INTRODUCTION

Carcinogenic effects of exposures to ionizing radiation, occupational chemicals, and tobacco smoke have been the primary sources of quantitative data on cancer risks in humans.^1-3 Effects of dose, duration of exposure, and time from exposure to cancer have been characterized.^2,4-8 As a new source of quantitative data on cancer risks, survivors of first primary cancers are followed for treatment-related risks of second cancers, with implications for therapeutic adverse effects and human susceptibility in general.^9,10 Active research on subsequent cancers has been carried out since the 1980s, but because of the numbers of patients enrolled and extended follow-up times, recent studies provide the most comprehensive evidence on the magnitude of the effects.^9-13 For example, two large cohorts of patients with Hodgkin's disease revealed high risk of many solid tumors, in addition to the well-known risks for leukemia and lymphoma in these patients.^14,15 One of these studies which covered patients diagnosed at a large age range provided some evidence on higher risk for subsequent neoplasms in patients diagnosed at an early age.¹⁵ Definite data on the inverse age effect was provided by a recent multicenter study.¹⁶

Non-Hodgkin's lymphoma (NHL) is more common than Hodgkin's disease and it is primarily treated with chemotherapy and local radiation therapy. Recent articles on subsequent tumors after NHL show increased risks of many solid and lymphohematopoietic neoplasms.^17-19 The largest of these studies reported subsequent neoplasms among 109,000 patients with NHL from 13 international data sets, including an early version of the Swedish Family-Cancer Database.¹⁷ The study found an inverse age affect for stomach, lung, and bladder cancers, and for myeloma and leukemia, but the covered age span was narrow as the youngest diagnostic age group was younger than 56 years. The other large study covered a full age range, showing the highest risk for second solid tumors in patients treated with radiation who were diagnosed before age 25 years and in other patients diagnosed between ages 25 and 49 years.¹⁸ This study addressed the question about age-group specific modification of risk since exposure but found no effect. This question was taken up in a recent study on 1,490 solid cancers after Hodgkin's diseases finding a strong effect of attained age.¹⁶

We use data from the Swedish Cancer Registry to assess risks for second cancers after NHL diagnosis in 28,131 patients by considering the age at NHL diagnosis and time since diagnosis, focusing on the 2,256 diagnosed solid tumors. The novel aspects of this study, in addition to the large sample size from a single country, include a wide age range at NHL diagnosis and an assumed uniform subtype-specific treatment of adult patients. Large differences in site-specific risks were observed. While the relative risk of solid tumors decreased by age overall, the response depended on tumor type. The effect of time since diagnosis was a complex function of diagnostic age and attained age.

PATIENTS AND METHODS

The Swedish Family-Cancer Database includes persons born in Sweden after 1931 with their parents, totaling more than 11.5 million persons and more than 1.2 million tumor notifications (for a detailed description of the Database and its last update see Hemminki et al).^20,21 Cancer cases were retrieved from the Swedish Cancer Registry, which relies on separate compulsory notifications from clinicians who diagnosed the neoplasms and pathologists/cytologists.^20,21 All registered NHL cases are histologically verified. An ad hoc study on the diagnostic accuracy of second neoplasms in the Swedish Cancer Registry found 98% to be correctly classified.²² In this analyses, 29,134 patients with first primary NHL were observed from diagnosis until death, detection of a second primary cancer, emigration or December 31, 2004, or whichever came first. Second cancers were classified as such by the Cancer Registry, including synchronous tumors. The incidences of second malignancies among patients with NHL were compared with the rates in the general Swedish population by standardized incidence ratios (SIRs) and 95% CIs, adjusting for covariates age (5-year bands), sex, socioeconomic index (six groups), region (four groups), and calendar year (1961 to 1964, 1965 to 1969, and so on to 2000 to 2004). Separate analyses were carried out according to age at NHL diagnosis (20 to 39, 40 to 49, 50 to 59, 60 to 69, and 70+ years), time since NHL diagnosis (0 to 10, 11 to 20, 21 to 30, and 31+ years), and period of NHL diagnosis (1961 to 2004 and 1975 to 2004). SIR applies indirect standardization, which is particularly suitable for cells with small numbers of subjects.^1,23 In this method the observed number of cases is divided by the expected number of cases, calculated from the whole background population of 11.5 million individuals.

