Errata: Corrigendum Volume 145, Issue 4, 551, Article first published online: 24 April 2009
This study has been presented in part at the Annual Meeting of the American Society of Hematology in December 2006 and at the IW-CLL Meeting in September 2007.
Jennifer R Brown MD PhD, Assistant Professor of Medicine, Harvard Medical School, Dana-Farber Cancer Institute, 44 Binney Street, Boston, MA 02115, USA. E-mail: firstname.lastname@example.org
Increasing evidence points to a heritable contribution in the development of lymphoma. The goal of this study was to determine the rate of familial lymphoproliferative malignancy among consecutive lymphoma patients presenting to a tertiary care center and to enrol families with multiple affected first-degree relatives on a data and tissue collection study. Beginning in 2004 all new patients presenting to the Dana-Farber Cancer Institute with non-Hodgkin (NHL) or Hodgkin lymphoma (HL) or chronic lymphocytic leukaemia (CLL) were asked to complete a one-page self-administered family history questionnaire. 55·4% of 1948 evaluable patients reported a first-degree relative with a malignancy, with the highest rate among CLL probands. Lymphoid malignancies were particularly common, with 9·4% of all probands reporting a first-degree relative with a related lymphoproliferative disorder (LPD). This frequency was again highest for CLL, at 13·3% of CLL probands, compared to 8·8% of NHL probands and 5·9% of HL probands (P = 0·002). The prevalence of CLL was significantly increased in parents of CLL probands (P < 0·05), and a greater risk of NHL was seen in fathers of NHL probands than in mothers (P = 0·026). We conclude that familial aggregation of LPDs is common among newly diagnosed patients, varies significantly by diagnosis and contributes meaningfully to the population disease burden.
The most recent pooled case–control analysis of the InterLymph Consortium confirmed an increased familial risk across the spectrum of lymphomas and found higher risks in particular for those with affected siblings or affected male relatives (Wang et al, 2007). The largest familial risk appears to be associated with CLL, based on an analysis of the Swedish Family Cancer database, which found a 7·5 times relative risk for CLL in a first-degree relative of a CLL patient, as compared to a 1·45 times risk for NHL and a 2·35 times risk for HL (Goldin et al, 2004). In this study no difference in risk was found between parents or siblings or related to the sex of the relative (Goldin et al, 2004), although earlier studies in CLL alone had suggested a higher familial risk in females, whether proband cases or relatives (Linet et al, 1989; Cannon-Albright et al, 1994).
Given these findings, significant effort has been devoted to attempting to identify genes or loci involved in the heritability of CLL and lymphoma. A number of small linkage studies were completed initially in CLL (Goldin et al, 2003; Ng et al, 2006), culminating recently in the identification of a region on chromosome 13 (13q21) as a possible site of susceptibility in familial CLL (Ng et al, 2007). Unfortunately, a systematic search for mutations in protein-coding genes in that region found none (Ng et al, 2007). Studying the minimal region of deletion at 13q14 has identified the role of MIRN15A and MIRN16-1 in the pathogenesis of sporadic CLL, but their involvement, if any, in familial CLL remains to be determined (Calin et al, 2002, 2005). Other recent genome-wide linkage searches have identified different possible susceptibility loci, including 2q21, 6p22 and 18q21 (Sellick et al, 2005, 2007). Recently a polymorphism in the death-associated protein kinase 1 gene (DAPK1) was associated with familial CLL in a single family, but this allele has not been found again in a larger number of cases (Raval et al, 2007). Thus, to date, the search for genes involved in a significant percentage of familial lymphoproliferative disease has been disappointing.
