Interleukin-6 is a potent lymphoid growth and differentiation cytokine that is produced by various types of cells, including benign and malignant B and T lymphocytes, monocyte macrophages, fibroblasts, and hepatocytes [1-3]. Its pleiotropic activities are reflected by its participation in the physiologic regulation of immune function, inflammatory responses, bone metabolism, and neural processes and by its effects on hematopoiesis [2]. Interleukin-6 has also been implicated in the pathogenesis of several lymphoproliferative disorders, including multiple myeloma, lymphoma, and Castleman disease [4-14], and it may be a prognostic factor for solid tumors, such as renal cell [15], epithelial ovarian [16], and prostate cancer [17].

Our preliminary studies suggested that high serum levels of interleukin-6 correlate with the presence of B symptoms in patients with relapsed lymphoma and are associated with poor outcome in relapsed Hodgkin disease, non-Hodgkin lymphoma, and newly diagnosed diffuse large-cell lymphoma [18, 19]. However, the small number of patients analyzed in these studies precluded determination of whether interleukin-6 levels had independent prognostic importance in a multivariate analysis and suggested the need for an expanded study. We therefore examined the relation between serum levels of interleukin-6 before treatment and outcome in 118 patients with newly diagnosed diffuse large-cell lymphoma.

Methods

Top Methods Results Discussion Author & Article Info References

Patients

All patients 1) had a diagnosis of diffuse large-cell lymphoma or large-cell immunoblastic lymphoma (International Working Formulation categories G and H [20]) confirmed pathologically at our institution; 2) were 16 years of age or older; 3) had not previously received therapy; and 4) had been enrolled in protocols at least 4 months earlier (this criterion was used to ensure uniformity of therapy and adequate follow-up evaluation). Serum samples obtained before therapy began were routinely frozen for all patients who gave consent and were entered into protocols. No serum samples from either the patients or the controls had been previously thawed. Informed consent was obtained in accordance with the guidelines of our institutional review board.

Controls

Serum samples from 45 healthy volunteers were processed and frozen in a manner identical to that used for the serum samples from patients with lymphoma. Both sets of samples were analyzed for interleukin-6 levels simultaneously in a blinded manner. Samples were collected from controls only if the controls had not had fever within 1 week, were not receiving any medications, were not known to be pregnant, and did not have a history of any chronic or acute illnesses.

Initial Staging Evaluation

Routine evaluations done before treatment began included a physical examination; laboratory studies (serum chemistry, measurement of ß₂-µglobulin and lactate dehydrogenase [LDH] levels, complete blood count, differential, and platelet count); chest radiography; computed tomography of the chest, abdomen, and pelvis; whole-body gallium scanning; and bilateral bone marrow aspiration and biopsies. At our institution, the upper limit of normal is 2.0 mg/L (169.5 nmol/L) for ß₂-µglobulin and 618 IU/mL (10.3 µkat/L) for LDH. On the basis of data derived at the time of initial evaluation of the extent of disease, Ann Arbor stage [21] and International Prognostic Index score [22] were determined for each patient.

Treatment

Patients were assigned to treatment on the basis of their prognostic features. Patients with poor prognosis received the alternating triple therapy regimen [23], and patients with more favorable features received a CHOP [cyclophosphamide, doxorubicin, vincristine, prednisone] and bleomycin-based regimen. Poor prognosis was defined as 1) the presence of an elevated serum LDH or ß₂-µglobulin level or a bulky tumor [7 cm] or 2) more than two extranodal sites of disease [24]. The alternating triple therapy regimen consisted of three different chemotherapy regimens administered alternately for a total of nine courses; the three chemotherapy regimens were ASHAP (doxorubicin, methylprednisolone, high-dose cytosine arabinoside, and cisplatin), M-BACOS (high-dose methotrexate, bleomycin, doxorubicin, cyclophosphamide, vincristine, and methylprednisolone), and MINE (mesna, ifosfamide, mitoxantrone, and etoposide).

Follow-up Evaluations

Once treatment began, follow-up physical examinations and blood tests for such tumor markers as LDH and ß₂-µglobulin [25] were done before each course. Complete restaging with appropriate radiologic tests and bone marrow examinations was done at least every 3 months for the first 3 years. On the basis of the results of these tests, response and failure were determined by the attending physician.

