SEARCH

SEARCH BY CITATION

Keywords:

  • epidemiology;
  • Epstein–Barr virus;
  • hepatitis C;
  • HIV;
  • infections;
  • malignant lymphomas

Abstract.

  1. Top of page
  2. Abstract.
  3. Introduction
  4. Infections and Hodgkin lymphoma
  5. Infections and non-Hodgkin lymphoma
  6. Conclusion
  7. Conflict of interest statement
  8. Acknowledgement
  9. References

Lymphomas constitute a heterogeneous group of malignant disorders with different clinical behaviours, pathological features and epidemiological characteristics. For some lymphoma subtypes, epidemiological evidence has long pointed to infectious aetiologies. A subset of Hodgkin lymphoma is strongly linked to Epstein–Barr virus (EBV) infection. In addition, infectious agents can directly infect and transform lymphocytes (e.g. EBV, human herpesvirus 8), induce immunosuppression (human immunodeficiency virus), or cause chronic immune stimulation (hepatitis C virus, Helicobacter pylori), all of which may play a role in the development of various non-Hodgkin lymphoma subtypes. Here, we review the epidemiological evidence linking infections with malignant lymphoma.


Introduction

  1. Top of page
  2. Abstract.
  3. Introduction
  4. Infections and Hodgkin lymphoma
  5. Infections and non-Hodgkin lymphoma
  6. Conclusion
  7. Conflict of interest statement
  8. Acknowledgement
  9. References

Malignant lymphomas make up an estimated 3–4% of all malignancies worldwide [1]. The lymphomas can be divided into two major groups, i.e. Hodgkin (20–30% of all lymphomas) [1,2] and non-Hodgkin lymphomas (HL and NHL respectively). Historically, attempts to understand the aetiology of NHL have been complicated by its heterogeneous composition with numerous subtypes displaying diverse clinical behaviours and histopathological and immunological phenotypes. Over the course of time, attempts have been made to incorporate this heterogeneity into classification systems. Because such classifications have reflected pathological or clinical considerations rather than epidemiological/aetiological characteristics, their application in aetiological research may have been confusing. This situation is in contrast to HL, for which characteristic epidemiological patterns long have suggested a modest number (two or three) of subtypes and an infectious aetiology. Because of these differences, HL and NHL have traditionally been studied separately.

Infections and Hodgkin lymphoma

  1. Top of page
  2. Abstract.
  3. Introduction
  4. Infections and Hodgkin lymphoma
  5. Infections and non-Hodgkin lymphoma
  6. Conclusion
  7. Conflict of interest statement
  8. Acknowledgement
  9. References

The suspicion of an infectious aetiology to HL is almost as old as the recognition of the disease as such [3]. Accordingly, both the disease’s typical histological presentation dominated by inflammatory cells and its typical clinical presentation with sudden onset of fever, night sweats and lymphadenopathy would be compatible with an infectious process.

As mentioned, the clinical suspicion of an infectious aetiology has been supported by epidemiological evidence. A hallmark of HL epidemiology is the bimodal age-specific incidence pattern that can be observed throughout the Western World (Fig. 1) [4]. Different from the age-dependent monotonic incidence increase seen for NHL, this remarkable age distribution fostered the suggestion that HL in children, younger adults and older adults were aetiologically distinct disease subtypes [4, 5]. More importantly, the typical histological presentation (nodular sclerosis HL), a more benign clinical behaviour than HL in older adults, together with the impression of synchronicity of onset of familial cases, led to the proposition of an infectious aetiology for HL in younger adults [5].

image

Figure 1.  Age-specific incidence of Hodgkin (HL, broken line) and non-Hodgkin lymphoma (NHL, solid line) in the United States, females and males combined. Data come from the National Cancer Institute’s Surveillance, Epidemiology, and End Results (SEER) Program (http://www.cancer.gov), SEER9 registries, 1973–2005. Note the different vertical scales for Hodgkin (left axis) and non-Hodgkin lymphomas (right axis).

