Familial risk for non-Hodgkin lymphoma and other lymphoproliferative malignancies by histopathologic subtype: the Swedish Family-Cancer Database

Non-Hodgkin lymphoma (NHL) comprises a heterogeneous group of malignant tumors of the lymphoid tissue with distinctive cell lineage, histopathology, genetic profile, clinical features, and possibly distinct causes.^1-6  The overall incidence of NHL increased steeply in most developed areas of the world between the 1970s and the mid-1990s and leveled off thereafter.^7-9  In young adults, lymphoma rates are still increasing.¹⁰  Distribution of the most common subtypes of NHL (diffuse large B-cell and follicular) appears to differ by geographic region, suggesting differences in etiologic or host factors.^5-7,9,11,12  The difference is particularly striking for follicular NHL, which is most common in North America and western Europe,⁵  and for Burkitt lymphoma, which is endemic in equatorial Africa, but Burkitt lymphoma constitutes only 1% to 2% of lymphomas in the US and western Europe.⁶ 

The causes of NHL are largely unknown. Severe immunodeficiency, various infectious agents, familial aggregation, blood transfusion, and occupational exposure to pesticides and solvents have been consistently reported to increase the risk for NHL.^6,12,13  A few studies suggest that some of these risk factors may be more strongly associated with specific subtypes of NHL. For example, Epstein-Barr virus infection has been consistently associated with endemic Burkitt lymphoma,^6,14,15  HIV infection has been consistently associated with Burkitt lymphoma, and immunoblastic and diffuse large-cell NHL,^6,16,17  human T-cell lymphotrophic virus-1 with adult T-cell lymphoma,^6,18  and pesticide exposure have been consistently associated with follicular, small lymphocytic, and low-grade NHL.^19-21 

The evidence from epidemiologic studies of different designs shows that a history of hematolymphoid tumors in close relatives increases the risk for NHL by a factor of approximately 2 to 4.^22-29  Occasionally, higher risk has been reported for specific subgroups, such as siblings,^22,26,27  males,^5,27,28  and patients with late- or early-onset disease.^25,27  At least 3 studies assessed familial risk for specific subtypes of NHL, but the findings were inconsistent.^22,27,30  In addition, further data are needed on the familial clustering of histopathology specific NHL, including the effects of age and sex. We consulted the 2004 update of the Swedish Family-Cancer Database, including medical histology-confirmed tumor diagnoses, and we assess here the familial associations of different histopathologic subtypes of NHL. The large nationwide coverage allowed us to consider a relatively recent and short period of time (10 years), when diagnostic indicators are thought to be relatively consistent.

The Swedish Family-Cancer Database was created in the mid-1990s by linking census information, death notifications, and the administrative family register at Statistics Sweden to the Swedish Cancer Registry.^31-33  Included are data on each child's parents at the time of birth. Each person is assigned a unique technical identification number, allowing the construction of family lineages. The database was updated at the end of 2004 to include persons born in Sweden after 1931 and their biologic parents, totaling 10.5 million persons. Housed in the database are details on 3.6 million nuclear families, parents and offspring. Data on patients with neoplasms from 1958 to 2002 were retrieved from the Swedish Cancer Registry. The Swedish Cancer Registry is composed of compulsory reports of patients provided by clinicians/pathologists or cytologists and is now considered to be almost 100% complete.³³  Pathologists and cytologists separately report every cancer diagnosis on surgically removed tissues, biopsies, cytologic specimens, bone marrow aspirates, and autopsies. The incidence of tumors according to the database has been previously validated.³¹  Data on parity are complete, and data on socioeconomic index and area of residence are based on population census figures from 1960, 1970, 1980, and 1990.

