Hydroxyurea and busulfan are the two agents most commonly used as the standard treatment in patients with chronic myelogenous leukemia. These drugs provide effective disease control in the chronic phase, inducing complete hematologic remissions in greater than 70% of patients, depending on the dose schedules used. They are attractive treatment options in community practice because they are administered orally, are relatively inexpensive, and have few chronic side effects. Busulfan may induce organ fibrosis (marrow, pulmonary), Addison-like disease, and unpredictable prolonged myelosuppression (in about 10% of patients) that may be fatal. A recent randomized trial [5] of hydroxyurea compared with busulfan therapy showed that hydroxyurea conferred a survival advantage (median survival, 58 compared with 45 months; P = 0.008). Patients having allogeneic bone marrow transplantation also have better survival if therapy before transplantation does not include busulfan [6]. Conventional therapy has been associated with a median survival range of 3 to 5 years, but neither busulfan nor hydroxyurea produced a substantial, durable suppression of the Ph-positive cells in patients with chronic myelogenous leukemia [7].

Investigations of interferon- therapy in chronic myelogenous leukemia were initiated by our group in 1982. Since then, several trials [8-12] have confirmed its efficacy in this disease. Interferon- therapy produces hematologic responses in 60% to 80% of patients and cytogenetic responses (suppression of the Ph-positive cells) in 35% to 55% of patients. These responses were major (<35% Ph-positive cells) and durable in about 15% to 25% of patients [7].

In a study [10] of chronic myelogenous leukemia by the Italian Cooperative Study Group (ICSG-CML), patients were randomly assigned to interferon- or conventional chemotherapy. The 218 patients assigned to interferon- had longer survival (median > 72 months compared with 52 months; P = 0.003) and time to transformation (median > 72 months compared with 45 months; P < 0.001) and a higher incidence of a major cytogenetic response (19% compared with 1%; P < 0.01) when compared with conventional chemotherapy. We determined whether achieving a cytogenetic response with interferon- therapy was independently related to prolonged survival in patients with chronic myelogenous leukemia.

Methods

Top Methods Results Discussion Author & Article Info References

Study Group

Adults with a diagnosis of Ph-positive, early chronic-phase, chronic myelogenous leukemia (diagnosis for less than 12 months) who received interferon- therapy as their primary treatment after referral to our institution from 1982 until 1990 were included in the analysis. Informed consent was obtained for treatment with the interferon- protocols according to institutional guidelines. The interferon- programs were sequential studies with human leukocyte interferon- (1982 to 1984; DM82-49), recombinant interferon- alone (1984 to 1987; DM84-38, DM86-86), recombinant interferon- and interferon- (1985; DM85-45), and recombinant interferon- plus hydroxyurea (1988 to 1990; DM88-99). The early analyses of the clinical results of DM82-49 and DM84-38 have been reported [8, 12]. Patients had pretreatment and follow-up evaluations of bone marrow aspirates, and had cytogenetic analysis every 3 months in the first 2 years and every 4 to 6 months thereafter.

Therapy

In the subsequent studies, the daily dose of interferon- was 5 x 10⁶ U/m². The interferon- dose was reduced by 25% for persistent grade 2 toxicity. For grade 3 to 4 toxicity, interferon- was interrupted until resolution of the toxicity, and the agent was resumed at 50% of the dose. Toxicity grading was according to the National Cancer Institute guidelines [13]. The dose was also reduced by 25% if the leukocyte count was lower than 2 x 10⁹/L or if the platelet count was lower than 60 x 10⁹/L. In the study combining interferon- and interferon- (DM85-45), the interferon- daily dose was 0.025 mg/m (²), and interferons- and – were administered daily. Oral hydroxyurea was adjusted in a daily dose range of 0.5 g to 2 g to keep the leukocyte count less than 5 x 10⁹/L with a platelet count greater than 60 x 10⁹/L.

