Risk of subsequent primary lymphoma in a cohort of 69,460 five‐year survivors of childhood and adolescent cancer in Europe: The PanCareSurFup study

Survivors of Hodgkin lymphoma (HL) are at risk of developing non‐Hodgkin lymphoma (NHL) after treatment; however, the risks of developing subsequent primary lymphomas (SPLs), including HL and NHL, after different types of childhood cancer are unknown. The authors quantified the risk of SPLs using the largest cohort of childhood cancer survivors worldwide.


INTRODUCTION
,000 children and adolescents are diagnosed with cancer in Europe each year. 1 Since the 1970s, 5-year survival rates have improved dramatically and now reach 80% in most European countries, mostly because of combined chemotherapy modalities and improvements in the delivery of radiotherapy. [1][2][3] There are currently over 500,000 childhood cancer survivors in Europe, and this number continues to increase. 1 The risk of premature morbidity and mortality for childhood cancer survivors compared with the general population is well documented, with many health risks arising 20-30 years after a childhood cancer diagnosis. Approximately 60% of childhood cancer survivors develop at least one chronic health condition during their lifetime, and more than one in four develop severe or life-threatening diseases. 4,5 A serious long-term consequence of treatment of childhood cancer is the increased risk of developing subsequent primary neoplasms, which is in the range of three to five times greater that of the general population. [6][7][8][9] Existing evidence suggests that survivors are at increased risk of several types of subsequent primary neoplasms, particularly central nervous system tumors; nonmelanoma skin cancer; and digestive, breast, bone, and genitourinary cancers. Limited numbers of studies have shown that the overall risk of subsequent primary lymphoma (SPL) is also increased 7,[9][10][11] ; however, to our knowledge, few previous large-scale studies have comprehensively investigated the risk of developing SPLs among childhood cancer survivors-with the largest previous study to date including 45 SPLs. 10 Studies of survivors of Hodgkin lymphoma (HL) have found an increased risk of non-Hodgkin lymphoma (NHL) but mostly included HL survivors who were diagnosed in adulthood. [12][13][14][15][16][17][18][19][20][21] To date, no large-scale study has characterized the risks of SPLs and of specific SPLs by type of childhood cancer, sex, age at diagnosis, decade of diagnosis, attained age, and treatment factors. Identification of those survivors at highest risk of an SPL would be important for informing survivors and health care professionals of who is at risk, potentially detecting SPLs early, and may give clues about potential biologic mechanisms.
The principal aim of this largest ever cohort study was to quantify the risk of developing an SPL, further subdivided into HL and NHL, among 69,460 survivors of childhood cancer within Europe and compare this risk with that in the general population, including three times the number of SPLs compared with the largest previous study to date. 10 Secondary aims included investigating variations in risk, which may be associated with certain demographic and oncologic factors, and assessing the level of risk sustained in the long term (beyond age 40 years).

PanCare Childhood and Adolescent Cancer Survivor Care and Follow-Up Studies
The Pan-European Network for Care of Survivors after Childhood and Adolescent Cancer (PanCare) is a network of health care professionals, researchers, childhood cancer survivors, and their families that aims to improve both the care and the quality of life for survivors of childhood cancer. 22 Table S1). Data were sourced from both population-based cancer registries and major treatment centers. Ethical approval and consent for data collection were obtained from the ethical and legal bodies in each of the respective countries contributing to PanCareSurFup.
Ethical approval was not obtained specifically for this study because it involved pooling of nonidentifiable data.

Childhood cancer classification
Because of various practices across different countries in terms of childhood cancer registration, the cancer site and the type of childhood cancer were coded using a range of classification systems. To standardize this across the pooled cohort, all childhood tumor classification codes were converted into codes from the third revision of the

Identification of subsequent primary lymphomas
SPLs were ascertained using various methods, primarily through population-based cancer registries and follow-up clinics and validated through pathology reports or other means of clinical diagnosis.
The SPLs were classified by site according to the International Classification of Diseases using the revision appropriate to the year of diagnosis (see Table S2). For inclusion as an SPL, the subsequent lymphoma had to have a different morphology classification than the original childhood cancer, as defined by the International Classification of Diseases for Oncology, third edition classification. In addition, we excluded subsequent NHLs diagnosed after a primary NHL in childhood (N = 8) and subsequent HLs after a primary HL (N = 1), regardless of any difference in morphology. Also, all NHL survivors were excluded from analyses relating to subsequent NHLs, and all HL survivors were excluded from analyses relating to subsequent HLs.