Despite this significant research effort focused on possible mechanisms of heritability in lymphoproliferative disorders (LPDs), the implications for clinical practice of small relative risks identified in epidemiological studies have been obscure. How often will physicians see patients with a LPD who also have a close relative with the same or another LPD? What should the physician tell a curious patient who wonders about familial associations in lymphoma? In order to help us better understand the answers to these clinically relevant questions, we initiated systematic screening of all patients with LPDs seen at the Dana-Farber Cancer Institute (DFCI), beginning in 2004, to determine the clinical significance of familial predisposition and the resulting impact on family members. As the frequency of other malignancies in first-degree relatives of probands with lymphomas was also unknown, we collected data on solid tumours as well.
Patients and methods
Selection of patients
Systematic screening of consecutive new patients seen in the DFCI Lymphoma Programme, which includes CLL, began in November 2004 and is ongoing through the present. The data presented here cover the period November 2004 to February 2007. Eligible patients included those with any NHL, HL or CLL/small lymphocytic lymphoma (SLL). A one-page Institute Review Board (IRB)-approved screening questionnaire was mailed to each patient prior to the visit and supplied again at the clinic, to be completed prior to their visit. The questionnaire asked about any history of malignancy in parents, siblings or children, and about any history of LPD in any other blood relative. Those patients who reported a LPD in a first-degree relative were invited to participate in a more detailed study, which included a more extensive family history questionnaire, medical history, confirmation of the presence of a pathologically-proven LPD in the proband and relative, and germline and tumour banking. The family history questionnaires employed in this study may be viewed at http://research3.dfci.harvard.edu/jbrown/.
During this period, 2012 patients with LPDs were scheduled to be seen, of whom 64 were ineligible, due to adoption or unknown status of parental cancers, leading to 1948 evaluable and screened patients. The questionnaire return rate was 71%. For the remaining patients, physician notes were reviewed. Of these 1948, 183 [9·4%; 95% confidence interval (CI) 8·1–10·8%] reported a first-degree relative (alive or dead) with a LPD. Of that group, 62 had been registered on the more detailed tissue banking study. Many have not been registered because the affected relatives were deceased and lacked available records or pathology.
This protocol was approved by the DFCI IRB, and informed consent was obtained from all patients enrolled on the more detailed study. Those patients who did not wish to return the screening questionnaire could choose not to, without follow-up from study staff; alternatively patients could return the questionnaire but decline follow-up contact.
Proband diagnoses were confirmed by DFCI physicians and independent haematopathology review at the Brigham and Women’s Hospital. For analysis purposes, subgroups were created that included all NHLs, HL (which included nodular lymphocyte predominant HL), and CLL/SLL. For those patients who reported a World Health Organization or Revised European-American Lymphoma Classification diagnosis in a relative, that diagnosis was recorded. For those who reported HL, or NHL not otherwise specified, those diagnoses were noted as such. ‘Lymphoma not otherwise specified’ was taken as NHL for the purposes of the analysis (Glaser et al, 2007). ‘Leukaemia not otherwise specified’ was excluded due to the ambiguity with acute leukaemias or chronic myeloid leukaemia. ‘Leukaemia not otherwise specified’ was noted in 26 first-degree relatives, 33 second-degree relatives and 12 third-degree relatives.
For solid malignancies, common primary malignancies were noted as such. If a single individual was listed as having both a common primary malignancy and a common site of metastasis for that malignancy (for example, lung with bone or liver), they were coded as lung cancer and not considered to have separate bone or liver cancer.
Frequency tables have been analysed by the chi-squared test. Unless otherwise noted, the nominal P values given are for the three-group comparison (CLL, NHL, HL). 95% CIs are exact binomial confidence intervals. The nominal P values for the comparison of reported solid tumour rates among parents of CLL probands and NHL probands were determined using the Fisher exact test. The P value for the comparison of maternal to paternal rates of NHL was corrected for multiple comparisons using the method of Bonferroni. P values <0·05 were considered significant.