Response Criteria

Response criteria have been previously described [8]. A complete hematologic response required normalization of the peripheral blood counts and differential, including a leukocyte count less than 10 x 10⁹/L; a platelet count less than 450 x 10⁹/L; an absence of immature peripheral blasts, promyelocytes, or myelocytes; and disappearance of all signs and symptoms of the disease, including palpable splenomegaly [8]. Patients achieving a complete hematologic response were further categorized by their best cytogenetic response [lowest percentage of Ph-positive metaphases]: 1) as complete, if the lowest Ph-positive percentage was 0%; 2) as partial, if it was between 1% and 34%; and 3) as minor, if it was between 35% and 90%. A major cytogenetic response included complete and partial cytogenetic responses (that is, patients with the lowest Ph-positive metaphase percentage below 35%).

Prognostic Models

Prognostic models for survival of patients with chronic myelogenous leukemia, previously proposed by our group and by Sokal and colleagues [14-16], were evaluated by comparing response and survival experiences of patients divided into different risk groups. The "overall" prognostic model was derived from an analysis of clinical, hematologic, and bone marrow features of 303 patients with Ph-positive, chronic-phase chronic myelogenous leukemia who were referred to our institution between 1965 and 1982 [14]. The multivariate analysis [14] identified five independent poor prognostic factors: 1) age of 60 years or greater; 2) black race; 3) marrow basophils of 3% or greater; 4) blood basophils of 7% or greater; and 5) cytogenetic clonal evolution (acquired chromosomal abnormalities other than Ph).

The "clinical" prognostic model was based on the analysis of the same population but was derived excluding the bone marrow findings. In addition to age, race, and peripheral basophilia, two other independent poor prognostic factors were selected: 1) platelet count of 700 x 10⁹ or greater or of less than 150 x 10⁹/L; and 2) increasing splenomegaly [14]. Both models divide patients into good, intermediate, and poor risk groups based on their hazard ratio (or risk for death per unit time compared with the average expected risk), derived from solving a mathematical equation (detailed in reference 14).

The Sokal model is based on a similar analysis. The population analyzed included 813 patients with "good risk" (essentially chronic-phase) chronic myelogenous leukemia who were treated in multiple institutions in the United States, Italy, and Spain [16]. It identified independent poor prognostic factors, including 1) older age; 2) higher platelet counts; 3) higher peripheral blast percentage; and 4) splenomegaly. The Equation tocalculate the hazard ratio and the analysis have been previously described [16].

The "synthesis" model attempted to synthesize the existing models, choosing the consistently reproducible prognostic factors and dividing patients into prognostic groups by a simple enumeration of the number of poor prognostic factors present at diagnosis. Patients are categorized as having stage 1, 2, or 3 disease if they have none or 1 (stage 1), 2 (stage 2), or 3 or more (stage 3) of the following poor prognostic factors: 1) age of 60 years or greater; 2) peripheral blasts of 3% or greater or bone marrow blasts of 5% or greater; 3) peripheral basophils of 7% or greater, or marrow basophils of 3% or greater; 4) platelet count of 700 x 10⁹/L or greater; and 5) splenomegaly of 10 cm or greater below the costal margin. Accelerated-phase chronic myelogenous leukemia, or stage 4 disease, is defined by the presence of any of the following: 1) peripheral blasts of 15% or greater; 2) peripheral basophils of 20% or greater; 3) peripheral blasts plus promyelocytes of 30% or greater; 4) a platelet count less than 100 x 10⁹/L that was unrelated to therapy; or 5) cytogenetic clonal evolution. Confirmation of the validity of the model was carried out in 406 patients with Ph-positive chronic myelogenous leukemia who were referred to our institution from 1975 through 1986. The details of the model and a discussion of the other models are available in a previous report [15].

Statistical Methods

Outcomes evaluated in this study were frequency of cytogenetic response and duration of survival. Survival analysis was based on Kaplan-Meier estimation [17], and groups were compared by log-rank test [18]. Except where noted, survival was dated from initiation of interferon- therapy. Patients who had allogeneic bone marrow transplantation while still in the chronic phase of the disease were counted as censored observations at the date of transplantation.

