In people with HIV infection, there have been few studies of risk factors for non-Hodgkin's lymphoma (NHL) and little is known of its aetiology. The one apparent strong risk factor is degree of immune deficiency. NHL incidence is 60- to 100-fold higher among people with HIV infection than in people without HIV infection, and increases markedly as immune function declines [1–3]. For the other common AIDS-associated cancer, Kaposi's sarcoma, epidemiological analyses pointed towards a sexually transmissible agent, later discovered to be human herpesvirus eight, as the likely cause [4,5]. In contrast, analyses of NHL occurrence in people with AIDS have not revealed any distinct epidemiological patterns .
Several aetiological hypotheses have been proposed to explain the increased risk of NHL in people with immune deficiency. DNA sequences of Epstein-Barr virus (EBV) are found in tumour tissue in nearly 100% of specimens of primary central nervous system NHL, in about 80% of systemic large cell-immunoblastic NHL, and in 30 to 50% of Burkitt's lymphoma occurring in people with HIV infection . Another factor may be increased B-cell stimulation caused by HIV antigens, leading to continuous B-cell division, which may increase the probability of a malignant clone developing [8–10]. Other risk factors for NHL in people with AIDS that have been suggested on the basis of epidemiological comparisons, include use of zidovudine  and a previous diagnosis of Kaposi's sarcoma [12,13]. This report describes a study of risk factors for AIDS-related NHL (AIDS-NHL) in people with HIV infection in Sydney, Australia.
We performed a matched case–control study of risk factors for NHL in people with HIV presenting for clinical care in eastern and central Sydney. Matching was performed for degree of immune deficiency, the strongest known risk factor for AIDS-NHL. Each control was required to have a CD4 positive cell count within the same stratum as the corresponding case at NHL diagnosis (0–19, 20–99, 100–199, 200–499 and more than 500 CD4 positive cells × 106/l) within 6 months of the date of diagnosis of NHL (the `match date'). The study protocol was reviewed and approved by the ethics committees of the New South Wales Cancer Council and the institutions at which the study was conducted.
Multiple overlapping sources were used to identify cases of AIDS-NHL among people with HIV infection who resided in the study area. Non-residents were included if they would have normally been referred for HIV-related illness to hospitals in the study area. Three hospitals, which represent all of the public institutions caring for people with AIDS in central and eastern Sydney, and their outpatient clinics were included in addition to the two largest private AIDS clinics. Participating centres generated lists of patients with AIDS-NHL based on International Classification of Disease (version 9) rubrics  coded on hospital discharge and other patient summaries. Outpatient diagnosis lists were searched to identify cases of AIDS-NHL in people who were not admitted to hospital. The extent of under-ascertainment of AIDS-NHL was assessed by comparing cases identified through these mechanisms with cases identified by linking the New South Wales AIDS and cancer registers for the period 1985–1992 . The CD4 positive cell count at or within 3 months of diagnosis of NHL was abstracted from medical records and used for matching to controls.
Potential controls were people with HIV infection but without diagnosed NHL, who resided in the study area or received regular HIV-related medical care there. A list was compiled of 5919 potential controls who had ever been seen at the hospitals and clinics that reported the NHL cases. Subjects were randomly selected from this list, and all CD4 positive cell counts for these patients were recorded from computerized and paper records at each study site. These potential controls were then matched with a case who fulfilled the CD4 positive cell count and match date criteria, or were discarded if there was no such match.
Data were collected by a trained medical records administrator or medical practitioner using a standardized data collection instrument. For each patient, data were collected relating to the time period beginning at HIV diagnosis and ending at the match date. For laboratory variables, values at 1, 2, 3, 4 and 5 years prior to match date were also sought. Such data was only collected if a value within 6 months of the target date was available. All data abstraction involving a diagnosis of a HIV-related condition was reviewed by a medical practitioner (A.G.).
A date of seroconversion was recorded if there had been a clinical diagnosis of a seroconversion illness or a clear statement of a period of less than 1 year of HIV-risk exposure reported in the medical record. Laboratory results were based on values recorded in the medical records.
Serum globulin levels were used as a surrogate marker of B-cell stimulation. This was because a large proportion of subjects did not have a recorded serum immunoglobulin measurement, and more than 50% of total globulin in normal individuals is immunoglobulin . Serum globulin was calculated as the difference between recorded total protein and albumin values.