Download figure to PowerPoint

In 1971, an international HL incidence survey indicated that the bimodal age distribution in the Western World constituted one extreme of a dynamic range of incidence patterns [6]. Specifically, data suggested that HL incidence patterns correlated with the underlying population’s level of socio-economic development (Fig. 2). In developing regions, HL incidence would be relatively high in boys but low in young adult men, as opposed to developed regions where HL incidence was low in children but high in young adults [6]. Other data suggested that HL in developing regions was predominantly of the mixed cellularity or lymphocyte depletion subtypes, and in developed regions predominantly of the nodular sclerosis subtype [6]. The ecological correlation between socio-economic level and HL incidence patterns suggested opposite roles of childhood environment to HL risk in children and younger adults [6]. One model suggested HL to be a rare consequence of a common infection [7], the risk of which would increase when age at infection was delayed, e.g. by improving living conditions [8].

image

Figure 2.  Age-specific incidence rates per 100 000 persons in different geographical areas. Reproduced from Correa and O’Conor Int J Cancer, 1971; 8: 192–201. Reprinted with permission of Wiley-Liss Inc., a subsidiary of John Wiley & Sons Inc. [6].

Download figure to PowerPoint

Evidence to support that childhood socio-economic environment influences the risk of HL in young adulthood has been provided from several studies. For instance, in an early U.S. case–control study, socio-economic affluence in childhood, as reflected in correlates such as type of housing, sibship size, maternal education and paternal social class, was clearly associated with an increased risk of HL in young adulthood [9]. The role of family structure has been further emphasized by Scandinavian register-based studies. In these, the risk of HL in younger adults correlated inversely with birth order [10] or with the number of older siblings [11]. In classical epidemiology, birth order and sibship size have been considered important for infectious disease pressure in early life [10]. Specifically, firstborn children are believed to be exposed to infections at later ages than their younger siblings, who may contract infections from their older siblings. Likewise, the number of children in a sibship has been shown to correlate with risk of infections in childhood [10]. Accordingly, for HL in young adults the observed associations would be consistent with an infectious aetiology, HL risk being associated with delayed infection perhaps with a commonly circulating virus.

Another and perhaps more direct indication of an infectious aetiology to HL is the increased occurrence of HL described in immune-deficient individuals, in particular persons with acquired immunodeficiency syndrome (AIDS) due to human immunodeficiency virus (HIV) infection [12], but presumably also other groups of immune-compromised patients [3].

Epstein–Barr virus (EBV) is among the infectious agents whose clinical manifestations vary with age at primary infection. Moreover, it behaves epidemiologically analogously to HL. When primary infection with the virus occurs in early childhood, as is typical of developing countries, it is accompanied by, few if, any symptoms. If primary infection in contrast is delayed to adolescence, as often is seen in industrialized countries, it is associated with the clinical syndrome, infectious mononucleosis [13].

That HL could be related to EBV infection has been explored almost since the discovery of the virus in 1964 [14]. First, a wealth of cohort and case–control studies has shown that mononucleosis is followed by an increased risk of HL. In the largest study to date, a Scandinavian cohort study of 38 000 mononucleosis patients, mononucleosis was associated with a more than 2.5-fold increased HL risk, which although it decreased with time remained significantly elevated for up to two decades. Moreover, because mononucleosis typically occurs in adolescence, HL risk was particularly high, 3.5-fold increased, in young adults [15].

Epstein–Barr virus has also been linked with HL in serological investigations. Hence, elevated levels of anti-EBV antibodies have repeatedly, although not unanimously, been demonstrated in HL patients [16]. Even more compellingly, elevated levels of EBV antibodies in healthy persons are associated with an increased risk of later HL [17].

Whilst the above evidence for long had indicated that EBV was associated with HL development, it was not until 1987 that Weiss et al. succeeded in demonstrating EBV in the tumour’s malignant Hodgkin/Reed-Sternberg cells [18]. The presence of EBV in the malignant Hodgkin/Reed-Sternberg cells, the vast majority of which are of B cell origin [2], is consistent with the normal course of EBV infection. Accordingly, following primary infection, EBV persists in a latent state in memory B cells for the lifetime of the host. The possibility that EBV can cause lymphoma is supported by laboratory evidence, including the fact that EBV readily transforms B lymphocytes in vitro through the expression of various EBV proteins, among which latent membrane protein 1 (LMP-1) is thought to be the most important.