Cytogenetic response did not occur, on average, until a patient had received 12 months of interferon- therapy. Two approaches were used to evaluate the association of a cytogenetic response with survival: 1) comparison of survival subsequent to a fixed time point according to cytogenetic response at that time (landmark method) [19] and 2) use of a proportional hazards regression model [20] for survival that included a time-varying term for cytogenetic response.

In the first approach, the "landmark" used for analysis was 12 months after initiation of interferon- therapy. Sixteen patients who died or were censored before 12 months were excluded. Patients were grouped according to their cytogenetic response at 12 months, and subsequent survival was compared for the overall group and within prognostic subsets based on the synthesis model (described above).

In the second approach, an indicator variable was defined in a proportional hazards model, taking the value 0 for each patient until the time at which a major cytogenetic response was detected and taking a value of 1 for each patient thereafter. A test for statistical significance of the coefficient for this term was, thus, a test for whether cytogenetic response status was associated with subsequent risk for death. Important pretreatment factors ("fixed" covariates) were also included in the model in order to adjust for their prognostic effects. Associations of pretreatment factors with survival were first assessed individually using log-rank tests, and then the joint effect of variables with prognostic potential was assessed in a proportional hazards model using a stepwise procedure. Pretreatment factors for which P 0.05 in that model were selected for inclusion in the final model along with the term for cytogenetic response.

Associations between some patient characteristics were evaluated using rank correlation coefficients. Performance status was defined on the Zubrod scale: 0 = asymptomatic, 1 = minor symptoms, 2 = bedridden less than 50% of normal day, 3 = bedridden greater than 50% of normal day, and 4 = bedfast.

Results

Top Methods Results Discussion Author & Article Info References

The characteristics of the 274 patients analyzed are shown in Table 1. Their median age was 41 years (range, 15 to 76 years, and 109 [40%] were women. The median time from diagnosis to start of therapy was 2 months (range, 0 to 11 months). Forty-nine patients were treated with human leukocyte interferon- alone, 136 patients received the recombinant form of interferon- alone (102 patients) or combined with recombinant interferon- (34 patients), and 89 patients received recombinant interferon- and hydroxyurea.

Of the 274 patients treated, 219 (80%) achieved complete hematologic response, 19 (7%) had a partial hematologic response, and 36 (13%) had resistant disease. Among the 219 patients with complete hematologic response, 159 (58% of the 274 patients) had a cytogenetic response that was complete (Ph, 0%) in 72 patients (26%), partial (Ph, 1% to 34%) in 32 patients (12%), and minor in 55 patients (20%). Thus, 104 patients (38%) were categorized as having a major cytogenetic response (complete or partial).

The median follow-up of patients alive on study was 52 months (range, 4 to 125 months); 92 of the 274 patients have died. The estimated median overall survival was 89 months (94% CI, 63 to 101 months), with an estimated 5-year survival rate of 63% (Figure 1). Thirty-one patients had allogeneic bone marrow transplantation in the chronic phase: Twelve have died at times ranging from 2 to 38 months; their estimated median survival from transplantation was 38 months. Counting the actual survival experience of these patients rather than censoring them at the time of transplantation had little effect on survival estimates for the total study group.

View larger version (13K): [in this window] [in a new window]   Figure 1. Survival of 274 patients with Philadelphia chromosome-positive, early chronic-phase chronic myelogenous leukemia who were treated with interferon- between 1982 and 1990. Data are censored for patients having allogeneic bone marrow transplantation in the chronic phase.

Factors Associated with Achievement of a Major Cytogenetic Response

The associations between host and leukemic-cell characteristics and achievement of a major cytogenetic response are summarized in Table 2. A good performance status (Zubrod score, 0), the absence of symptoms at diagnosis, small spleen size, high hemoglobin levels, low leukocyte counts, and low peripheral blast percentage were all associated with higher incidences of a major cytogenetic response (P 0.01).

View this table: [in this window] [in a new window]   Table 2. Associations between Patient Characteristics and Achievement of a Major Cytogenetic Response and Survival with Interferon- Therapy

When patients were subgrouped according to risk using the prognostic models, significant differences in the incidences of a major cytogenetic response were noted among various risk subgroups (Table 3). Depending on the prognostic model used, the incidence of a major cytogenetic response ranged from 46% to 52% with good prognosis, from 32% to 38% with intermediate prognosis, and from 14% to 26% with poor prognosis (Table 3).