Main Study Measures

The relations between interleukin-6 levels and three outcomes-complete response, failure-free survival, and overall survival-were examined. Outcomes were determined through review of patients' medical records. Complete response was defined as the disappearance of all disease for at least 4 weeks. Survival and time to treatment failure were calculated from the date on which the serum sample was collected. Treatment failure was defined as 1) the development of recurrent or progressive lymphoma or death from any cause or 2) removal of the participant from the study because of lack of response. All causes of death were included in the survival analysis. (No deaths were due to totally unrelated causes, such as suicide or accident.)

Interleukin-6 levels were assayed in duplicate by using a validated commercial enzyme-linked immunosorbent assay (Quantikine R & D Systems, Minneapolis, Minnesota), and all values were expressed as the mean of the two measurements. A standard curve was generated by using known-concentrations of recombinant human interleukin-6. All values for interleukin-6 were converted to the NIBSC/World Health Organization standard by using a correction factor according to the manufacturer's instructions. The lower limit of sensitivity of the enzyme-linked immunosorbent assay was 0.7 pg/mL.

Statistical Analysis

Categoric data were compared by using the chi-square or the Fisher exact test, as appropriate [26, 27]; ordinal data were compared by using the Mann-Whitney test or the Kruskal-Wallis test, as appropriate [28]. Survival and time to treatment failure were analyzed by using the Kaplan-Meier method [29], with statistical significance determined by the log-rank test [30]. Patients who did not have treatment failure at the most recent follow-up point were censored at that date in the analyses of time to treatment failure, and patients who were alive at the most recent follow-up point were censored at that date in the analyses of overall survival. For statistical purposes, all interleukin-6 levels that were below the detection limit of the assay were considered to be equivalent (that is, zero).

Multivariate analysis of the time to treatment failure and overall survival data was done by using the Cox proportional-hazards model [31] to assess the predictive value of several factors: age, sex, performance status, presence of B symptoms, Ann Arbor stage, bulky disease, number of extranodal sites, LDH level, ß₂-µglobulin level, interleukin-6 level, and International Prognostic Index score. A multivariate logistic regression analysis was used to identify the independent factors that predict response. Because the two types of treatment (the alternating triple therapy regimen and the CHOP-based regimen) are not equivalent, the proportional hazards analysis was stratified according to treatment and treatment was always included as a variable in the logistic model. For all multivariate analyses, both forward and backward stepwise procedures were used. All P values given are two sided.

Results

Top Methods Results Discussion Author & Article Info References

Interleukin-6 Levels

Thirty-five of the 45 controls (77%) had serum levels of interleukin-6 below the detection limit of the assay. The serum levels of interleukin-6 of the entire group ranged from undetectable to 4.3 pg/mL (median, undetectable) (Table 1). The upper limit of the normal range closely approximates that previously found in a smaller cohort of normal persons [18, 19] when the previous value is converted to the NIBSC/World Health Organization standard.

We identified 118 patients who were treated on protocol and had serum samples stored from the time of diagnosis; all 118 were studied. These patients included the 57 evaluable persons previously reported by our group [19]. Lymphomas were of T-cell origin in only three cases. Serum levels of interleukin-6 in the patients with lymphoma ranged from undetectable to 224.8 pg/mL (median, 4.6 pg/mL) (Table 1) and were significantly higher than those in controls (P < 0.009, Mann-Whitney test) (Table 1).

Univariate Relations among Interleukin-6 Levels, Traditional Prognostic Features, and Outcome

Elevated serum levels of interleukin-6 before treatment correlated with poor prognostic features, such as older age, presence of B symptoms, elevated serum ß₂-µglobulin levels, elevated serum LDH levels, bulky tumor mass, Ann Arbor stage III or IV, poor performance status, and International Prognostic Index score of 2 or more (P 0.02 for all comparisons, Mann-Whitney test) (Table 2).

The overall complete response rate was 80% for all patients. The rate of complete response was significantly higher (P < 0.001, chi-square test) for patients with normal serum levels of interleukin-6 (4.3 pg/mL) (complete response rate, 95%) than for patients with high levels of interleukin-6 (complete response rate, 66%).