Statistical analysis
Individuals entered the cohort at 5-year survival from childhood cancer diagnosis and remained at risk of an SPL until the first occurrence of loss to follow-up, death, or study exit date. Standardized incidence ratios (SIRs) were calculated as the ratio of the observed to expected numbers of lymphomas. Expected rates were calculated by accruing person-years at risk stratified by age, country, sex, and calendar year, and multiplying by the equivalent lymphoma incidence rates for the general population. 29 General population incidence rates (also stratified by age, country, sex, and calendar year) were obtained from the Cancer Incidence in Five Continents project. 30 Site-specific incidence rates do not exist for Hungary, hence these were estimated using Slovakian incidence rates because it is their neighboring country with similar demographic characteristics. 31 Absolute excess risks (AERs) were calculated as the observed minus the expected number of lymphomas, divided by person-years at risk, and multiplied by 100,000. This can be interpreted as the number of excess lymphomas observed beyond those expected per 100,000 person-years. Multivariable Poisson regression models were fitted to estimate the relative risk (RR) of all SPLs and subsequent primary NHLs and HLs while adjusting for the effect of potentially confounding factors: sex, childhood cancer diagnosis, country, decade of childhood diagnosis, age at childhood diagnosis, and attained age.
RRs can be interpreted as the ratio of SIRs, adjusted for potential confounders. Likelihood-ratio tests were applied to generate p values for linear trend for ordinal factors of interest or heterogeneity for nominal variables. Because of a relatively small number of observations in subsequent primary HLs, AERs could not be calculated.
SIRs, AERs, and RRs were calculated for the treatment factors radiotherapy (yes/no) and chemotherapy (yes/no) for those countries where <30% of treatment data were missing. Therefore, the Nordic countries and an Italian population-based cohort were completely excluded from analyses involving treatment variables regardless of whether treatment data were available to avoid potential bias. By using this approach, for 10.5% of survivors, the radiotherapy data were missing; and, for 12.4% of survivors, the chemotherapy data were missing. In all, information on radiotherapy was available for 90 SPLs, and information on chemotherapy was available for 88 SPLs.
To assess the probability of developing an SPL with increasing attained age, the cumulative incidence accounting for the competing risk of death was estimated for types of childhood cancer with ≥20 observed SPLs. All statistical analyses were conducted in Stata software (Stata Corporation). 32 All analyses were based on complete case analysis (i.e., list-wise deletion). A two-sided p value < .05 was considered statistically significant.

Cohort characteristics
Among 69,460 five-year survivors of childhood cancer, in total, 1,264,624 person-years were accrued, with loss to follow-up not exceeding 6% in any country (see Table S1). Over the follow-up period, 140 SPLs were observed in survivors across an age range from 5 to 88 years. From the 140 SPLs, 104 (74.2%) were classified as NHL, and 36 (25.7%) were classified as HL. Most of the SPLs occurred in males (66.4%) and in those surviving HL (n = 33) and leukemia (n = 31; Table 1).

Risk of subsequent primary Hodgkin lymphoma
Overall, childhood cancer survivors were no more likely to develop HL compared with the general population (SIR, 1.1; 95% CI, 0.8-1.5;   -431 T A B L E 2 Standardized incidence ratios, relative risks, and absolute excess risks, with 95% confidence intervals, for developing a subsequent primary lymphoma among a cohort of childhood cancer survivors   -433 T A B L E 3 Standardized incidence ratios, relative risks, and absolute excess risks, with 95% confidence intervals, for developing a subsequent primary non-Hodgkin lymphoma among a cohort of childhood cancer survivors (excluding non-Hodgkin lymphoma survivors)

Study limitations
One of the potential limitations of this study is the lack of detailed data concerning childhood treatment variables, such as cumulative doses of radiotherapy (including radiotherapy field) and chemotherapy. To fully assess the effect of treatment modalities on SPL risk, it would be beneficial to perform further analysis in the form of a nested case-control study in which detailed treatment information would be collected so that the risk of developing an SPL by cumulative doses of both radiation and chemotherapy agents can be calculated.
Another potential limitation of the study is the relatively small number of subsequent primary HLs observed, meaning that some components of the study may lack statistical power for further detailed analyses. That said, with more than three times the number of SPLs compared with the next largest study, this study is the largest cohort of childhood cancer survivors with subsequent HLs to date and hence provides the most accurate estimates of these risks. 10 Furthermore, it is feasible that, in some instances, the first primary lymphoma was misdiagnosed as HL instead of NHL. 12  RRs were derived from a model including sex, childhood cancer diagnosis, country, decade of childhood diagnosis, age at childhood diagnosis, and follow-up time.