Pretreatment Factors Associated with Survival

Pretreatment characteristics associated with statistically better survival experience included good performance status (Zubrod scale, 0), small spleen size, high hemoglobin level, low leukocyte count, low percentage of circulating blasts or nucleated erythrocytes, and low percentage of marrow basophils (Table 2). Prognostic models were used to categorize patients into risk groups, and comparison of survival indicated statistical differences in survival outcomes (Figure 2). Survival experience, on average, was prolonged for patients classified as having a "better risk" and was similar for different models, although more patients were assigned to the favorable subset by the synthesis model. Depending on the prognostic model used, median survival ranged from 102 to 104 months among patients with good prognosis, from 82 to 95 months among patients with an intermediate prognosis, and from 47 to 62 months among patients with a poor prognosis (Table 3); (Figure 2).

View larger version (30K): [in this window] [in a new window]   Figure 2. Survival of patients in different prognostic groups according to the three models derived from our institution and the Sokal model. The overall (top left) and clinical (top right) prognostic models have been previously described [14], as have the synthesis model (bottom left [15]), and the Sokal model (bottom right [16]). The four models show a good discrimination of patients treated with interferon- into good-, intermediate-, and poor-risk groups (P = 0.01 for the overall model, P < 0.01 for the clinical and synthesis models, and P = 0.10 for the Sokal model).

Factors for which the univariate analysis yielded evidence of an association with survival were further evaluated by fitting a proportional hazards regression model; these factors also corresponded to known prognostic factors for chronic myelogenous leukemia. Two variables were excluded from consideration: 1) leukocyte count because of its close correlation with spleen size, hemoglobin level, and percentage of circulating blasts; and 2) nucleated erythrocytes because they were not measured in 17% of patients. Terms evaluated in a regression model included 1) performance status (Zubrod scale) at diagnosis [0, 1]; 2) spleen size at diagnosis [<5 cm, 5 cm]; 3) hemoglobin level (g/L) at diagnosis [<100, 100 to 119, 120]; 3) peripheral blasts at diagnosis [0, 1%]; and 4) bone marrow basophils (<3%, 3%). After a stepwise fitting procedure, terms for the pair of variables—spleen size and bone marrow basophils—were both significant at the 0.01 level, and tests for addition of a third factor yielded P values greater than 0.15. These findings were in accord with the strong correlations among splenomegaly, anemia, leukocytosis, and peripheral blasts (Spearman r 0.43 for all pairs).

Comparison of Interferon- Regimens

The characteristics of patients in the different trials, and their response and survival outcomes are summarized in Table 4. Patients entered on the initial human interferon- trial tended to have somewhat worse prognosis, particularly with regard to performance status and spleen size, and only 18% achieved a major cytogenetic response. More patients achieved a hematologic response when receiving the regimen that included hydroxyurea and interferon-, and, also, more patients achieved a hematologic response when receiving the regimen of recombinant interferon- alone, but no real evidence indicated that survival was correspondingly improved.

No major differences were noted in survival or duration of major cytogenetic response among the four regimens. Survival experience was also compared by testing treatment indicator terms in a proportional hazards regression model that adjusted for prognostic factors identified in the previous section. The resulting P value was 0.62, confirming that small observed differences in survival are attributed to prognostic differences and not to differences in therapy. Because only nine patients achieved a major cytogenetic response with the human interferon- regimen, it was not possible to draw definitive conclusions about the durability of response in that group; however, six of the nine patients still have durable responses, suggesting the quality of their response is similar to that of other responsive patients.

Achievement of a Cytogenetic Response and Prognosis

For the 104 patients who achieved a major (complete or partial) cytogenetic response, the median duration of interferon- therapy required to reach that status was 12 months (range, 3 to 75 months). The median time to achievement of a complete cytogenetic response for 72 patients was 16 months (range, 3 to 70 months).