The Swedish Cancer Registry is lacking historic clinical and treatment data. The therapeutic principles for NHL in Sweden (online-only Appendix) initially applied to radiation therapy.^24,25 Multidrug cyclophosphamide, doxorubicin, vincristine, and prednisone (CHOP) chemotherapy was instituted in the 1970s and it has been extensively used since. In the late 1980s, autologous bone marrow transplantation with very high-dose chemotherapy (often with cyclophosphamide or other alkylating agents) was introduced for patients with relapsing high-grade NHL. HIV infections are relatively uncommon in Sweden and their contribution to the incidence of NHL is thought be small, in line with the rare occurrence of the HIV-related Kaposi's sarcoma.^26,27

RESULTS

According to Table 1, among 29,134 patients with NHL, 2,290 developed second solid tumors during 141,915 years of follow-up. Men were more common among patients with NHL and even more so among those diagnosed with a subsequent solid tumor. Most of the patients (88.5%) were diagnosed after 1974 when the CHOP therapy had become standard in Swedish hospitals. The most common follow-up duration (64.6%) was less than 5 years but 14.7% of the patients were observed over 10 years; 87% of the second solid tumors were diagnosed in a different calendar year from the first NHL.

The SIR for solid tumors after NHL was 1.65 (2,290 patients) and that for lymphohematopoietic neoplasms was 5.36 (369 patients; Table 2). Among the 25 most common solid tumors (only sites with > five cases are shown in Table 2), the SIRs were increased for all but nine sites; for endometrial cancer the SIR was decreased (0.52). The highest risk was observed for spinal meningioma (40.8) and even the nervous system, excluding the brain, showed a remarkably high SIR of 18.1. Other sites with SIRs over 5.0 were the nose and the thyroid gland. The SIRs for aerodigestive tract (3.31), small intestinal (4.46), anal (3.35), and skin (squamous cell, 4.28) tumors were also high.

Table 3 shows the relative risks for all patients with solid and lymphohematopoietic tumors after NHL, diagnosed at various ages. The risk for younger patients is not presented in Table 3 because only 19 solid tumors were recorded (SIR for solid tumors 4.52; 95% CIs, 2.88 to 7.09) and because the treatment may be more aggressive for young patients. The SIRs declined in an age-dependent manner from 3.02 in diagnostic age 20 to 39 years to 1.12 in diagnostic age 70+ years. The length of follow-up also declined but the SIR is not sensitive to differences in the follow-up time. For patients with lymphohematopoietic neoplasms, the relative risks were higher and the decline was also dependent on diagnostic age, declining from 10.9 (20 to 39 years) to 1.73 (70+ years). The pattern of declining SIR was shared for leukemia and lymphomas whereas for myeloma an increased SIR was noted for one diagnostic age group only. Because of the relatively small number of patients with subsequent lymphohematopoietic neoplasms, they will not be analyzed further.

SIRs for individual solid tumors are presented in Table 3 for sites with more than 50 cases and a significant increase in Table 1. All the 11 listed sites showed significant increases at least for some diagnostic age groups. Squamous cell skin cancer and nervous system tumors were in excess in patients at each diagnostic age group; the SIR for patients with brain meningeal tumors was 19.7 in the youngest diagnostic group (the nervous system tumors outside the brain, with a very high risk in Table 1, numbered only 11 and were not analyzed). In the youngest diagnostic age group, also the SIR of gastric cancer was high (12.1). Breast cancer was increased only in patients diagnosed in the youngest age group. The youngest diagnostic groups showed the highest risks for all cancers (nine sites), except for prostate cancer (age group, 40 to 49 years) and squamous cell skin cancer (age group, 60 to 69 years). However, notably, even those treated at age 70+ years were at an increased risk at many sites, yet with an overall SIR of 1.12 only. In this age group the SIR for prostate cancer was decreased to 0.72.