In healthy infected individuals, EBV replication and the proliferation of EBV-transformed B cells is prevented by the presence of intact T cell immunity [19]. Therefore, the increased HL risk in immune-deficient individuals could be explained with the loss of this surveillance, and indeed HL in immune-suppressed patients typically harbour EBV [20]. Importantly, however, EBV in not ubiquitously present in HL occurring in apparently otherwise healthy individuals [21] (Figs 3 and 4).

image

Figure 3.  Percentage of Hodgkin lymphomas positive for EBV by 5-year age group. Reproduced from Glaser et al. Int J Cancer, 1997; 70: 375–82. Reprinted with permission of Wiley-Liss Inc., a subsidiary of John Wiley & Sons Inc. [21].

Download figure to PowerPoint

image

Figure 4.  Distribution of Epstein–Barr virus-positive tumours by histological subtype. Reproduced from Glaser et al. Int J Cancer, 1997; 70: 375–82. Reprinted with permission of Wiley-Liss Inc., a subsidiary of John Wiley & Sons Inc. [21].

Download figure to PowerPoint

The absence of EBV in 60–70% of HL in general, and in particular the low prevalence in HL in younger adults, the age group where HL is most strongly associated with childhood socio-economic affluence and mononucleosis, has questioned the role of EBV in HL development [22]. However, when present the virus is monoclonal in Hodgkin/Reed-Sternberg cells, indicating that the infection occurred before the malignant transformation was completed [18]. Moreover, in addition to the findings in immune-deficient patients, there is mounting evidence that other markers linking EBV infection with HL risk are specific to the EBV-positive HL subgroup. Characterization of cases observed in the Scandinavian mononucleosis cohort [15] showed the increased HL risk to be specific to EBV-positive HL [23] with no increased risk of EBV-negative HL (Fig. 5). Likewise, in other investigations, mononucleosis has been found to be associated with an increased risk specifically of EBV-positive HL, with no [24, 25] or a less increased [26] risk seen for EBV-negative HL.

image

Figure 5.  Relative risk of Epstein–Barr virus (EBV)-positive and EBV-negative Hodgkin lymphoma after infectious mononucleosis. Blue lines represent the relative risks of EBV-positive and red lines EBV-negative Hodgkin lymphoma after infectious mononucleosis in different estimation methods. Reproduced with permission after Hjalgrim et al. N Engl J Med, 2003; 349: 1324–32. Copyright® 2003 Massachusetts Medical Society. All rights reserved. [23].

Download figure to PowerPoint

The suggested differences in risk profiles between EBV-positive and EBV-negative HL suggest that tumour EBV status may be a more appropriate marker for aetiologically diverse HL subtypes than age and histological subtypes (Fig. 4). Additional support for this notion comes from genetic investigations, showing that markers in the HLA-I region (HLA-A1*01) carry a significantly increased risk of EBV-positive HL and markers in the HLA-III region an increased risk of EBV-negative HL [27, 28].

According to this view on HL, two major subtypes of HL exist – one related to EBV infection and one not [29, 30]. The high prevalence of EBV-positive HL in childhood might be attributed to primary EBV infection, and in older adults to loss of immunological control of latent infection (Fig. 3). The relatively few cases of EBV-positive HL in younger adults (Fig. 3) may be attributed to delayed primary EBV infection as suggested by the association with mononucleosis. Of note, the genetic markers associated with risk of EBV-positive HL were recently found to be associated with an increased risk of mononucleosis upon primary EBV infection, further supporting an association between the two diseases [31]. This model leaves the cause(s) for the majority of HL in younger adults, i.e. EBV-negative HL, to be identified. Although more recent studies have been unable to demonstrate previous associations with childhood socio-economic environment [24, 32, 33] for (EBV-negative) HL, this may reflect that secular changes in society have reduced the importance of family structure for infectious disease transmission. Based on earlier studies an infectious aetiology therefore remains likely also for this subgroup of HL, although as yet no strong candidates have been identified [30].

Infections and non-Hodgkin lymphoma

  1. Top of page
  2. Abstract.
  3. Introduction
  4. Infections and Hodgkin lymphoma
  5. Infections and non-Hodgkin lymphoma
  6. Conclusion
  7. Conflict of interest statement
  8. Acknowledgement
  9. References

As already mentioned, NHL comprises numerous distinct entities, whose epidemiology may differ considerably. Importantly, these epidemiological differences often reflect different aetiologies and different natural histories. Infectious causes of NHL therefore have to be considered in the context of NHL subtypes.