In an initial evaluation of the possible effect of a cytogenetic response on subsequent survival, survival was compared for patients grouped according to response status 12 months after initiation of interferon- therapy. One hundred and twenty patients had achieved at least a minor cytogenetic response at that time, and most of these eventually achieved a major cytogenetic response. Survival curves by cytogenetic response at 12 months, dated from the 12-month point, are shown in Figure 3 for the total group (excluding 6 patients who died and 10 who were censored before 12 months). Survival was prolonged among patients with a cytogenetic response compared with the remaining patients (P < 0.001; (Figure 3); only 19 deaths have occurred among patients in responding categories.

View larger version (21K): [in this window] [in a new window]   Figure 3. Landmark analysis of survival dated from 12 months into therapy by the cytogenetic response at 12 months. P < 0.001 for survival of patients with a cytogenetic response compared with those without a response. CR = complete response; IFN- = interferon-; PR = partial response.

To determine whether the prognostic effect of achieving a cytogenetic response to interferon- was selectively beneficial to specific prognostic risk groups or whether outcomes were consistent over all risk groups, this analysis was repeated within patient prognostic groups based on the synthesis model [16]. Results are summarized in Table 5. Because only a few patients were available, patients were categorized as cytogenetic responders (complete, partial, minor) compared with others. Within each prognostic subset, a cytogenetic response was associated with better survival by landmark analysis. In the most favorable group (stage 1), a cytogenetic response was associated with a 4-year survival rate [from 12 months into therapy] of 79% compared with 62% if no cytogenetic response was obtained (Table 5); P < 0.01). The 4-year survival rates with and without a cytogenetic response in stage 2 were 82% and 35%, respectively (P < 0.01), and in stages 3 and 4 the survival rates were 83% and 39%, respectively (P < 0.01).

In a second approach to evaluating the effect of a cytogenetic response on survival, a time-dependent proportional hazards model was analyzed, with a time-varying term indicating major cytogenetic response as the only predictor of survival (P < 0.001). When the previously identified fixed terms for spleen size and bone marrow basophils were included along with the time-varying term, cytogenetic response retained its association with survival (P < 0.001). The model, including the two pretreatment factors plus the term defining cytogenetic response, is summarized in Table 6. P values for each covariate indicate its statistical significance in the model, adjusting for the remaining two factors. According to the model, patients who achieved a major cytogenetic response were at 0.21 times the risk for death of remaining patients. In a further effort to accurately adjust for prognosis, an analysis (synthesis model) that adjusted for stage of disease (rather than spleen size and bone marrow basophils) gave similar results, showing cytogenetic response to retain its independent association with survival (P < 0.001).

Discussion

Top Methods Results Discussion Author & Article Info References

In patients with solid tumors or acute leukemia, the achievement of a "minimal tumor burden," defined classically as a partial or complete remission, has been associated with survival prolongation. The equivalent of such a minimal leukemic burden in patients with chronic myelogenous leukemia (that is, the achievement of a major cytogenetic response) has not yet been proved to prolong survival. Our study confirms the independent prognostic association between achievement of a cytogenetic response with interferon therapy and survival prolongation in patients with chronic myelogenous leukemia.

Among 274 patients with Ph-positive, early chronic-phase chronic myelogenous leukemia who were treated with interferon--based regimens, 80% achieved a complete hematologic response and 38% had a major cytogenetic response. The estimated median survival for the total study group was 89 months. Achievement of a major cytogenetic response, entered as a time-dependent variable after accounting for other important factors, was associated with survival prolongation (P < 0.001). Similarly, a landmark analysis showed that a cytogenetic response at 12 months was associated with survival (P < 0.001): The 4-year survival rates (dated from 12 months into therapy) were 93% with a complete cytogenetic response, 88% with a partial cytogenetic response, 75% with a minor cytogenetic response, and 50% with no cytogenetic response. The beneficial prognostic effect of achieving a cytogenetic response with interferon- therapy was not restricted to specific risk groups but was evident in all risk groups (Table 5).