Frequency of hypothesized risk factors was tabulated for cases and controls. Differences between cases and controls were tested using t tests for continuous variables and χ2 tests for categorical variables. Matched odds ratios were calculated. As unmatched odds ratios allowed inclusion of more data for those variables with missing data and gave very similar results, unmatched odds ratios are presented as the primary outcome of this study (matched odds ratios are available on request from the authors).
Risk associated with anti-herpes and antiretroviral agents was assessed by comparing cases and controls with respect to total doses received in the past year (anti-herpes agents) or ever (antiretroviral agents) and total days on combination therapy (zidovudine plus either of didanosine or zalcitabine).
Multivariate analysis (unconditional logistic regression) was carried out using a forward stepwise technique. Variables with few missing values that were significant in univariate analyses were added first and others were added in order of statistical significance. Variables were retained only if they attained statistical significance for difference from an odds ratio of 1 (dichotomous variables) or had a significant test for trend (ordered categorical variables with more than two levels). Variables that showed univariate relationships with NHL risk but had substantial missing data were then adjusted for the significant variables.
Two hundred and twenty-one cases of NHL were identified, and a CD4 positive cell count within 3 months of diagnosis was available for 219, who were matched with 219 controls. Matching for date of diagnosis and CD4 positive cell count was successful. The mean CD4 positive cell count was 137 × 106 cells/l in cases and 144 × 106 cells/l in controls (P = 0.66). CD4 count ranged from zero to 1265 × 106 cells/l in cases and zero to 1470 × 106 cells/l in controls (medians of 50 × 106 and 56 × 106 cells/l respectively).
Demographic risk factors
Subjects with NHL were significantly older than controls (means of 40.0 versus 38.2 years respectively, P = 0.05). Over 97% of cases and controls were male, and the predominant reported mode of HIV acquisition was male homosexual contact [94% of cases and 89% of controls, odds ratio (OR), 2.04; 95% confidence interval (CI), 0.97–4.35].
Sexually transmissible diseases
More than 90% of both cases and controls had a recorded history of at least one sexually transmissible disease other than HIV. Neither seropositivity to syphilis, hepatitis A or B nor clinical history of sexually transmissible diseases (non-specific urethritis, gonorrhoea, genital herpes, and genital warts) was significantly related to NHL risk (data not shown).
Clinical history of HIV-related diseases
An AIDS-defining illness was recorded prior to the match date in 46% of cases and 44% of controls (OR, 1.10; 95% CI, 0.74–1.63). Of 11 HIV-associated illnesses occurring in more than 20 study subjects, only Mycobacterium avium complex disease (MAC) and non-Pneumocystis carinii pneumonia had occurred significantly more often in cases than controls (OR, 3.18; 95% CI, 1.18–9.96 and OR, 2.33; 95% CI, 1.17–4.70; respectively). Prior oral hairy leukoplakia was not related to NHL risk (OR, 0.85; 95% CI, 0.52–1.38). No HIV-associated illnesses occurred significantly less often in cases than controls.
Anti-herpesviral and antiretroviral treatment
There was no significant relationship between the amount of acyclovir, ganciclovir or foscarnet received and risk of NHL (Table 1). Foscarnet had been used prior to match date by less than 2% of cases and controls. Of the antiretrovirals, zidovudine had been used by 72% of cases and 71% of controls, didanosine by 24 and 25%, respectively, and zalcitabine by 16 and 14%, respectively. There was no relationship between total dose of each antiretroviral received and risk of NHL. Combination nucleoside antiretroviral therapy had been received by 13% of cases and 18% of controls (OR, 0.68; 95% CI, 0.34–1.18).
Duration of immune deficiency
Cases were 3.56 times (95% CI, 1.14–11.4) more likely to have been infected with HIV for 8 or more years (Table 2). A CD4 count approximately 1 year prior to diagnosis was available in 190 cases and 204 controls, and was collected at a median of 365 days prior to match date in controls and 353 days prior to match date in cases (P = 0.102). This CD4 count was markedly lower in cases than controls (P for trend 0.003, Table 2), and CD4 count was lower in cases than controls for several years prior to diagnosis (Fig. 1).