The relation between serum levels of interleukin-6 before treatment and failure-free and overall survival is shown in Figure 1, Figure 2, and Figure 3. With a median follow-up period among surviving patients of 31 months (range, 4 to 54 months), the proportion of patients who remain failure-free is significantly higher (P < 0.001, log-rank test) among patients with normal interleukin-6 levels (3-year failure-free survival rate, 77% for persons with normal interleukin-6 levels and 42% for persons with high interleukin-6 levels) (Figure 1). In addition, the relation between increasing serum levels of interleukin-6 and the relapse rate was studied by dividing the cases into quartiles according to serum levels of interleukin-6: 1.45 pg/mL or less, more than 1.45 to 4.63 pg/mL, more than 4.63 to 12.01 pg/mL, and more than 12.01 pg/mL. (The midpoint of 4.63 was the median interleukin-6 level in the patients with lymphoma.) The 3-year failure-free survival rates were 86%, 66%, 46%, and 36%, respectively, for each of these quartiles (P < 0.001, log-rank test).

View larger version (13K): [in this window] [in a new window]   Figure 1. Failure-free survival of participants with high and low interleukin-6 levels. Data were analyzed by using the Kaplan-Meier method. Tick marks indicate the point of last follow-up for one or more participants who did not have treatment failure. Low = an interleukin-6 level of 4.3 pg/mL or less (56 participants had such a level, and 11 of them had treatment failure); high = an interleukin-6 level greater than 4.3 pg/mL (62 patients had such a level, and 35 of them had treatment failure) (P < 0.001, log-rank test).

View larger version (19K): [in this window] [in a new window]   Figure 2. Failure-free survival of participants with favorable and unfavorable prognosis. Data were analyzed by using the Kaplan-Meier method. Tick marks indicate the point of last follow-up for one or more patients who did not have treatment failure. Low = an interleukin-6 level of 4.3 pg/mL or less; high = an interleukin-6 level greater than 4.3 pg/mL. Top. Participants with favorable prognosis (treated with the CHOP [cyclophosphamide, doxorubicin, vincristine, prednisone]-based regimen). Forty-one participants had low interleukin-6 levels, and 6 of them had treatment failure; 17 participants had high interleukin-6 levels, and 7 of them had treatment failure (P = 0.007, log-rank test). Bottom. Participants with unfavorable prognosis (treated with the alternating triple therapy regimen). Fifteen participants had low interleukin-6 levels, and 5 of them had treatment failure; 45 participants had high interleukin-6 levels, and 28 of them had treatment failure (P = 0.02, log-rank test).

View larger version (12K): [in this window] [in a new window]   Figure 3. Overall survival of participants with high and low interleukin-6 levels. Data were analyzed by using the Kaplan-Meier method. Tick marks indicate points at which one or more participants were alive. Low = an interleukin-6 level of 4.3 pg/mL or less (56 participants had such levels, and 6 of them died); high = an interleukin-6 level greater than 4.3 pg/mL (62 participants had such levels, and 28 of them died) (P < 0.001, log-rank test).

The overall survival rate of patients with normal interleukin-6 levels was also superior (P < 0.001, log-rank test) to that of patients with high interleukin-6 levels (Figure 3). The median overall survival rate of patients with high serum levels of interleukin-6 was 35 months; at 3 years, 46% (95% CI, 31% to 61%) are projected to be alive. The median survival rate of patients with a normal serum level of interleukin-6 has not been reached; at 3 years, 90% (CI, 82% to 98%) are projected to be alive.

To determine whether the differences in outcome would be retained within treatment groups, we analyzed outcome within these groups. Those patients with a high serum level of interleukin-6 had an inferior failure-free survival rate whether they received the CHOP-based regimen (P = 0.007) or the alternating triple therapy regimen (P = 0.02) (Figure 2). Similarly, when analyzed according to treatment, the proportion of patients who died remained significantly higher among CHOP recipients with high interleukin-6 levels (3-year survival rate, 77% for persons with high interleukin-6 levels and 100% for persons with normal interleukin-6 levels; P = 0.011) (data not shown). Among persons receiving alternating triple therapy, there was a trend toward inferior overall survival in the subgroup with high interleukin-6 levels, but this did not reach statistical significance (3-year survival rate, 35% for persons with high interleukin-6 levels and 68% for persons with normal interleukin-6 levels; P = 0.08) (data not shown).

