Response to highly active antiretroviral therapy strongly predicts outcome in patients with AIDS-related lymphoma
AIDS 2003; July 2003 17(10):1521-1529
Christian Hoffmann a; Eva Wolf b; Gerd Ftkenheuer c; Thomas Buhk d; Albrecht Stoehr e; Andreas Plettenberg e; Hans-Jrgen Stellbrink d; Hans Jaeger b; Uwe Siebert f; Heinz-August Horst a. University Hospital Kiel, the bKIS, Curatorium for Immunedeficiency, Munich, the cUniversity of Cologne, the dUniversity Clinic of Eppendorf, Hamburg, the eSt. Georg Hospital, Hamburg, Germany and the fHarvard School of Public Health, Boston, Massachusetts, USA.
HIV-infected patients are at high risk for non- Hodgkin's lymphoma (NHL). When compared to the general population, the relative risk for high-grade NHL in adults with HIV/AIDS is 100-400-fold higher according to registry-linkage studies. Within the EuroSIDA study, almost 16% of AIDS-defining illnesses (ADIs) in 1998 were ARL. The impact of highly active antiretroviral therapy (HAART) on the incidence of AIDS-related lymphoma (ARL) remains
controversial. Early observations suggested that the decrease in ARL was either lacking or less profound than seen with other ADIs. These findings raised concern that the events triggering the emergence of ARL were different from those preventing opportunistic infections or Kaposi's sarcoma and were not reversible by improved immune responses. Although newer data suggest that the risk of developing ARL has now started to decrease significantly, albeit slowly, it seems likely that ARL will remain one of the most common ADIs.
ARL is a biologically heterogeneous disease. For its development and progression more than one pathogenetic mechanism is operational. HIV-induced immunosuppression with a dysfunctional cellular immunity against
Epstein-Barr virus (EBV) infected B-cells is only one pathogenic factor among others. The different subtypes of ARL vary substantially with regard to EBV and other viral infections, to the extent of immune deficiency, to oncogenic
mutations, to cytokine dysregulation and to histogenetic derivation. However, ARL shares several distinctive features when compared to NHL in the general population. These factors include not only an almost exclusive B-cell derivation, a propensity towards advanced disease and an extensive extranodal involvement with unusual localizations but also clinical aggressiveness with poor therapeutic outcome. The median survival reported in various studies in the pre-HAART era ranged from 1.5 to 18 months.
Studies addressing the impact of HAART on survival of patients with ARL have yielded conflicting data. At least five large cohort studies reported sobering results. This contrasts to the results of smaller and (in part) prospective,
non-randomized studies showing a significant improvement in survival and to our own observations in patients with primary central nervous system lymphoma. In the present study, we analyzed the outcome of patients with
systemic ARL in a large multicentric cohort with respect to the use and efficacy of HAART and other potential prognostic factors. Our study also focussed on the changing clinical spectrum of ARL in the HAART era.
Background: AIDS-related lymphoma (ARL) remains a frequent complication of HIV infection. We analyzed the outcome of patients with ARL with respect to the use and efficacy of highly active antiretroviral therapy (HAART) and to potential prognostic factors.
Methods: This multicenter cohort study included patients with systemic ARL diagnosed between 1990-2001. We evaluated overall survival and the effects of several variables on overall survival using the Kaplan-Meier method and the extended Cox proportional hazards model. Response to HAART was used as a time-dependent variable and was defined as a CD4 cell count increase of >100 x 106 cells/l and/or at least one viral load <500 copies/ml during the first 2 years following diagnosis of ARL.
This ongoing, multicenter cohort study of HIV- infected patients with systemic high-grade or intermediate NHL diagnosed histologically between January 1990 and December 2001 in five German HIV centers was started in 2000. Patients were identified by computed database analysis in each center: University of Cologne (68 cases); St. Georg Hospital, Hamburg (52 cases); University of Eppendorf, Hamburg (37 cases); KIS - Curatorium for Immunedeficiency, Munich (32 cases); University of Kiel (14 cases). Patients with low-grade lymphoma or primary central nervous system lymphoma were excluded.