Correlation between serum globulin and IgG
Only 32% of cases and 26% of controls had an IgG measurement available at match date, compared with 99% of cases and 92% of controls that had serum globulin measurements available. In those subjects for whom both serum globulin and IgG was available, globulin was correlated with immunoglobulin G (IgG) in both cases and controls (Fig. 2, pooled r = 0.57, P < 0.001).
Serum IgG, globulin, and p24 antigen
The risk of NHL rose with increasing levels of serum globulin at NHL diagnosis (P for trend = 0.015, Table 3). Mean globulin levels were higher in cases than controls in each of the 5 years prior to match date (Fig. 3). IgG levels at match date were available in 71 cases and 56 controls, and higher levels of serum IgG were associated with increased NHL risk (P for trend 0.036, Table 3).
Cases were significantly more likely to be HIV-p24-antigen positive at match date than controls (P for trend 0.0001, Table 3) and this was also the case during each of the 3 years prior to diagnosis of NHL (data not shown).
Multivariate analysis of NHL risk
Variables significantly related to NHL risk in the multivariate analysis were lower CD4 count 1 year prior to NHL diagnosis and serum globulin at match date (Table 4). When those variables with substantial missing data were individually adjusted for previous CD4 count and serum globulin at match date, a significant effect remained for longer time since seroconversion and p24 antigenaemia at NHL diagnosis (Table 5).
This study has demonstrated two previously unreported risk factors for AIDS-NHL: prolonged immune deficiency (as reflected by lower CD4 count prior to diagnosis, and longer period since HIV infection), and chronic B-cell stimulation (as indicated by raised serum globulin and HIV p24 antigenaemia). The study is only the second large-scale case–control study of risk factors for AIDS-NHL, and the first to closely control for the major risk factor for NHL, immune deficiency [3,11, 12,17,18]. Clinical indicators of EBV infection, clinical and serological indicators of sexually transmissible diseases, and use of nucleoside analogue antiretroviral therapy were not associated with NHL risk.
It is unlikely that biases introduced by the study design can explain the patterns seen. Cases and controls were closely matched for immune deficiency at time of NHL diagnosis, and for period of diagnosis, so differences in immune status at match date and temporal changes in HIV disease presentation have probably not influenced the results. Selection bias is unlikely to have had a major effect. For cases, we verified that all cases of NHL in people with HIV infection in the study area that were registered with the NSW cancer register by the end of 1992 were included in this study . For controls, there was no identifying register of HIV-positive individuals, but the number of potential controls identified (5919) was of a very similar magnitude to the estimated number of people with diagnosed HIV in the study area, based on HIV and AIDS registration [19,20]. Although 155 controls were known to have died without diagnosed NHL, and the remainder did not have NHL by mid-1996, autopsy series suggest that up to 20% of people with AIDS develop NHL [13,21]. If some controls did develop NHL, the true associations would be likely to be stronger than those reported in this study.
Immune deficiency is well described as a major risk factor for AIDS-NHL. Rates of NHL increase exponentially with time in HIV-infected individuals  and in one report were more than 20 times higher in those with a CD4 count of less than 100 × 106cells/l compared to those with a CD4 count of greater than 500 . Although the present study was closely matched for degree of immune deficiency at NHL diagnosis, we have identified duration of immune deficiency as an important risk factor.
The major new finding of this study, namely, that raised serum globulin (a surrogate marker of serum immunoglobulin) is associated with risk of NHL and that this relationship is present for several years prior to NHL diagnosis, provides strong support for the theory that chronic B-cell stimulation is a risk factor for NHL . Serum immunoglobulin is increased from early in the course of HIV disease, reflecting B-cell stimulation [8,9], but there have been few empirical data comparing B-cell stimulation in people with AIDS-NHL to people with HIV who do not develop NHL. No previous reports have controlled closely for the major confounding factor, degree of immune deficiency. A small study found that serum levels of the cytokine interleukin 6 (which stimulates B-cell replication) were significantly higher in those HIV-positive patients who later developed NHL compared with those who did not . As serum globulin is only an approximate surrogate marker of B-cell stimulation, we have probably under-estimated the true strength of the relationship with NHL risk due to measurement error leading to non-differential misclassification.
We found no association between recorded or serological history of sexually transmissible diseases and risk of NHL. These data support analyses of surveillance data [6,23] and a case–control study , which concluded that a sexually transmissible agent is unlikely to cause a substantial proportion of AIDS-NHL.