Four-digit diagnostic codes from the seventh revision of the International Statistical Classification of Diseases (ICD-7) and subsequent ICD classifications are available. In the current study, NHL was indicated by ICD-7 code 200 or 202. Beginning in 1993, histopathology has been recorded according to the College of American Pathologists Systematized Nomenclature of Medicine (SNOMED).³⁴  In the present analysis, we assessed the risks of histopathology-specific subtypes of NHL and other related lymphoproliferative malignancies, including multiple myeloma (ICD-7 code 203) and chronic lymphocytic leukemia (ICD-7 code 204.1) for the offspring whose parents or siblings were affected by tumor at the same site. Thus, the analyses of histopathology-specific risks considered diagnoses made between 1993 and 2002 and included 4455 patients diagnosed with NHL from 6.97 million offspring and 23 875 patients with NHL from 6.4 million parents. In total, 50 387 offspring had a parental history of NHL, contributing 487 004 person-years of follow-up. NHL subtypes classified according to the SNOMED and translated in World Health Organization (WHO) categories⁶  consisted of 589 diffuse large B-cell, including 508 centroblastic (96833), 57 anaplastic large B-cell (97143), 20 B-cell immunoblastic (96843), and 8 histocytic (96803). Follicular NHL numbered 942 and included 484 centroblastic-centrocytic (96923), 426 follicular NHL not otherwise specified (NOS) (96903), 24 centroblastic (96973), 5 nodular lymphocytic well differentiated (96933), 2 nodular lympho-histocytic mixed (96913) and 1 nodular lymphocytic poorly differentiated (96963). T-cell NHL included 158 T-cell NOS (95903), 67 anaplastic large T-cell (97143), 64 peripheral (97023, 97053, 97063, 97073), 53 mycosis fungoides (97003), and other less common subtypes. In the present analysis we also included 752 B-cell NOS (95903), 903 NHL NOS (95913, 97603), and 894 lymphomas grouped as “others” representing 212 lymphoplasmacytic lymphoma/Waldenström macroglobulinemias (96713, 97613), 113 mantle cell lymphomas (96743, 96723), 151 hairy cell leukemias (99403), 74 Langerhans cell lymphomas (77860), 65 Burkitt lymphomas (96873), and less frequent miscellaneous subtypes. Related lymphoproliferative malignancies were 1291 multiple myelomas, including 1211 plasma cell (97323, 98303) and 80 plasmacytoma NOS (97313), and 990 chronic lymphocytic leukemias. The age of parents was not limited, but the maximum age of offspring was 70 years.

Follow-up was started from the date of birth, the date of immigration, or January 1, 1993, whichever occurred last. Follow-up ended on the date of diagnosis of the first primary neoplasm, the date of death, the date of emigration, or the closing date of the study, December 31, 2002, whichever occurred first. Standardized incidence ratios (SIRs) were used to estimate familial relative risks. SIRs were calculated in offspring for whom only parents were affected and offspring for whom only siblings were affected. SIRs were computed as the ratio of the number of observed cases to the number of expected cases. Expected numbers were based on incidence rates standardized for sex, age (5-year bands), socioeconomic status (6 groups), residential area (4 groups), and parity (6 groups, ranging from no births to more than 5 births). Age was the most important standardization variable, whereas successive standardization for sex, socioeconomic status, residential area, and parity did not materially alter the risk estimates. Confidence interval (95% CI) and P values (P) were calculated under the assumption of a Poisson distribution.³⁵ 

The distribution of offspring with NHL, multiple myeloma, or chronic lymphocytic leukemia according to age at diagnosis and sex is summarized in Table 1. Overall mean age at diagnosis was 51 for NHL, 57 for multiple myeloma, and 57 for chronic lymphocytic leukemia. Male-to-female ratios ranged between 1.2:1 and 2.4:1.

Table 2 shows the risk for histopathology-specific NHL for offspring with a parent or a sibling affected with NHL. SIRs for persons with a parental history of NHL were significantly increased for NHL (SIR = 1.8; 95% CI, 1.4-2.2), diffuse large B-cell (SIR = 2.3; 95% CI, 1.2-3.8), and NHL NOS (SIR = 1.8; 95% CI, 1.1-2.8). SIRs for siblings of affected patients were significantly increased for NHL (SIR = 1.9; 95% CI, 1.1-3.2), follicular NHL (SIR = 2.3; 95% CI, 1.1-4.8), and B-cell NOS (SIR = 3.4; 95% CI, 1.5-6.8). No siblings with diffuse cell NHL were observed. No increased risk was found for T-cell NHL or for multiple myeloma.