Today, several infectious agents are known or suspected to cause malignant lymphoma. The agents may be classified according to mechanism into three broad groups (Table 1). First, some lymphocyte-transforming viruses can infect lymphocytes and, by disrupting normal cell functions, directly promote cell division. Secondly, a mechanism, unique to HIV infection, is induction of immune suppression through profound depletion of CD4+ T lymphocytes, leading to the condition known as AIDS. AIDS is associated with a markedly elevated risk for several types of lymphoma, including in particular high-grade B cell NHLs. Thirdly, some infections increase NHL risk by causing chronic immune stimulation and persistent activation of lymphocytes.

Table 1.   Infections associated with non-Hodgkin lymphoma
Infectious agentLymphoma subtype
  1. NHL, non-Hodgkin lymphoma; AIDS, acquired immunodeficiency syndrome; CNS NHL, central nervous system non-Hodgkin lymphoma; DLBCL, diffuse large B cell lymphoma; MCD, multicentric Castleman disease; MALT, mucosa-associated lymphoid tissue.

Lymphocyte-transforming viruses
 Epstein–Barr virusBurkitt lymphoma
AIDS-associated NHLs (especially CNS NHL, DLBCL)
Posttransplant lymphoproliferative disorder
Extranodal NK/T-cell NHL
 Human herpesvirus 8Primary effusion lymphoma and related DLBCLs
MCD-associated plasmablastic NHL
 Human T lymphotropic virus type IAcute T-cell leukaemia/lymphoma
Agents that cause immunosuppression
 Human immunodeficiency virusAIDS-associated NHLs
Agents that cause chronic immune stimulation
 Plasmodium falciparumBurkitt lymphoma
 Hepatitis C virusDLBCL, lymphoplasmacytic NHL, marginal zone NHL
 Hepatitis B virusUncertain
 Helicobacter pyloriGastric MALT NHL
 Campylobacter jejuniSmall intestine MALT NHL
 Chlamydia psittaciOcular adnexa MALT NHL
 Borrelia burgdorferiCutaneous MALT NHL

As discussed in connection with HL, EBV is an example of a transforming virus and in addition to HL it has been implicated in several forms of NHL [16]. An association between EBV and Burkitt lymphoma dates to the initial discovery of EBV in Burkitt lymphoma cells [14]. The incidence of Burkitt lymphoma is especially high in sub-Saharan Africa, where the tumour characteristically presents in young children as a rapidly growing extranodal mass (‘endemic’ Burkitt lymphoma). Burkitt tumour cells demonstrate a characteristic t(8;14) translocation which activates the c-myc oncogene. The highly consistent detection of EBV DNA and proteins in tumour cells implicates EBV in endemic Burkitt lymphoma. In addition, EBV is itself monoclonal in tumour cells, suggesting that the initial EBV infection occurs in a single precursor B cell as an early tumour-inducing event [34]. Furthermore, African children who develop Burkitt lymphoma manifest elevated antibody titres against the EBV capsid several years prior to diagnosis, consistent with a role for dysregulated immunity and uncontrolled viral replication in the development of this malignancy. Based on the geographical distribution of endemic Burkitt lymphoma, malaria caused by Plasmodium falciparum is also postulated to play a key role in the development of this tumour by causing chronic immune stimulation or, perhaps, immunosuppression [35] (Table 1).

Epstein–Barr virus is also strongly implicated in other subtypes of NHL that occur in congenital or acquired immunodeficiency states. In organ transplant recipients, immunosuppression induced by medications used to prevent graft rejection leads to loss of control of EBV infection and a spectrum of EBV-driven lymphoproliferation, ranging from hyperplasia to frank NHL [36]. In addition, the occurrence of primary EBV infection after organ transplant, as seen in children, is associated with poor control of EBV and markedly high risk for EBV-related NHL. Transplant-associated NHLs, when EBV-positive, express a broad range of EBV proteins, including LMP-1. In addition, EBV is almost always found in cases of extranodal NK/T-cell NHL [37], a rare NHL subtype.

Also noteworthy are two other transforming viruses, human herpesvirus 8 (HHV8) and human T lymphotropic virus type I (HTLV-I) (Table 1). HHV8 is evolutionarily related to EBV and is the causative agent of Kaposi sarcoma. HHV8 is found in all cases of primary effusion lymphoma (PEL), a rare variant of NHL that arises in serous body cavities [38]. HHV8 is also closely linked with multicentric Castleman disease. Patients with multicentric Castleman disease have a high risk of developing plasmablastic NHLs, which are frequently HHV8 positive [39]. Although HHV8 infection is highly prevalent in sub-Saharan Africa, HHV8-associated NHLs appear to be uncommon in this region of the world [40]. HTLV-I, a retrovirus, is the established cause of adult T-cell leukaemia/lymphoma [41].