Although molecular data have consistently implicated EBV in the pathogenesis of AIDS-NHL, no association was found between EBV-related clinical disease (oral hairy leukoplakia) or anti-herpesviral treatment at doses sufficient to inhibit EBV replication and risk of NHL. This finding does not exclude a role of EBV in lymphomagenesis, as EBV infection in NHL is latent whereas anti-herpesviral agents are active only against lytic EBV infection . Previous studies have found conflicting results, reporting either an association  or no association  with anti-herpesvirus therapy, but control for degree of immune deficiency in these studies was less stringent than in the current study.
The present study provides further evidence that zidovudine use is not associated with an increased risk of AIDS-NHL. Since the original publication suggesting that zidovudine use may be associated with NHL risk , subsequent studies have found no association [12,26], and use of zidovudine is not associated with NHL risk in randomized controlled trials . In the current study, receipt of nucleoside analogue combination antiretroviral therapy (i.e. zidovudine plus either of didanosine or zalcitabine) was associated with a non-significant reduction in NHL risk.
If, as this study suggests, AIDS-NHL results from chronic stimulation of B cells in the absence of effective T-cell control, then there are two possible avenues of prevention. Effective prevention of NHL in people with HIV will depend on the prevention or reversal of the strongest risk factor, immune deficiency. On the other hand, our data suggests that reduction of B-cell stimulation, as may be possible by effective inhibition of HIV replication in lymphoid tissues  will reduce the risk of NHL. The overall effect on NHL risk will depend on the relative contribution of these two factors. Thus, the increases in survival associated with the use of combination antiretroviral therapies  may be associated with increases in lifetime risk of NHL unless immune deficiency can be reversed. Data from cohorts of people receiving combination antiretroviral therapy suggest that the incidence of AIDS-NHL has fallen, but by less than that of other AIDS-associated illnesses [29,30]. If combination therapy remains only partly effective in immune system reconstitution, it appears likely that NHL will become proportionately more important as a cause of morbidity and mortality in people with HIV and AIDS.
The authors thank Ms Stella Bullard for her help in data collection, and the medical records staff of participating hospitals and clinics.
1. Rabkin CS, Hiltgartner MW, Hedberg KW. et al. Incidence of lymphoma and other cancers in HIV-infected and HIV-uninfected patients with hemophilia.
JAMA 1992, 267: 1090 –1094.
2. Munoz A, Schrager LK, Bacellar H. et al. Trends in the incidence of outcomes defining AIDS in the MACS study: 1985–1991.
Am J Epidemiol 1993, 137: 423 –438.
3. Pluda JM, Venzon DJ, Tosato G. et al. Parameters affecting the development of non-Hodgkin's lymphoma in patients with severe human immunodeficiency virus infection receiving antiretroviral therapy.
J Clin Oncol 1993, 11: 1099 –1107.
4. Beral V, Peterman TA, Berkelman RL, Jaffe HW. Kaposi's sarcoma among persons with AIDS: a sexually transmitted infection?
Lancet 1990, 335: 123 –128.
5. Chang Y, Cesarman E, Pessin MS. et al. Identification of Herpesvirus-like DNA sequences in AIDS-associated Kaposi's Sarcoma.
Science 1994, 266: 1865 –1869.
6. Beral V, Peterman T, Berkelman R, Jaffe H. AIDS-associated non-Hodgkin's lymphoma.
Lancet 1991, 337: 805 –809.
7. IARC working group on the evaluation of carcinogenic risks to humans. Human Immunodeficiency Viruses and Human T-cell Lymphotrophic Viruses.
IARC Monographs on the evaluation of carcinogenic risks to humans, Volume 67. Lyon: IARC; 1996.
8. Amadori A, Chieco-Bianchi L. B-cell activation and HIV-1 infection: deeds and misdeeds.
Immunol Today 1990, 11: 374 –378.
9. Middleton GW, Lau RK. AIDS lymphomas.
Int J STD AIDS 1992, 3: 173 –181.
10. Dolcetti R, Boiocchi M. Cellular and molecular bases of B cell clonal expansions.
Clin Exp Rheumatol 1996, 14 (Suppl. 14): S3 –S13.
11. Pluda JM, Yarchoan R, Broder S. The occurrence of opportunistic non-Hodgkin's lymphoma in the setting of infection with the HIV.