Table 3 shows the familial risk in offspring for the occurrence of histopathology-specific malignancies. The risk for NHL in persons with a parental history of diffuse large B-cell was 4.5 (95% CI, 1.9-8.9), and the SIR for a parental histology-specific concordant neoplasm was 11.8 (95% CI, 2.2-34.8). We found no parent/offspring pair with a diagnosis of diffuse large B-cell and follicular NHL. Parental history of follicular NHL significantly increased the risk for NHL (SIR = 3.6; 95% CI, 1.5-7.1) and follicular NHL (SIR = 6.1; 95% CI, 1.1-18.0). The SIR for a parental history of B-cell NOS was 4.9 (95% CI, 0.5-17.9) for a concordant malignancy. None of the histopathologic subtypes were significantly associated with a parental history of T-cell lymphoma and NHL NOS. The risks for multiple myeloma (SIR = 2.3; 95% CI, 1.4-3.5) and chronic lymphocytic leukemia (SIR = 5.9; 95% CI, 3.9-8.6) were increased for a parental history of a malignancy at the same site. The magnitude of the risk tended to be higher for a histopathology-specific concordant neoplasm than for any other subtype; however, because of the sparse number of cases, the difference reached statistical significance only for multiple myeloma and chronic lymphocytic leukemias. An analysis of specific subtypes among siblings was not feasible because of the small number of familial cases.

Familial risks for a parental history of NHL according to offspring sex are presented in Table 4. No significant association emerged for different subtypes of NHL in males with a parental history of NHL. SIRs in females were significantly increased for NHL NOS (SIR = 2.6; 95% CI, 1.3-4.9) and chronic lymphocytic leukemia (SIR = 2.3; 95% CI, 1.1-4.4). Risks for NHL were not significant for brothers. In contrast, sisters with NHL were at risk for NHL (SIR = 3.1; 95% CI, 1.3-5.9), follicular NHL (SIR = 4.2; 95% CI, 1.5-9.3), and B-cell NOS (SIR = 6.3; 95% CI, 2.3-13.9). When we repeated the analyses for males with only a father affected and the risk for females with only a mother affected, the numbers of cases were small and the risks were not significantly increased compared with the general population (data not shown).

We have confirmed and further quantified a significant familial aggregation of NHL in family members of patients affected with NHL.^22-25,27,28  In the present study, family history of NHL was based on registered family structures covering virtually the entire Swedish population and medical diagnoses with histopathologic confirmation, excluding any type of recall bias. The ascertainment of relatives was complete, giving further reassurance.

Familial aggregation of NHL has been associated with a rare inherited defect of immune function and an autosomal-dominant transmission in some families. However, in most cases, the pattern of inheritance remains poorly understood.³⁶  In the present analysis, we found a strong familial risk for NHL for most common subtypes of NHL, with approximately a 2- to 4-fold increased risk. T-cell NHL was the only subtype that appeared to be not associated with a familial history of NHL. When the histopathologic codes where applied to parents and offspring, a parental history of T-cell NHL increased significantly the overall risk for NHL. With reference to other related lymphoproliferative malignancies, we found no association with multiple myeloma and a significant association with chronic lymphocytic lymphoma. In the Swedish Family-Cancer Database, diffuse large B-cell was the second most common subtype of NHL and the one that showed the highest familial risks; however, it is noteworthy that no siblings with diffuse large B-cell lymphoma were observed. For the other subtypes of NHL, the risk in persons with a sibling affected with NHL was generally higher than it was in persons with a parental history. Higher risks in siblings without affected parents, in comparison with risks to parents, may indicate recessive components of the disease.^37,38  Our finding of a higher familial risk in females compared with males is tentative because of a small number of familial cases. However, as previously reported,^22,26,27  this pattern of sibling-specific familial aggregation for NHL would also be consistent with shared environmental exposure during childhood. It is plausible that NHL may cluster in families to some degree of shared environmental determinants, including various infectious agents, dietary habits, pesticides, and occupational exposure.^21,38-40  However, a recent report from this database estimating the contribution of environmental effects comparing familial risks among siblings close or distant in age suggests that, in this population, the environmental component in the etiology of NHL and other lymphoproliferative malignancies is likely to play a marginal role.³⁸ 