Table 4 presents SIRs for patients with solid tumors since NHL diagnosis by diagnostic age groups. In the youngest diagnostic age group, the SIRs were equal through three first decades. In diagnostic age group 20 to 39 years, the trend for SIRs was decreasing whereas for the next older diagnostic age group it was relatively even. For all the higher diagnostic age groups the trend was increasing, with the exception of last follow-up time (31+ years). We carried out a separate analysis for NHL diagnosed after 1974 (Table 5) in order to cover the period of the modern chemotherapy (see Patients and Methods). In the youngest diagnostic age group, the SIRs for solid tumors appeared to be higher in Table 5 compared with Table 4 and the risk reached a maximum 11 to 20 years after diagnosis. In the diagnostic age group, 20 to 39 years, the SIRs appeared to be lower than in Table 4 but the trend of decline since diagnosis remained. In all diagnostic age groups over 50 years the SIRs were almost equal in Tables 4 and 5 until 20 years after NHL diagnosis, thereafter they increased more sharply in Table 5.

Figure 1 shows the incidence of solid tumors in the Swedish general population (bottom) and in patients with NHL diagnosed at various ages, which are shown with the numbers in the brackets. The points for the incidence curves were adjusted to the median age in each age band, for example 16 years for patients diagnosed in age band 0 to 19 years. The visual impression from Figure 1 agrees with the data in Table 3 and 4: at young diagnostic ages the relative difference between the incidence rates patients/general population is stable or decreasing with time, while at more advanced diagnostic ages the relative difference increases with time. An incidence plot for the data that contributed to Table 5, showed even a clearer upward deviation over the background incidence for patients diagnosed at ages older than 50 years and observed over 20 years (data not shown).

DISCUSSION

The risks for subsequent tumors after NHL were higher for lymphohematopoietic neoplasms than for solid tumors, in agreement with previous studies observing NHL patients.^17-19 The focus of this article was on solid tumors and we discuss lymphohematopoietic neoplasms no further. The risk of solid tumors are known to be related to the length of the follow-up time after therapy,¹⁷ which explains some of the differences in the literature, together with the population size. For example, a small study on subsequent neoplasms after NHL found an increase only for lung cancer among solid tumors.¹⁹ Another study, based on the US Surveillance, Epidemiology and End Results (SEER) Program, found an increase for eight solid cancers but also a decrease for breast and prostate cancers.¹⁸ The largest published study found an increase for 21 solid sites and a decrease for none.¹⁷ This study detected an increase for 16 sites and one decrease (endometrium). Biologic reasons for a decreased risk for subsequent cancer could be, for example, the castrating effects of NHL therapy which could reduce the risks at hormone dependent sites, such as the endometrium. However, a decrease in risk would also be expected if subsequent tumors were underascertained, which may be the reason for the reported decreased risks based on the SEER Program.¹⁰ The overall risks of this SEER data based study, 1.1 for solid tumors,¹⁸ were also lower than 1.65 in this study and about 1.4 in the largest study with relatively short follow-up times.¹⁷ It should be noted that the reporting of multiple neoplasms is generally considered high in the Swedish Cancer Registry.¹⁰ However, we have previously observed deficits of second neoplasms after cancers with higher fatality than NHL (stomach and pancreatic cancers).²⁸ The limitations and the potential pitfalls of the study include lack of data on treatment, short follow-up times, small numbers of subjects after the subdivisions of the initially large study population, and large increase in the incidence of NHL during the study period for unknown reasons.²⁶

These data on the reverse age effects of the risk for solid tumors after NHL refine the previous data on subsequent tumors after NHL and Hodgkin's disease.^14-19 In this adult population the therapeutic regimens have probably been similar over time and age groups, however with the caveat that older patients (ie, older than 70 years) may have received less treatment in terms of dosages and duration than younger patients, due to more concomitant other diseases and less bone marrow reserve capacity.^29-31 There were large differences in relative risks by individual cancer sites and tissues, with spinal meningioma showing the remarkably high SIR of 40.8. Meningiomas in brain showed a smaller SIR of 3.10 overall, but the SIR for brain meningeal tumors was 19.7 in the youngest diagnostic group. Increased numbers of meningiomas have also been reported in survivors of childhood acute lymphoblastic leukemia.¹² Squamous cell skin cancer and nervous system tumors were in excess in patients at each diagnostic age group; associations of nonmelanoma skin cancers and NHL have been observed earlier.^17,32-34 Breast cancer was only increased in patients diagnosed in the youngest age group. In this diagnostic age group, also the SIR of gastric cancer was high (12.1). The youngest diagnostic groups showed the highest risk for all cancers, except for prostate cancer and squamous cell skin cancer. However, even patients treated at age 70+ years were at an increased risk of many cancers. The overall SIRs for these sites were 1.43, 1.27, and 1.02, respectively.