Chemotherapy regimens consisting of cyclophosphamide, doxorubicin, vincristine and prednisone (CHOP), or comparable or more intensive regimens were considered to be curative. Complete remission (CR) was defined as the
absence of clinical and radiological evidence of ARL. Histological diagnosis was classified according to the WHO classification. As most cases were classified by German referral centres, specimens were not reviewed. Any
antiretroviral treatment given for a period of at least 4 weeks was considered as antiretroviral therapy. HAART was defined as a combination of at least three antiretroviral drugs. Response to HAART was used as a time-dependent
variable and was defined as an increase in the CD4 cell count of >100 x 106 cells/l and/or at least one viral load <500 copies/ml during the first 2 years following ARL diagnosis. The definition of response was adapted and slightly
modified from previous studies
Results: Among 203 patients with ARL, median overall survival was 9.0 months [95% confidence interval (CI), 7.6-12.4 months]. In the univariate analyses, age <60 years, no previous AIDS, CD4 cell counts >200 x 106 cells/l, hemoglobin > 11 g/dl, Ann Arbor stages I-II and A, no extranodal lesion, response to HAART, and complete remission showed statistically significant association with prolonged overall survival. In the multivariate Cox model, the only factors independently associated with overall survival were response to HAART [relative hazard (RH), 0.32; 95% CI, 0.16-0.62], complete remission (RH, 0.24; 95% CI, 0.15-0.36), previous AIDS (RH, 1.92; 95%CI, 1.23-3.01) and extranodal involvement (RH, 2.85; 95% CI, 1.47-5.51).
Conclusions: Efficacy of HAART was independently associated with prolonged survival in this large cohort of patients with ARL. Information on patient's response to HAART is crucial for the evaluation of future treatment strategies.
A total of 203 patients with ARL were identified. The most frequent mode of HIV-1 infection was male-to-male sexual transmission (65%), followed by heterosexual transmission (11%). Relatively few patients reported a history of intravenous drug use (3%) or the reception of blood products (2%) or came from pattern II countries (3%). In 15%, the mode of infection was unknown or data was not available. The median time from the first positive HIV test to ARL diagnosis was 3.6 years. Profound immunosuppression was a common finding, with 48% having a CD4 cell count below 100 x 106 cells/l. Of 77 patients with an evaluable plasma viremia at the time of ARL diagnosis, only 15 (19%) had a viral load < 500 copies/ml. Among the 75 patients with previous AIDS, 142 ADIs had occurred prior to ARL diagnosis. Extranodal manifestation was a common feature as was advanced stage. Many patients displayed an elevated LDH (lactate dehydrogenase) and an anemia. ARL were mostly of B-cell origin. Within the WHO classification, 57% of the B-cell lymphoma were classified as diffuse large cell lymphoma (DLCL), 27% were classified as Burkitt- or Burkitt-like lymphoma (BL/BLL), and 5% were otherwise classified. The remaining 11% were not classifiable.
Associations between different risk factors and changes of clinical spectrum over time
Patients with extranodal disease had a significantly higher percentage of elevated LDH (67 versus 40%, P = 0.01). Patients with BL/BLL also had a higher percentage of elevated LDH (80 versus 57%, P = 0.02) and were more likely to
present with disease stage III-IV (84 versus 66%, P = 0.04) when compared with patients with DLCL. In contrast, the median CD4 cell count was higher in patients with BL/BLL (164 versus 90 x 106 cells/l, P = 0.04). Patients with a
prior AIDS diagnosis had a lower hemoglobin level with 61 versus 27% having a value of <11 g/dl (P < 0.0001).
Analyzing changes over time, ARL was found more likely to be the first ADI in the years 1997-2001 than in the years 1990-1996 (73 versus 58%, P = 0.047). The baseline median CD4 cell count tended to be slightly higher in the years
1997-2001 (122.5 versus 80 x 106 cells/l, P = 0.08) and in patients who had received HAART prior to ARL diagnosis (145 versus 91.5 ? 106 cells/l, P = 0.25).
There were no differences in the ARL manifestation pattern with respect to time period or HAART use. Low plasma viremia was not associated with early stages of ARL in this cohort. 12 of 15 patients with < 500 copies/ml at the time of ARL diagnosis (median time since the last viral load > 500 copies/ml: 9 months, ranging from 2-39 months) presented with disease stage III/IV, and 14 of 15 patients had at least one extranodal manifestation. All five patients with a sustained viral load < 500 copies/ml of > 12 months before ARL diagnosis showed an extranodal manifestation, with four of five presenting disease stage IV.
Chemotherapy and antiretroviral therapy
Following ARL diagnosis, curative chemotherapy was initiated in 159 patients (78%). Of the remaining 44 patients, 22 received no ARL chemotherapy and 11 were treated palliatively. Nine patients were diagnosed post-mortem, and in two patients data on chemotherapy was insufficient.
The by far most frequently used protocol in patients with curative chemotherapy was the CHOP regimen in standard or reduced dosage (84%). The mean number of cycles was 4.3 with 72% of the patients receiving at least four cycles. A CR was achieved in 57% of patients starting curative chemotherapy. Patients attaining CR were less likely to be profoundly immunosuppressed compared to patients without CR, with 53 versus 72% having a CD4 cell count of less than 200 x 106 cells/l. The CR rate did not change over time.