Limited and inconsistent epidemiologic evidence is available on the familial risk for specific histopathologic subtypes of NHL. In a case-control study, Chatterjee et al²⁷  reported a significant familial association for follicular NHL (hazard ratio [HR] = 4.5; 95% CI, 1.3-15.6) but not for diffuse NHL in case patients with a family history of NHL. Conversely, a case-control study from Iowa and Minnesota showed an increased risk of 5.4 (95% CI, 1.4-20.2) for diffuse cell NHL, but not for follicular NHL in persons with a sibling affected with any lymphoma.²²  Chiu et al³⁰  found a significantly increased risk for both follicular (odds ratio [OR] = 2.8) and diffuse large-cell NHL (OR = 2.8) in subjects reporting a family history of any hematopoietic malignancy. In our study, the familial risks for a parental history of NHL ranged between 1.0 for T-cell NHL and 2.3 for diffuse large B-cell NHL. We could not estimate the risk for diffuse cell NHL in siblings because there were no affected sibling pairs with this diagnosis. For follicular NHL, however, the sibling risk was 2.3.

We found remarkably high familial risks for parental concordant cancer for diffuse large B-cell, follicular, and B-cell NOS. However, because of the rarity of the disease and the low proportions of familial cases, some of these associations were based on small numbers. To our knowledge, none of the studies published provide familial risk estimates for histopathology-specific subtypes in parents and offspring.

The availability of histopathology-specific data only for a 10-year period is a limitation of the current study. However, a short follow-up period is also an advantage because diagnostic classifications have not changed much. Diagnostic criteria for lymphomas have changed extensively in the past decades. Reexamination of the pathologic specimens from our database has led to adjustments of the original Hodgkin disease diagnoses to NHL in some 10% of cases, in broad agreement with results from other population-based studies.^6,41-43  These data suggest that the proportion of diagnostic misclassification for NHL is not likely to substantially affect risk estimates from cancer registry-based studies. Moreover, the histopathology-specific diagnoses of parents and offspring were made between 1993 and 2002. In such a short period of time, we assumed diagnostics to be relatively consistent and the histopathologic associations not likely to be biased by different diagnostic criteria between parents and offspring. Proportions of different types of histopathologic diagnosis did not substantially vary across different medical regions or for the entire period of follow-up, providing further reassurance against misclassification bias. The power of our study is, at least in part, limited by the consistent proportion of cases that were diagnosed as either NHL or B-cell NOS, with no specific information on the histopathologic subtype. However, this lack of information is not likely to affect the familial risk estimates, given that there is no reason to believe that familial cases of specific subtypes of NHL were systematically included in the NHL and B-cell NOS groups.

In conclusion, the current investigation represents the largest population-based study on familial risk for NHL and the only study assessing familial risk for concordant cancer in parents and offspring by histopathologic subtype. Our results provide evidence of a remarkable familial clustering for NHL, with few differences in the magnitude of risks for various histopathologic subtypes, consistent with an autosomal-dominant component of the disease. The excess familial risks for different histopathologic subtypes suggest that they share, at least in part, common etiologies. The patterns of risks in parents and siblings support the hypothesis of an autosomal-dominant component for diffuse large B-cell NHL and a recessive component for follicular NHL.

The Family-Cancer Database was created by linking registries maintained by Statistics Sweden and the Swedish Cancer Registry.