available either from the respondent or as noted on the pathology report.

Collection and assembly of data: Gregory T. Armstrong, Wei Liu, Wendy Leisenring, Sue Hammond, Smita Bhatia, Joseph P. Neglia, Marilyn Stovall, Deokumar Srivastava, Leslie L. Robison

Data analysis and interpretation: Gregory T. Armstrong, Wei Liu, Wendy Leisenring, Yutaka Yasui, Sue Hammond, Smita Bhatia, Joseph P. Neglia, Marilyn Stovall, Deokumar Srivastava, Leslie L. Robison

Manuscript writing: All authors