Univariate analysis showed that a high interleukin-6 level (>4.3 pg/mL), advanced Ann Arbor stage, high LDH level, and high International Prognostic Index score were all associated with a decreased incidence of complete response (Table 3). Interleukin-6 level, age, Ann Arbor stage, LDH level, ß₂-µglobulin level, International Prognostic Index score, and bulky disease were found in the univariate analyses to be significantly correlated with failure-free survival (Table 3). Finally, the following factors correlated with an increased risk for death: poor performance status, advanced Ann Arbor stage, elevated LDH level, elevated ß₂-µglobulin level, high International Prognostic Index score, B symptoms, bulky disease, and high interleukin-6 levels (P 0.01 for all comparisons).

Because the two treatments (the alternating triple therapy regimen and the CHOP-based regimen) were not considered to be equivalent, the univariate analysis was repeated after adjustment for treatment. For complete response, the statistical method used was logistic regression with treatment included as a covariate in the model. Thus, although this analysis was univariate in the sense that only one factor was examined a time, the statistical method was multivariate logistic because treatment type was also considered. After correction for treatment, only two factors were found to be significantly related to the complete response rate: interleukin-6 level (P = 0.013) and age (P = 0.033). For overall and failure-free survival, treatment was considered in the univariate analyses by using the Cox proportional-hazards model stratified according to treatment. This approach allows the shape of the underlying hazard function to vary with treatment type, if necessary. By using this approach, three factors were found in the univariate analyses to be significantly correlated with poor failure-free survival: older age (P = 0.017), unfavorable International Prognostic Index score (P = 0.006), and high interleukin-6 level (P = 0.004). Finally, four factors correlated with an increased risk for death: age 60 years or greater (P = 0.045), performance status of 3 or greater (P = 0.018), International Prognostic Index score of 3 or greater (P = 0.008), and interleukin-6 level greater than 4.3 pg/mL (P = 0.024).

Response, Failure-Free Survival, and Overall Survival: Multivariate Analysis

Multivariate analysis was done by using a forward as well as a backward stepwise procedure with a significance level of 0.05. The variables included in the analysis as potential prognostic factors were sex, age (included as three variables: age >40 years or not, age >60 years or not, and age >70 years or not), Zubrod performance status (included as three variables: status 1 or not, status 2 or not, and status 3 or not), Ann Arbor stage (included as three variables: stage II or not, stage III or not, and stage IV or not), LDH level (dichotomized at 1 times the upper limit of normal), ß₂-µglobulin level (dichotomized at 1 times and at 1.5 times normal), International Prognostic Index score (included as three variables: score 1 or not, score 2 or not, and score 3 or not), extranodal sites (included as two different variables: 1 site or not and 2 sites or not), B symptoms, bulk (dichotomized at 7 cm) and interleukin-6 level (dichotomized at 4.3 pg/mL).

Because the two treatments (the alternating triple therapy regimen and the CHOP-based regimen) were not considered equivalent, the proportional hazards analysis was stratified according to treatment, allowing for the possibility of a different baseline hazard for each treatment, and treatment was always included as a variable in the logistic model for complete response. Considering treatment in the analysis also allowed identification of prognostic factors independent of known factors because patients were assigned to treatment on the basis of established prognostic features; thus, treatment correlated closely with these prognostic features. When the multivariate logistic analysis (both forward and backward stepwise procedures) was done, the only factor significantly correlated with the incidence of complete response was interleukin-6 level (Table 4). The influence of interleukin-6 level on the complete response rate was not detectably different in the two treatment groups. Ann Arbor stage, LDH level, and International Prognostic Index score, which were selected in the univariate analysis (Table 3) but not retained in the multivariate analysis, were found to be significantly correlated with high interleukin-6 levels (odds ratio, 3.40 [CI, 1.58 to 7.28], P = 0.002 for advanced Ann Arbor stage [III or IV]; 10.9 [CI, 4.38 to 27.1], P < 0.001 for high LDH level; and 7.16 [CI, 3.13 to 16.37], P < 0.001 for unfavorable International Prognostic Index score [2]). This may partly explain the nonsignificance of these variables in the multivariate analysis.