Reflecting the growing availability of antiretroviral regimens, between 1997-2001 71% of the patients received HAART during the first 2 years following ARL diagnosis (compared to 9% between 1990-1996). Of the remaining 29% of the patients without HAART, 18 of 20 died within the first 2 years. Median survival of the 13 curatively treated patients in this group was 4.7 months.
Patients receiving HAART (n = 61) were significantly more likely to achieve CR (71 versus 48%, P = 0.006). In total, 47 of 61 patients were regarded as HAART responders. In the HAART responders, the CR rate was 77%, compared to
50% in the non-responders (P = 0.09). Patients with previous AIDS were less frequently HAART responders (50 versus 83%, P = 0.03).
Overall survival in this cohort was relatively poor with a median of 9.0 months (95% CI 7.6-12.4 months). The 1-year and 2-year overall survival rates were 44 and 32%, respectively. By the end of 2001, 151 patients (74%) had died, 39
patients (19%) were alive, and 13 (6%) were lost to follow-up.
In the univariate analyses, several pre-existing risk factors were identified to be predictive for overall survival. These factors included age > 60 years, previous AIDS, a CD4 cell count <200 x 106 cells/l, a hemoglobin level <11 g/dl and also ARL features such as extranodal manifestation, elevated LDH and advanced stage. Patients with BL/BLL had only a non-significant trend for a poorer prognosis. There were no differences between B- and T-cell origin of ARL, gender, or different HIV transmission risks. In the extended Cox model, the only pre-existing factors independently associated with overall survival were a prior AIDS diagnosis and extranodal involvement.
Two non-baseline factors remained predictive in the Cox model, namely a response to HAART and a CR after chemotherapy. The median survival in patients with both CR and response to HAART was not reached with
83% surviving after 39 months (Kaplan-Meier estimate). The significant survival benefit of HAART response was also seen when only patients achieving a CR, or when only patients with a relatively preserved immune competence >200 x 106 CD4 cells/l), were analyzed. When reviewing our data in the light of other, more stringent definitions of HAART response (for example, when the response group was extended to include also patients with a sustained decline of > 1.0 log or > 2.0 log copies/ml of plasma viremia), there were no significant differences concerning risk reduction.
The results of this large cohort of HIV patients with high-grade or intermediate NHL confirm previously described features of ARLAdvanced and extranodal disease at manifestation were common findings as was profound
immunosuppression. There have been reports suggesting changes in ARL characteristics over time with a reduced probability of advanced and extranodal stage and an increase in the CD4 cell count. However, these observations were not made uniformly. In our cohort, there were no changes over time in different ARL features, according to time of diagnosis, ART use, or immunological or virological parameters at baseline. For example,
nearly all patients with a complete, in some cases long-standing, suppression of HIV plasma viremia at ARL diagnosis presented with advanced and extranodal disease. Several in vitro studies elegantly demonstrated that HIV infection of non-malignant accessory cells plays an essential role in the homing and outgrowth of lymphoma cells at extranodal sites and may therefore explain the extensive extranodal involvement in ARL. Although the number of patients with sufficient suppression of HIV plasma viremia was low, these preliminary observations suggest that the HIV triggered mechanisms may occur earlier in the course of the disease and perhaps years before ARL onset.
However, ARL in our cohort, as reported by some others, were more likely to be AIDS-defining in the HAART era indicating that the decrease in ARL is less profound than seen with other ADIs. Thus, ARL will remain an
important health problem in HIV infected patients, and the data presented may have some implications on future treatment strategies.
The role of pre-existing risk factors
Before the introduction of HAART, several pre- existing risk factors were found to be predictors of poor outcome in ARL patients. In many studies, survival was most strongly related to factors indicating the degree of
immunosuppression such as low CD4 cell counts, a prior AIDS-defining illness and a poor performance status, but also to ARL factors such as extranodal disease, particularly bone marrow involvement, advanced stage, and elevated LDH. Older age, low hemoglobin level, stomach involvement, the presence of immunoblastic subtype, monoclonality, and the expression of multidrug resistance genes represent additional adverse prognostic factors, although their predictive value has been described more inconsistently.
In the multivariate analysis of this large cohort, only two pre-existing risk factors, namely a prior AIDS diagnosis and extranodal involvement, still had a significant predictive value. This suggests that the importance of many risk factors, concerning both immunological and lymphoma features, may be reduced in the HAART era, apparently dominated by the efficacy of antiretroviral therapy.