Human immunodeficiency virus causes NHL by inducing severe cell-mediated immunodeficiency, which permits dysregulated activation and proliferation of B cells. EBV is frequently involved in AIDS-associated NHLs. Three NHLs are characteristic of AIDS: central nervous system NHL (CNS NHL, almost universally EBV-positive), diffuse large B cell lymphoma (DLBCL, ∼50% EBV-positive) and Burkitt lymphoma (∼30% EBV positive) [16]. In CNS NHL and DLBCL, tumour cells usually express EBV LMP-1 [42]. People with AIDS have substantially elevated risk for these subtypes of NHL, with relative risks compared with the general population of 5000 for CNS NHL, 100–140 for DLBCL and 60 for Burkitt lymphoma [43]. In AIDS, risk for CNS NHL and DLBCL increases linearly as the CD4 lymphocyte count declines [44]. Furthermore, risk of AIDS-associated NHL has declined markedly in recent years with the availability of increasingly effective HIV therapies that partially restore immune function [43, 45].

Human herpesvirus 8 is also implicated in some subtypes of AIDS-associated NHL. Although PEL is uncommon in people with AIDS, risk for this subtype is increased compared with the general population, highlighting the importance of HIV-induced immunosuppression in permitting HHV8-driven lymphoproliferation. Among people with AIDS, Kaposi sarcoma is associated with an increased risk of developing immunoblastic DLBCLs (related histologically to PEL, but not involving body cavities) [46], and HHV8 can be detected in a subset of these NHLs [47, 48]. Of final note, an elevated risk of T cell NHL of various subtypes is also observed among persons with AIDS [49].

An example of an infectious agent implicated in causing NHL through chronic immune stimulation is hepatitis C virus (HCV) [50]. HCV is usually acquired through blood-borne exposures, particularly injection drug use or (before 1990) blood transfusion, and prevalence in the general population of most western nations is typically less than 5%. HCV produces chronic hepatitis and viraemia following primary infection. Ongoing HCV infection is associated with the development of essential mixed cryogrobulinaemia, a low-grade lymphoproliferative disorder that can progress to NHL [51]. HCV may induce lymphoproliferation by binding to the CD81 receptor on the surface of B lymphocytes [52], which could lower the threshold for antigen response or induce DNA mutations. Furthermore, HCV-infected individuals have an elevated prevalence of circulating lymphocytes with abnormal chromosomal translocations associated with NHL, e.g. t(14;18) [53].

The potential association between HCV infection and NHL has been examined in many retrospective case–control studies [54]. Most of these studies, although not all, have found a positive association. Notably, however, there has been substantial heterogeneity in the magnitude of association across studies, with odds ratios ranging from 2 to 10 [54], and the strength of the association has appeared to diminish in more recent studies. It has been difficult to determine the degree to which this variation in results across studies can be attributed to demographic differences in study populations or, instead, methodological differences in the studies (e.g. choice of control subjects). A recent large cohort study of U.S. military veterans (including 146 394 HCV-infected individuals and 572 293 uninfected individuals) demonstrated that HCV infection precedes the development of NHL and is associated with an increased NHL risk over a period of at least 7 years [55]. Nonetheless, the overall magnitude of association between HCV and NHL was small in that study (hazard ratio 1.28, 95% CI 1.12–1.45). HCV infection was also associated with an increased risk of Waldenstrom macroglobulinaemia, a lymphoproliferative condition related to lymphoplasmacytic lymphoma [55].

The wide range in magnitude of associations between HCV and NHL documented in various studies suggests a need to consider NHL subtypes separately, but whether HCV is associated with an increased risk of all NHL subtypes or only certain subtypes is an incompletely resolved question. A pooled analysis of data from seven case–control studies showed associations between HCV and DLBCL, lymphoplasmacytic NHL and marginal zone NHL [56]; however, heterogeneity in the associations with HCV was observed even for some individual NHL subtypes. Of interest, among HCV-infected patients with splenic marginal zone NHL, treatment of the HCV infection with α-interferon-based regimens can lead to resolution of HCV and, simultaneously, regression of NHL [57]. This observation suggests that HCV directly contributes to lymphoproliferation in at least a subset of NHL cases. Finally, chronic hepatitis B virus infection may also be associated with increased NHL risk [58, 59], although data are fewer than for HCV.