The screened population consisted of 62% NHL patients, 23% CLL/SLL patients and 15% HL patients, as shown in Table I. The breakdown of NHL subtypes is also shown in Table I. The median age of screened NHL patients was 58 years (range 16–98), comparable to the median age of screened CLL patients, which was also 58 years (range 28–90). HL probands were significantly younger, with a median age of 36 years (range 18–85). The NHL and CLL probands were younger than the median age at diagnosis in the Surveillance, Epidemiology, and End-Results (SEER) population (67 and 72 years respectively), possibly reflecting referral bias, although the age of the HL probands was comparable to SEER (38 years). The age of the probands was significant, because the data from the NHL and CLL groups were directly comparable, whereas the HL group was likely to have fewer evaluable first-degree relatives and fewer first-degree relatives affected with any cancer, due to the younger age of the probands.
Of the entire screened population, 55·4% reported a first-degree relative with cancer. This frequency differed significantly by diagnosis, with 63·7% (95% CI 59·1–68·2%) of CLL patients reporting at least one affected first-degree relative, compared to 56·1% (95% CI 53·3–58·9%) of NHL patients and 39·2% (95% CI 33·5–45·1%) of HL patients (P < 0·0001 for the three group comparison; P = 0·04 for CLL compared to NHL). If second-degree relatives with cancer were considered in addition, 65·0% of all probands reported a relative with malignancy: 70·1% (95% CI 65·7–74·3%) of CLL probands, which was significantly greater than NHL probands at 64% (95% CI 61·3–66·8%) and HL probands at 61·2% (95% CI 55·3–66·9%) (P = 0·02 for the three group comparison and P = 0·02 for CLL compared to NHL). It is important to note that these differences were significant between CLL and NHL, where the probands were of similar age, not just between HL and CLL/NHL.
The median number of first-degree relatives affected with cancer was one for all groups, among probands reporting any first-degree relatives with cancer, but the distribution shifted significantly by diagnosis (Fig 1), with more CLL patients reporting at least one affected relative and fewer HL patients doing so. A total of 382 probands reported 505 siblings with cancer; 21·9% of CLL probands reported a sibling with cancer, as compared to 21·1% of NHL probands and 9·8% of HL probands. This difference is probably driven by the younger age of the HL probands. However, for parents, at least one parent with cancer was reported by 53·1% (95% CI 48·4–57·8%) of CLL patients, 46·2% (95% CI 43·4–49·1%) of NHL patients, and 34·3% of HL patients (P = 0·000003 for the three group comparison). This difference between CLL and NHL patients alone was also statistically significant (P = 0·03), indicating a difference not attributable to the age of the probands, which was comparable.
Frequency of affected relatives with lymphoproliferative malignancy
Of these 1948, 183 reported a first-degree relative (alive or dead) with a LPD. Among all probands 9·4% reported a first-degree relative with a LPD, and 13·6% a first or second-degree relative with a LPD (Table II). These rates differed significantly by diagnosis, with 13·3% (95% CI 10·3–16·8%) of CLL probands reporting a first-degree relative with a LPD, 8·8% (95% CI 7·2–10·5%) of NHL probands, and 5·9% (95% CI 3·5–9·3%) of HL probands (P = 0·002). The number of families with two or more affected first-degree relatives with LPDs was small, and included eight with CLL (1·8%), 10 with NHL (0·83%) and none with HL (0%).
Table II. Lymphoproliferative malignancy among first-and second-degree relatives.
This difference in frequency of affected relatives by proband diagnosis was true also for the proportion of those with affected first-or second-degree relatives, with 17·5% (95% CI 14·1–21·3%) of CLL probands, 12·6% (95% CI 10·7–14·6%) of NHL probands and 11·9% (95% CI 8·4–16·2%) of HL probands reporting an affected first-or second-degree relative (P = 0·026). The number of probands with two or more relatives of any type affected with LPDs broke down as follows: CLL [n = 19, 4·2%, including four with three affected relatives (0·9%)]; HL [n = 4, 1·4%, including two with three affected relatives (0·7%)]; NHL [n = 31, 2·5%, including eight with three affected relatives (0·7%)].