One major limitation of this study was that, due to the retrospective nature of the data, the evaluation of performance status was not possible. Performance status has been shown to be a relevant risk factor for poor survival, which may have influenced treatment decisions for both HIV infection and ARL. To reduce this bias, we excluded all patients from the multivariate analysis who were not started on curative chemotherapy.
Some smaller studies also suggested reduced predictive values for most risk factors on survival. These observations, together with the strong predictive value of the efficacy of HAART in our cohort, indicate that the decision
on ARL therapy should focus not only on the efficacy of chemotherapy but also on optimal ART, while pre-existing risk factors may be less important.
The role of chemotherapy
Achieving CR of ARL after chemotherapy was an independent predictive factor in the multivariate analysis, emphasizing the importance of initiating curative chemotherapy regimens. The few randomized trials to date employing CHOP or other approaches in ARL patients resulted in CR rates of 43-67% [53, 54], resembling the overall CR rate of 57% in our cohort, in which the CHOP regimen was used almost uniformly. The CR rates in this study did not change over time. However, they did so with respect to HAART use and response after ARL diagnosis. The CR rate of 77% in HAART responders was remarkably higher than that achieved with numerous chemotherapy regimens in the pre-HAART era. One could argue that this difference may be simply due to the effect that patients achieving CR might be in a better condition to receive HAART, and that CR enabled patients to receive HAART and to take it successfully. However, our analysis showed that response to HAART was an independent predictor for prolonged survival even after adjustment for CR. This did not change significantly when other criteria of virological and immunological response were chosen (data not shown). Therefore, even if the above argument is true, it cannot explain the entire effect of efficacious HAART in our cohort. Furthermore, there is strong evidence from other trials, that the response to chemotherapy is improved by HAART. In one study, CR was achieved in 71% of HAART responders and 30% of non-responders. In another small prospective trial, the immune function was better maintained when combining chemotherapy with HAART. These findings indicate an improved efficacy of chemotherapy in ARL patients on sufficient ART. As the immune system plays a major role in effectively surveying and protecting against malignant lymphoid transformation,
correcting the immune deficit may be essential not only to reduce the risk of AIDS-defining diseases and other infectious complications but also to enhance the tolerability of chemotherapy and to improve its efficacy. We were able to demonstrate in patients with primary central nervous system lymphoma that the dramatic improvement in survival in the HAART era may be at least partially due to a direct anti-tumor effect of HAART-induced immune recovery. Taken together, our results suggest that detailed information on the patient's response to antiretroviral therapy is crucial for the evaluation of the efficacy of different chemotherapy regimens in ARL.
An increasing use of hematological growth factors may also have contributed to a better tolerability of chemotherapy and to a better efficacy. In our cohort, dose-intensified CHOP regimens with a prophylactic use of granulocyte colony-stimulating factor (G-CSF) were not used. However, G-CSF was infrequently utilized in some patients. Therefore, we cannot exclude definitely an effect of G-CSF in this retrospective cohort.
The role of antiretroviral therapy and success
Our results demonstrate a strong predictive value of the efficacy of HAART on survival in ARL patients. The median survival time in patients with CR and response to HAART was not reached, with 83% surviving after 39 months. Our
observations contrast with at least five large cohort studies which recently reported on sobering results in the HAART era. This discrepancy may be partially due to an insufficient observation period which could have hindered these studies from fully assessing the benefit of HAART. However, these studies mainly used the design of a correlational analysis merely determining the influence of HAART by dividing the study period into different time intervals according to changes in the availability of antiretroviral compounds. Instead, we were able to analyze HAART effects on the individual patient data level and to adjust for potential confounding introduced through the effect of baseline characteristics on the physician's treatment decision. Indeed, if our follow up had ended in 2000, and if patients were divided by different time periods, only a trend would have been detected. This approach may therefore be inappropriate since many patients in the HAART era do not receive any, never mind a successful, HAART. In this cohort, only 71% of patients diagnosed with ARL after January 1997 received HAART. Of note, the significant survival benefit of HAART response was also seen when only patients with a CD4 cell count of > 200 x 106 cells/l at the time of ARL
diagnosis were analyzed. This indicates that even patients with a relatively preserved immune competence may benefit from HAART.
Efficacy of HAART was a strong predictor of survival in this large cohort of ARL patients. In the multivariate analysis, response to chemotherapy and response to HAART were independently associated with improved survival. Our findings also indicate that the predictive value of many classical risk factors for poor survival in ARL is reduced in the HAART era. Detailed information on the patient`s response to HAART is crucial for the evaluation of different treatment strategies in NHL.