Mucosa-associated lymphoid tissue (MALT) NHLs arise in extranodal aggregates of lymphocytes associated with sites of chronic inflammation [60, 61]. An emerging body of evidence links certain bacteria that cause localized infection and chronic inflammation with MALT NHLs at those sites (Table 1). The strongest data associate gastric MALT NHL with chronic gastritis caused by Helicobacter pylori. Helicobacter pylori seroprevalence is higher among individuals with gastric MALT NHL than among controls. Furthermore, in many NHL cases, antibacterial therapy directed at eradication of H. pylori can lead to regression and remission of the associated tumour [61]. Recent studies have linked three other site-specific MALT NHLs with bacteria causing chronic local infections: small intestine NHL with Campylobacter jejuni, ocular adnexa NHL with Chlamydia psittaci and cutaneous NHL with Borrelia burgdorferi [62–64]. Because these NHL subtypes are rare, it has been difficult to confirm and extend these initial findings.

Of the three mechanisms outlined above (i.e. lymphocyte transformation, immunosuppression, chronic immune stimulation), it is clear that the first two can act synergistically to greatly increase NHL risk as discussed. On the other hand, immunosuppression does not appear to amplify the effects of infections that act through chronic immune stimulation to cause NHL, and indeed, there may be antagonism. For example, the risk of AIDS NHL is not higher in HIV risk groups with elevated HCV prevalence (e.g. injection drug users) than in HIV risk groups with low HCV prevalence (e.g. homosexual men) [65]. This lack of association between HCV and NHL among HIV-infected persons has also been demonstrated in case–control studies [66]. Analogously, HCV infection is not associated with an increased risk of NHL among immunosuppressed solid organ transplant recipients [67]. Furthermore, MALT NHLs are infrequently recognized in HIV-infected persons and are not considered a major component of the spectrum of AIDS NHLs. These observations suggest that a minimum level of host immune competence may be required to respond to chronic HCV and bacterial infections in a manner that leads to the related NHL subtypes.

Because there are several mechanisms by which infectious agents may cause NHL, and a diversity of proposed agents, it is useful to consider criteria that can be utilized to assess the evidence that these associations are due to causal relationships. One set of criteria, proposed by Bradford Hill, is built upon an epidemiological perspective and has been utilized to evaluate a wide variety of noninfectious exposure–disease associations [68]. A second set, proposed by Fredericks and Relman [69], takes a molecular perspective and is aimed specifically at determining whether detection of infectious agents in diseased individuals implicates the agents in the aetiology of the disease. Table 2 lists these criteria and applies them to illustrative examples of associations between infectious agents and NHL.

Table 2.   Criteria for an aetiological association between an infectious agent and non-Hodgkin lymphoma
Criterion for causalityCommentMicrobial agent and evidence for causality
  1. HIV, human immunodeficiency virus; HCV, hepatitis C virus. Criteria are derived from Bradford Hill [68] and Fredericks and Relman [69]. A ‘+’ sign indicates that there is evidence supporting an aetiological association between the infection and non-Hodgkin lymphoma based on the specified criterion.

Bradford HillHIVHCV
 1. Biological plausibilityIn vitro, molecular, or animal data++
 2. AnalogyComparison with other disease models++
 3. CoherenceLack of conflict with other information++
 4. Specificity∼1 : 1 correspondence of agent to disease+ 
 5. ConsistencyRepeated demonstration across studies+ 
 6. Strength of associationMagnitude of relative risk+ 
 7. Biological gradientDose–response relationship+ 
 8. TemporalityExposure precedes disease++
 9. Experimental evidenceHuman experiment  
Fredericks and RelmanBorrelia burgdorferiChlamydia psittaci
 1. Molecular detection in diseaseMicrobial sequences in host tissues++
 2. LocalizationMicrobe localized within tumour++
 3. Only low level detection in the     absence of diseaseQuantitative testing++
 4. Resolution/relapseDetection parallels disease +
 5. TemporalityDetection precedes disease  
 6. CoherenceMicrobe characteristics consistent with disease++
 7. ReproducibilityRepeated demonstration across studies  

Both sets of criteria for causality emphasize the overall plausibility of an aetiological association, based on a range of outside considerations, and the compatibility of the association with what is known from other settings, such as from laboratory experimentation or findings for similar agents and diseases. Both sets of criteria also require that the observations linking the agent to NHL be reproducible in multiple study settings and across different groups of investigators. This criterion of reproducibility has not been met for some of the associations discussed herein, in part because the associations have been proposed recently and, in part, because the implicated NHL subtypes are rare and thus hard to study.