The distribution of LPDs among first-degree relatives is shown in Fig 2. Although for both CLL and NHL the same LPD is most commonly seen in the affected first-degree relative, nonetheless there are significant rates of other LPDs. For HL probands, both HL and NHL are seen at high rates in first-degree relatives of the screened population.
Rates of malignancy in parents of probands
Reported prevalences of common solid tumours and lymphoid malignancies were determined in the parents of probands (Fig 3). Lymphoid malignancies appeared relatively enriched, particularly in CLL probands, whose parents had higher rates of lymphoid malignancies than lung or colon cancer. Amongst NHL and HL parents, lymphoid malignancies were more commonly reported than colon cancer.
Although these reported prevalences of parental malignancy were not true lifetime estimates, because many of these individuals were still alive and at risk, nonetheless a comparison to US SEER estimated complete prevalence counts as of 1 January 2004 was performed. These results suggested systematic undercounting of solid tumours (breast, colon, lung, prostate; data not shown), which was probably related to the lack of lifetime data and reduced sensitivity of self-reporting of familial solid tumours among probands with haematopoietic malignancy (Chang et al, 2006). We therefore compared the reported prevalences of these solid tumours in parents of CLL probands to the reported prevalences in parents of NHL probands. Although no significant difference was observed in the rates of breast cancer, prostate cancer or lung cancer, a higher rate of colon cancer was reported in parents of CLL probands (3·9%) than in parents of NHL probands (2·3%) (P = 0·016).
We also calculated the reported prevalences of LPDs in parents of probands with CLL and NHL (Table III), in which the sensitivity of reporting was better (Chang et al, 2006; Glaser et al, 2007). These data showed that the reported prevalences of CLL in parents of CLL probands, even without full lifetime data, were increased compared to SEER lifetime estimates (P < 0·05) and increased compared to parents of NHL probands (P = 0·001 for mothers; 0·007 for fathers). Among parents of NHL probands, the rates were greater for a mother with CLL or a father with NHL, but the 95% CIs did overlap with SEER; the importance of this is unclear given the relatively small number of events and shorter period of observation. A significant difference was observed, however, between the rates of NHL in fathers and mothers of NHL probands, with a greater risk in the fathers (P = 0·026). The numbers of parents with HL in either group was too small to permit adequate analysis.
Table III. Selected estimates of LPD prevalence in parents of probands.
Estimated lifetime risk in parents
Statistically significant findings in this table are italicized. The rates of CLL reported in both mothers and fathers of CLL probands are statistically greater than the rates reported in SEER (P < 0·05). The rate of NHL in fathers of NHL probands is greater than the rate of NHL in mothers of NHL probands (P = 0·026).
CLL, chronic lymphocytic leukaemia; LPD, lymphoproliferative disorder; DFCI, Dana Farber Cancer Institute; HL, Hodgkin lymphoma; SEER, Surveillance, Epidemiology, and End-Results; NHL, non-Hodgkin lymphoma.
Confirmation of reported diagnoses
As many of the patients with self-reported familial LPDs were enrolled on our more detailed cohort tissue banking study, we were able to determine the accuracy of reported lymphoma diagnoses. To date, of 284 enrolled probands and affected family members, 229 have pathology reports available at present, and 96% of those reports confirm the reported diagnosis. Of the remaining 10 pathology reports, nine confirmed lymphoma of a different subtype (3·9% misclassification rate), with only one failing to document lymphoma. Thus, the accuracy of reported diagnoses for the probands and their first-degree relatives was quite high in this patient population.
This study systematically evaluated the frequency of familial LPDs in a consecutive population of affected patients. Our results confirmed the clinical experience that led to the initiation of this study, namely that a significant fraction of affected patients have a relative who also has a LPD. Approximately one in six CLL patients and one in eight lymphoma patients have a first-or second-degree relative with a related LPD. Although awareness of familial predisposition is increasingly common in CLL, this finding in lymphoma is not as widely known.