Among survivors of Hodgkin's disease, the relative risk of breast cancer declined rapidly in those diagnosed before age 20 years and with a much slower space in those diagnosed at age 40 or older.¹⁶ For other solid tumors, the relative risks were almost unchanged decades after diagnosis. These data provide insight into the carcinogenic process showing that age at treatment determines both the magnitude of the initial relative risk and also the time-dependent modulation of the response after exposure. The young patients (diagnosis < 20 years) had a high initial relative risk of solid tumors with hardly any subsequent decline; in young adults (diagnosis, 20 to 39 years), the initial response was lower and the relative risk declined in the course of two decades; in older patients (diagnosis, 50+ years), the initial response was weak but the relative risk increased, particularly 20 years or more after diagnosis. The trends were reproduced but with even stronger effects in the period 1974 to 2004 when the CHOP therapy or its variants have been the standard treatment for high-grade NHL in Sweden.

Understanding the biologic rationale for the findings would probably teach us some fundamental principles in human carcinogenesis. The initially high relative risks in young patients probably reflect sensitivity of the growing organism to an aggressive and potentially carcinogenic therapy. The background incidence of NHL is very low through the first three decades of life but at around 30 years it starts increasing (Fig 1); thus, the relatively even SIRs for the youngest patients through their first two or three decades after diagnosis (with attained age 30 to 50 years) imply somewhat decreasing excess numbers of cases (subtracting the background incidence). For young adults (20 to 39 years at diagnosis), the SIRs declined through the following two or three decades. However, the decline in SIR took place in the attained age 30 to 60 years when the incidence of NHL increased rapidly and the excess number of cases was actually somewhat increasing up to 30 years since NHL diagnosis (Fig 1). Similarly, in diagnostic age groups 40 to 69 years the excess number of cases increased more than the SIR implied because of the increasing background incidence. Large numbers of excess cases were observed for patients diagnosed between ages 50 and 69 years when followed 21 to 30 years. Thus, these persons, with attained age 71 to 99 years, appeared to be particularly vulnerable. For patients diagnosed at age 70+ years the excess number of cases was limited but, based on Tables 4 and 5, the number of surviving patients was also small. Thus, the conclusion from the data of Figure 1 is that, with the exception of the youngest patients, the excess number of therapy-related cases increased in every decade since NHL diagnosis, with a relatively higher increase in older patients. Thus the therapy-related damage persists at least 30 years and the toll of solid tumors as therapy-related cases is largest 21 to 30 years after diagnosis, which need to be considered in the medical surveillance schemes of the patients with NHL. However, longer follow-up periods of patients will be needed to fully characterize the long-term effects of NHL therapy. Moreover, it is likely that the responses of the individual solid tumors are variable,¹⁶ which calls for detailed studies on large patient series.

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

The author(s) indicated no potential conflicts of interest.

AUTHOR CONTRIBUTIONS

Conception and design: Kari Hemminki, Justo Lorenzo Bermejo

Financial support: Kari Hemminki, Jan Sundquist

Collection and assembly of data: Kari Hemminki, Jan Sundquist, Justo Lorenzo Bermejo

Data analysis and interpretation: Kari Hemminki, Per Lenner, Justo Lorenzo Bermejo

Manuscript writing: Kari Hemminki, Justo Lorenzo Bermejo

Final approval of manuscript: Kari Hemminki, Per Lenner, Jan Sundquist, Justo Lorenzo Bermejo

Appendix