In taking an epidemiological perspective, Bradford Hill further argued that the strength of the association (i.e. the magnitude of the relative risk associated with the exposure) should be large if the relationship is to be considered causal. Along these lines, a difficulty with concluding that HCV causes NHL has been that the magnitude of the association is variable across studies and somewhat modest in the most recent studies (i.e. there is a lack of consistency and strength of association) [50].

Because NHL is a heterogeneous malignancy in terms of molecular characteristics, treatment response and clinical outcome, one might expect that particular infections cause unique NHL subtypes. The expectation is formalized under the Bradford Hill criteria as specificity, i.e. that one agent causes one or only a few specific outcomes. Such a close correspondence would occur if an agent has a specific molecular mode of action or if the agent’s effects are restricted to the site of infection (e.g. B. burgdorferi and cutaneous MALT NHL). A difficulty with concluding that the association between HCV and NHL is causal is that it has been uncertain whether there are unique NHL subtypes related to HCV infection, although recent data provide some support for such specificity [56]. Alternatively, one might argue that infectious agents that broadly affect the immune system (e.g. HIV and HCV) could plausibly increase the risk for many NHL subtypes. Thus, whilst specificity of association between agent and NHL subtype supports an aetiological relationship, it should not be viewed as necessary.

The Bradford Hill criteria also include the presence of a biological gradient (i.e. a dose–response relationship of increasing risk with increasing exposure). Although it may be difficult to define the ‘dose’ of an infection, a possible approach to incorporate this criterion for viral infections would be to consider viral load (i.e. the level of virus in plasma), in order to assess whether, for example, high levels of viral replication are associated with the highest NHL risk. Similarly, the aetiological nature of the association between HIV infection and NHL is supported by the clear dose–response relationship between the degree of HIV-induced immunosuppression (as measured by falling CD4 count) and increasing NHL risk, and by the decline in NHL risk associated with effective HIV therapies that reduce viral replication.

The Fredericks and Relman criteria are readily applied to evaluate associations involving transforming viruses and MALT agents, where it is postulated that microbes are present within tumour cells or the surrounding tissues and should therefore be detectable. According to Fredericks and Relman, an aetiological explanation becomes increasingly plausible based on the weight of evidence that the infection is present in all cases of the implicated NHL subtype, that the microbe is consistently found within the tumour tissue and that infection is absent or rare in individuals without NHL. Fredericks and Relman also suggest that a response of NHL paralleling treatment directed at the infection (along with relapse of NHL with any reappearance of the microbe if such treatment fails) provides additional supporting evidence.

Conclusion

  1. Top of page
  2. Abstract.
  3. Introduction
  4. Infections and Hodgkin lymphoma
  5. Infections and non-Hodgkin lymphoma
  6. Conclusion
  7. Conflict of interest statement
  8. Acknowledgement
  9. References

Since the first description of EBV in a Burkitt lymphoma cell culture more than 50 years ago [14], the list of infectious agents aetiologically linked with lymphoma has continuously increased. Based on current understanding, it has been estimated that infections annually account for more than 85 000 malignant lymphoma cases worldwide, or nearly 25% of all cases [70]. For some lymphoma subtypes, associations with infectious agents have been established guided by strong epidemiological evidence. Aetiological research, including epidemiological studies, should adequately address this diversity. With increasing accessibility to advanced molecular laboratory techniques, the number of infectious agents linked to lymphoma development is likely to continue to rise, potentially increasing as a corollary the opportunities for both prevention and treatment.

References

  1. Top of page
  2. Abstract.
  3. Introduction
  4. Infections and Hodgkin lymphoma
  5. Infections and non-Hodgkin lymphoma
  6. Conclusion
  7. Conflict of interest statement
  8. Acknowledgement
  9. References