Our findings are consonant with the increased relative risks of LPDs suggested for first-degree relatives in prior epidemiological studies. As no truly appropriate comparison group exists for our data, it was difficult to correlate our results directly with these relative risk estimates. Similarly, it was difficult to translate those relative risks into easily understandable clinical expectations, which is a strength of our data.
Substantial evidence suggests that familial aggregation of LPDs has a significant genetic component. The early literature in these diseases contains reports of families that have virtually Mendelian inheritance of LPDs, and it is well known that a number of single gene immunodeficiency diseases are associated with high rates of lymphomas. The challenge in the case of lymphoproliferative diseases is defining the phenotype(s) that we expect within families. The high rates of any cancer observed in first-degree relatives in this study suggest a greater susceptibility to cancer in general. The variation in cancer and LPD rates by diagnosis, greater for CLL than NHL and HL, further suggests that the underlying biological predisposition may vary among the diseases. The differences observed between CLL and NHL indicated that this finding is real and not merely an artefact of different proband ages. At the same time, however, the clear crossover of LPDs among families regardless of proband diagnosis strongly suggests that similar genes may predispose to any of these diseases.
Patterns of inheritance may provide some clues to pathogenesis. It is of interest that we observed a higher rate of NHL in fathers of NHL probands than in mothers; this effect is larger than would be expected based on the relative incidence of NHL in men and women. A similar association was recently reported by the InterLymph Consortium (Wang et al, 2007). In contrast, among parents of CLL probands, the relative increase in risk compared to SEER was greater for mothers than fathers, which may reflect the lower baseline risk for mothers. Similarly, Mauro et al (2006) reported that a family history of haematological malignancy is more common among women than men with CLL. These phenotypes, including any differences between CLL and NHL, will need to be clarified in family studies, including our ongoing cohort study of affected families, and through enhanced understanding of the underlying biology.
The major possible weakness of our study is its reliance on self-reported family history data. Self-reported data on familial cancer has the potential for both misstating the frequency of disease and misclassifying disease. For example, we observed probable under-reporting of common solid tumours in parents compared to SEER frequencies. In a previous case–control study of self-reported family history in lymphoma patients, the sensitivity of reporting of common solid tumours ranged from 50% to 70% (Chang et al, 2006), which is not dissimilar from our observations. Reports of the accuracy of lymphoma reporting in similar studies has varied but is significantly better; the above study reported a sensitivity of 85% with 89% specificity for LPDs. A separate study of familial HL reported 100% validation of affected first-degree relatives, but dropped to 81% if more distant relatives were included (Glaser et al, 2007). In our study, the accuracy of reported diagnoses was quite high, with all but one enrolled individual documented to have lymphoma, and only 3·9% of those misclassified by type. These data confirm the validity of our reported frequencies.
This study has focused on a selected patient population seen at a tertiary care referral centre. As a result of that, the observed rates of LPDs in this population may not be fully generalizable to community practice. Nonetheless the observation of a consecutive population of lymphoma patients with this high frequency of familial LPDs is significant, and further study of these patients will prove critical to elucidating the genetic basis of familial LPDs. The genes involved may well prove to be those in which polymorphisms have been associated with lymphoma risk: cytokines, DNA repair, folate metabolism. Systematic genomics approaches applied to large numbers of families are most likely to clarify this question in the coming years.
This work was supported in part by NIH grants K23 CA115682 to JRB and the Okonow-Lipton Fund. ASF is supported in part by NIH grant 2P01CA092625. The authors are indebted to Donnette Bailey and Sharinda Roper, the New Patient Coordinators of the Lymphoma Program, and to the staff of the Dana 1 hematological malignancies clinic, for their assistance with the distribution and collection of the family history questionnaires. We also wish to thank Elke Raderschall PhD for her assistance with sample processing on the more detailed cohort study.