Similarly, with regard to failure-free survival, Cox proportional-hazards multivariate regression analysis with stepwise forward and backward inclusion and exclusion of variables revealed that interleukin-6 was the only significant factor after stratification according to treatment (P = 0.004) (Table 4). International Prognostic Index score came close to having significance if it was forced into the model next after interleukin-6 (relative risk, 1.91 [CI, 0.96 to 3.77]; P = 0.06), and the relative risk for interleukin-6 remained significant with both factors in the model (P = 0.02). Age, Ann Arbor stage, LDH level, ß (₂) -µglobulin level, International Prognostic Index score, and bulky disease were selected in the univariate analysis but not retained in the multivariate analysis for failure-free survival, probably at least in part because of their correlation with interleukin-6 level (Table 2). In addition, because some of these prognostic variables were the basis for choosing a treatment regimen and hence were correlated with treatment type, stratification of the analysis according to treatment probably contributed to the failure to select them. For instance, bulky disease and International Prognostic Index score were both selected as independent variables (in addition to interleukin-6 levels) if the multivariate analysis for failure-free survival was done without taking treatment into account, but only interleukin-6 level was retained in the model when stratification according to treatment was done.

For overall survival, an International Prognostic Index score of 3 or more was the only factor selected, by both the forward and backward multivariate procedures, after stratification according to treatment (Table 4). However, interleukin-6 would have been the next factor entered in the forward analysis (P = 0.08) and was the last factor removed in the backward procedure (P = 0.08); the relative risk for either factor does not change appreciably if one or both are included in the model. Interleukin-6 is therefore only marginally nonsignificant after International Prognostic Index score is accounted for.

International Prognostic Index Stratification by Interleukin-6

For failure-free and overall survival, patients stratified by International Prognostic Index score could be further stratified into two groups with favorable and unfavorable prognosis when analyzed according to interleukin-6 levels (P 0.03 for all comparisons) (Figure 4) (data not shown). In addition, the complete response rates for patients with an unfavorable International Prognostic Index score could be further stratified by interleukin-6 levels. Among the patients with an unfavorable International Prognostic Index score (2), 100% of those with a low serum level of interleukin-6 and 61% of those with a high serum level of interleukin-6 achieved a complete response (P = 0.007, Fisher exact test). Among the patients with a favorable International Prognostic Index score (<2), there was a trend toward higher complete response rates in the subgroup with a low interleukin-6 level (complete response rate, 93%) compared with the subgroup with a high interleukin-6 level (complete response rate, 76%), but this difference did not reach statistical significance (P = 0.06, Fisher exact test).

View larger version (18K): [in this window] [in a new window]   Figure 4. Failure-free survival of participants with good and poor prognosis stratified according to interleukin-6 levels. Data were analyzed by using the Kaplan-Meier method. Tick marks indicate the point of last follow-up for one or more participants who did not have treatment failure. Low = an interleukin-6 level of 4.3 pg/mL or less; high = an interleukin-6 level greater than 4.3 pg/mL. Top. Participants with good prognosis (International Prognostic Index score <2). Forty-four participants had low interleukin-6 levels, and 8 of them had treatment failure; 21 participants had high interleukin-6 levels, and 10 of them had treatment failure (P = 0.009, log-rank test). Bottom. Participants with poor prognosis (International Prognostic Index score 2). Twelve participants had low interleukin-6 levels, and 3 of them had treatment failure; 41 participants had high interleukin-6 levels, and 25 of them had treatment failure (P = 0.01, log-rank test).

Discussion

Top Methods Results Discussion Author & Article Info References

Serum levels of interleukin-6 were previously found to be elevated in both relapsed and newly diagnosed cases of Hodgkin and non-Hodgkin lymphoma [18, 19, 32, 33], and interleukin-6 has been implicated in the pathogenesis of several lymphoid disorders [6-12]. Our study shows that the serum level of interleukin-6 before treatment was an important prognostic factor in our sample of 118 patients with newly diagnosed diffuse large-cell lymphoma. Interleukin-6 levels were commonly elevated in these persons; 52% were categorized as having more circulating interleukin-6 than healthy volunteers (>4.3 pg/mL). High interleukin-6 levels were significantly correlated with several other established clinical and laboratory adverse features, such as B symptoms, elevated serum levels of ß₂-µglobulin or LDH, bulky disease, poor performance status, advanced Ann Arbor stage, age older than 60 years, and an unfavorable International Prognostic Index score (the latter is the most widely accepted prognostic factor model for patients with aggressive non-Hodgkin lymphoma [22]) (Table 2). Of note, the serum level of interleukin-6 was shown to be an independent prognostic factor for complete response rate and failure-free survival. Indeed, in the multivariate regression procedure, interleukin-6 level was the only variable retained by both a stepwise forward and a stepwise backward analysis of prognostic features. An high interleukin-6 value (>4.3 pg/mL) was associated with lower complete response rates (P = 0.013) and an inferior failure-free survival rate (P = 0.004) (Table 4). Furthermore, high interleukin-6 levels were able to identify a subgroup of patients with unfavorable prognosis among those with a favorable International Prognostic Index score (Figure 4, top). The presence of normal compared with high interleukin-6 levels could also be used to further stratify outcome (failure-free survival) in patients considered to have an unfavorable prognosis according to International Prognostic Index score (Figure 4, bottom). In addition, interleukin-6 selected a patient population with a lower complete response rate in the group with International Prognostic Index scores of 2 or more (complete response rate, 61% compared with 100%; P = 0.007), and a similar trend was seen in the group with International Prognostic Index scores less than 2 (complete response rate, 76% compared with 93%; P = 0.06).