The prognostic factors associated with response and survival when receiving interferon- therapy were similar to those of earlier studies. Existing prognostic models were able to segregate patients into risk groups with differences in response to interferon- therapy and in survival outcome (Table 3). The differences were not strong enough to implement risk-oriented therapeutic strategies. However, the models can be used to assess the benefit of newer compared with older regimens within risk groups for chronic myelogenous leukemia.

Our results are strikingly similar to those obtained by the Italian Cooperative Study Group [10]. The median survival was 89 months compared with greater than 72 months when using interferon- therapy. The incidence of a major cytogenetic response was, however, higher in our studies (38% compared with 19%). This may be because of differences in the study groups or dose schedules of interferon- delivered. As noted with chemotherapy trials, single institution studies (restricted number of investigators and faster experience acquisition with the new treatment) generally deliver higher dose schedules of the treatments and obtain better results. In the Italian study, the interferon- dose schedule was reduced by 25% to 50% in 29% of patients and by greater than 50% in 18% of patients; 31% of patients had their interferon discontinued. As suggested by the investigators [21], both factors may have adversely affected the results in the interferon arm. Similar to our study, the Italian group [10] also observed (using landmark analysis) that achieving a hematologic response at 8 months or a major cytogenetic response at 24 months was independently associated with survival prolongation.

A third study [11] by the Cancer and Leukemia Group B of interferon in patients with chronic myelogenous leukemia found a median survival of 66 months with interferon- therapy; the incidence of a complete cytogenetic response was lower than that in our study (13% compared with 26%). In contrast to our findings, and those of the Italian trial, no difference was noted in survival by cytogenetic response [11]. As discussed by the investigators, several factors may have accounted for the different results. Their study included fewer evaluable patients (78 patients) and a lower percentage of patients achieving a complete cytogenetic response (14 patients) who were entered in the analysis of the association. The definition of a partial cytogenetic response was less strict (Ph-positive cells less than 50%), so that the quality of cytogenetic response in measuring minimal tumor burden may be inferior. Twenty-seven percent (29 of 107 patients) of their eligible group was excluded from comparisons of cytogenetic response categories because cytogenetic analyses were not done for patients not achieving certain levels of hematologic response. Finally, the complete hematologic response rate in that study was 22%, whereas the cytogenetic response rate was 29%, which is somewhat unusual. Meaningful cytogenetic responses were seen only in patients who had achieved a complete hematologic response in our study group and in the Italian trial.

We have confirmed the independent importance of obtaining a cytogenetic response for survival prolongation by two statistical methods. A third (and indirect) way to support this observation is to compare the survival of patients receiving different trials who achieved different incidences of a major cytogenetic response. In our study, the major cytogenetic response rate was 38%, and the median survival was 89 months. In the studies by the Italian group and the Cancer and Leukemia Group B, the major cytogenetic response rates were 20% and 13% to 29%, respectively (using different response criteria for major cytogenetic responses of less than 50% Ph-positive cells), and the median survival was 72 months or greater and 66 months, respectively. By extension, studies with interferon- that do not produce major cytogenetic responses may not be associated with survival prolongation compared with conventional hydroxyurea therapy because the survival advantage may be confined only to the cytogenetic responders.

The results from this study have practical applications to the community practice and future investigations. Among patients receiving interferon- therapy at the maximally tolerated individual dose schedule of interferon (not to exceed a daily dose of 5 x 10⁶ U/m²), achieving a cytogenetic response after 12 months would be the determinant for continuing interferon- therapy or for trying a different approach. The results also support the need to develop new programs and to refine existing ones in order to increase the percentage of patients who have major and durable cytogenetic responses with therapy.

We have established the importance of achieving a cytogenetic response with interferon- therapy and the contribution of this response toward survival prolongation, overall and within prognostic risk groups. In addition, we have shown that similar prognostic factors determine the outcome of patients whether they have been treated with interferon- or with other modalities (hydroxyurea, busulfan, combination chemotherapy), and we have validated the relevance of existing prognostic models in predicting outcome with interferon- therapy.

Appendix

Members of the Leukemia Service include Lester Robertson, MD; Charles Koller, MD; Elihu Estey, MD; and Michael J. Keating, MD.