Ann Oncol 1991, 2 (Suppl. 2): 191 –200.
12. Moore RD, Kessler H, Richman DD, Flexner C, Chaisson RE. Non-Hodgkin's lymphoma in patients with advanced HIV infection treated with zidovudine.
JAMA 1991, 265: 2208 –2211.
13. Ridolfo AL, Santambrogio S, Mainini F. et al. High frequency of non-Hodgkin's lymphoma in patients with HIV-associated Kaposi's sarcoma.
AIDS 1996, 10: 181 –185.
14. World Health Organization. Manual of the International Statistical Classification of Diseases, Injuries, and Causes of Death
, 9th revision. Geneva: World Health Organization; 1977.
15. Grulich AE, Wan X, Coates M, Day P, Kaldor JM. Validation of a non-identifying method of linking cancer and AIDS register data.
J Epidemiol Biostat 1996, 1: 207 –212.
16. Tietz NW. Fundamentals of Clinical Chemistry
. Philadelphia: WB Saunders; 1976.
17. Armenian, Armenian, HK, Hoover, Hoover, DR, Rubb S. et al. Risk factors for non-Hodgkin's Lymphoma in AIDS.
Am J Epidemiol 1996, 143: 374 –379.
18. Holly EA, Lele C. NHL in HIV-positive and HIV-negative homosexual men in the San Francisco Bay Area: allergies, prior medication use and sexual practices.
J Acquir Immune Defic Syndr Hum Retrovirol 1997, 15: 211 –222.
19. National Centre in HIV Epidemiology
and Clinical Research. HIV/AIDS and Related Diseases in Australia: Annual Surveillance Report, 1997
. Sydney: NCHECR; 1997.
20. Li Y, Gold J, McDonald A, Kaldor J. Demographic patterns of AIDS in Australia, 1991 to 1993.
Aust J Public Health, 1996, 20: 421 –426.
21. Wilkes MS, Fortin AH, Felix JC, Godwin TA, Thompson WG. The value of necropsy in AIDS.
Lancet 1988, ii: 85 –88.
22. Scherr PA, Mueller NE. Non-Hodgkin's lymphoma.
In Cancer Epidemiology and Prevention
. Edited by Schottenfeld D, Fraumeni JF. New York: Oxford University Press; 1996: 920– 945.
23. Serraino D, Salamina G, Franceschi S. et al. The epidemiology of AIDS-associated non-Hodgkin's lymphoma in the WHO European region.
Br J Cancer 1992, 66: 912 –916.
24. Kedes DH, Ganem D. Sensitivity of Kaposi's sarcoma herpesvirus replication to antiviral drugs.
J Clin Invest 1997, 99: 2082 –2086.
25. Fong MW, Lo B, Ho J, Toy I, Fong IW. Chronic high dose acyclovir may prevent AIDS-related lymphomas (non-Hodgkin's lymphoma): a case control study.Fourth Conference on Retroviruses and Opportunistic Infections
. Washington, 1997 [abstract no 721].
26. Levine AM, Bernstein L, Sullivan-Halley J, Shibata D, Mahterian B, Nathwani BN. Role of zidovudine antiretroviral therapy in the pathogenesis of AIDS-related lymphoma.
Blood 1995, 86: 4612 –4616.
27. Andersson J, Fehniger TE, Patterson BK. et al. Early reduction of immune activation in lymphoid tissue following highly active HIV therapy.
AIDS 1998, 12: F123 –F129.
28. Egger M, Hirschel B, Francioli P. et al. Impact of new anti-retroviral combination therapies in HIV infected patients in Switzerland: prospective multicentre study.
BMJ 1997, 315: 1194 –1199.
29. Moore R, Keruly JC, Gallant J, Chaisson RE. Decline in mortality rates and opportunistic disease with combination anti-retroviral therapy.World AIDS Conference,
Geneva, 1998 [abstract no 22374].
30. Costagliola D. Clinical manifestations of HIV infections in the era of highly active antiretroviral treatment (HAART) in France.World AIDS Conference
, Geneva 1998 [abstract no 12345].
Keywords:© 2000 Lippincott Williams & Wilkins, Inc.
non-Hodgkin's lymphoma; epidemiology; risk factors; immune deficiency; immune stimulation