Multivariate analysis of overall survival selected only International Prognostic Index scores of 3 or more. However, interleukin-6 would have been the next factor entered in the model with either the forward or backward analysis (P = 0.08) and was the only factor selected if International Prognostic Index score was not included (P = 0.02). Several reasons may explain why other variables, such as International Prognostic Index score, failed to be retained as significant factors in the multivariate analysis of failure-free survival and why the similar lack of retention of interleukin-6 occurred in the overall survival analysis. First, there is codependence between interleukin-6 and Ann Arbor stage, LDH level, age, and performance status (Table 2) (these factors are four of the five components of the International Prognostic Index score). Second, because the P value for International Prognostic Index score (in the multivariate analysis of failure-free survival) was close to significance (P = 0.06), as was the P value for interleukin-6 level in the overall survival model (P = 0.08), it is plausible that the higher statistical power of an analysis with a larger cohort of patients might select these variables as well. Finally, some prognostic variables probably failed to be selected as independent factors because the proportional hazards model was stratified according to treatment (allowing for the possibility of a different baseline hazard for each treatment) and treatment type correlated with established prognostic features (a consequence of patients being assigned to treatment type on the basis of these prognostic features).

In summary, classic predictive models in lymphoma incorporate clinical variables that are believed to be surrogate markers for underlying pathogenic determinants of biologic heterogeneity [34]. Some of the cellular and molecular features recently implicated in the development of aggressive lymphomas and linked to outcome include expression of Ki-67 (a nuclear antigen reflecting tumor cell proliferation), immunophenotype, expression of the CD44 adhesion molecule, and karyotypic and oncogenic aberrations [34-36]. Interleukin-6 is a multifunctional cytokine that may play a role in lymphomagenesis [6-12]. Our current studies indicate that interleukin-6 is an important independent prognostic variable in patients with diffuse large-cell lymphoma and that high interleukin-6 levels correlate with a lower complete response rate and shorter failure-free survival. An important finding is that interleukin-6 seems to be able to further stratify the patients stratified according to scores on the widely accepted International Prognostic Index. On the basis of this data, it may be worthwhile to study the effect of incorporating pretreatment levels of interleukin-6 into the International Prognostic Index score. In addition, we have previously noted that interleukin-6 levels return to normal in patients with Hodgkin lymphoma who achieve complete response after treatment [33]. It may therefore be of interest to determine whether serial monitoring of interleukin-6 levels after remission can be useful for the early prediction of relapse. Finally, evidence indicates that interleukin-6 plays an autocrine or paracrine role, or both, in the growth of lymphoid malignant conditions [12, 14, 37, 38] and that inhibition of interleukin-6 may have antitumor effects [39]. Therefore, identification of patients with high serum levels of interleukin-6 may eventually be relevant to the rational development of novel therapeutic strategies that use cytokine antagonists [40].

Dr. Cabanillas: Department of Hematology, The M.D. Anderson Cancer Center, 1515 Holcombe Boulevard, Box 68, Houston, TX 77030.

Drs. Talpaz and Kurzrock: Department of Bioimmunotherapy, The M.D. Anderson Cancer Center, 1515 Holcombe Boulevard, Box 302, Houston, TX 77030.

Dr. Tucker: Department of Biomathematics, The M.D. Anderson Cancer Center, 1515 Holcombe Boulevard, Box 237, Houston, TX 77030.

Dr. Seymour: Department of Clinical Haematology and Medical Oncology, The Royal Melbourne Hospital, Grattan Street, Parkville 3050, Australia.