British Journal of Haematology, 1997, 97, 29 37 Health-related quality of life assessed before and during chemotherapy predicts for survival in multiple myeloma FINN WISLØFF 1 AND MARTIN HJORTH 2 for the Nordic Myeloma Study Group 1 Department of Haematology, Medical Clinic, Ullevål University Hospital, Oslo, Norway, and 2 Department of Medicine, Lidköping Hospital, Sweden Received 16 September 1996; accepted for publication 7 January 1997 Summary. Measurement of health-related quality of life was integrated into a randomized trial (NMSG 4/90) comparing melphalan/prednisone to melphalan/prednisone + interferon a-2b in newly diagnosed multiple myeloma. One of the aims of the study was to assess the prognostic significance of quality-of-life scores, using the EORTC QLQ-C30 questionnaire. Univariate analysis showed a highly significant association with survival from the start of therapy for physical functioning as well as role and cognitive functioning, global quality of life, fatigue and pain. In multivariate analysis, physical functioning and W.H.O. performance status were independent prognostic factors (P values ¼ 0 : 001 for both) when analysed in a Cox regression model with the somatic variables b-2 microglobulin, skeletal disease and age. The best prediction for survival from the start of therapy was obtained by combining the b-2 microglobulin and physical functioning scores in a variable consisting of three risk factor levels with an estimated median survival of 17, 29 and 49 months, respectively. At a 12 months landmark analysis, the relative risk for patients with physical functioning score 0 20 v 80 100 was 5. 63 (99% CI 2. 76 11. 49), whereas the relative risk for patients without an objective response to chemotherapy compared to those with at least a minor response was 2. 32 (99% CI 1. 44 3. 74). Quality-of-life assessment may be an independent and valuable addition to the known prognostic factors in multiple myeloma. Keywords: multiple myeloma, quality of life, questionnaire, prognosis, multivariate analysis. During recent years, measurement of patients healthrelated quality of life has been introduced in clinical cancer research as an addition to traditional end points such as response, disease-free survival and overall survival (Weeks, 1992). Extensive research has led to the development of a variety of instruments for quality-of-life assessment in malignant disease. Multiple myeloma is characterized by pronounced distress due to skeletal pain, spontaneous fractures, fatigue and reduced physical functioning. We have previously demonstrated that the quality-of-life questionnaire developed by the European Organization for Research and Treatment of Cancer (EORTC QLQ-C30) is a reliable and valid instrument for the measurement of quality of life in these patients, with a high sensitivity to changes in objective disease status over time (Wisløff et al, 1996a). Using this instrument, we assessed the impact of treatment with interferon on the health-related quality of life in a randomized trial (NMSG 4/90) comparing melphalan/prednisone to melphalan/ prednisone + a-interferon (Wisløff et al, 1996b). Correspondence: Professor Finn Wisløff, Department of Haematology, Medical Clinic, Ullevål University Hospital, 0407 Oslo, Norway. 1997 Blackwell Science Ltd In addition to their value as end points, it has been suggested that quality-of-life scores may be important predictors of survival (Weeks, 1992). Data supporting this suggestion have emerged from studies in lung cancer (Kaasa et al, 1989; Ganz et al, 1991), advanced breast cancer (Coates et al, 1992) and in various terminal cancer patients (Tamburini et al, 1996). One of the aims of the NMSG 4/90 trial was to evaluate the prognostic significance of quality-oflife scores for survival in multiple myeloma. METHODS Patients. From 1 June 1990 to 3 November 1992, 581 patients with newly diagnosed multiple myeloma, recruited from 107 hospitals in Denmark, Finland, Norway and Sweden, were enrolled in the NMSG 4/90 trial. This was a randomized, prospective, unblinded study comparing melphalan/prednisone to melphalan/prednisone + interferon a-2b (Introna, Schering-Plough) 5 million units, subcutaneously, three times weekly. Criteria for the diagnosis, eligibility for the study, characteristics of the study cohort, response criteria, and the results of the trial have been 29
30 Finn Wisløff and Martin Hjorth presented elsewhere (The Nordic Myeloma Study Group, 1996). Clinical data and W.H.O. performance status were recorded before the start of chemotherapy and at 4 6 weeks intervals during treatment. As there was no significant difference in survival between the two treatment arms, all patients are included in the analyses in the present report. All patients were followed until death or until November 1994. None was lost to follow-up. Quality-of-life study. The EORTC QLQ-C30 v 1.0 was used (Aaronson et al, 1993; Wisløff et al, 1996a). This questionnaire incorporates five functioning scales (physical, role, cognitive, emotional, and social), three symptom scales (fatigue, nausea/vomiting, and pain), a global health and quality-of-life scale and a number of single items (dyspnoea, appetite loss, sleep disturbance, constipation and diarrhoea, as well as the financial impact of the disease and its treatment). The questionnaires were presented to the patients prior to treatment and after 1, 6, 12, 24, 36 and 48 months. 524 (90. 2%) of the 581 patients in the NMSG 4/90 trial took part in the quality-of-life study, and 484 (83. 3%) completed all the forms administered to them. Calculation of quality-of-life scores. Each item is graded on an interval scale. The scores for the individual items within each scale were summed, and then divided by the number of items within the scale. All scale and item scores were linearly transformed so that the results ranged from 0 to 100. For the five functioning scales and the single global health/quality-of-life scale, higher scores represent higher levels of functioning. For the symptom and single item scores, higher scores represent higher levels of symptoms (Fayers et al, 1995). Selection of prognostic variables. b-2 microglobulin, age and W.H.O. performance status were chosen on the basis of a previous preliminary analysis (The Nordic Myeloma Study Group, 1996). Based on standard X-ray examinations, the skeletal status had been recorded prior to treatment as normal, limited disease or extensive disease. S-creatinine and s-calcium were significant predictors in the univariate, but not in the multivariate analysis; results are not presented for these two variables. For the analysis of quality-of-life variables with possible prognostic significance, we selected physical and role functioning, fatigue and pain. Previous analysis had demonstrated pronounced impairment in these aspects of quality of life in multiple myeloma (Wisløff et al, 1996a), and our hypothesis was that the reductions in these quality-of-life scores might have prognostic importance. Emotional and social functioning were included because it has been suggested that Table I. Univariate analysis of variables predicting for survival from the start of therapy. No. with Median survival, Variable missing value Category n months (95% CI) P value b-2 microglobulin (mg/l) 34 0. 1 3. 7 161 49 (40 58) < 0. 0001 3. 8 6. 2 168 36 (31 41) 6. 3 62. 0 161 20 (16 24) Skeletal disease (X-ray) 0 Normal 88 51 (not calculable) 0. 0012 Limited 253 32 (28 37) Extensive 183 26 (19 33) Age (years) 0 67 267 38 (32 44) 0. 0118 > 67 255 28 (23 32) W.H.O. performance status 0 0 1 214 39 (33 45) < 0. 0001 2 179 36 (26 46) 3 4 131 17 (11 22) Physical functioning 23 0 20 146 20 (15 25) < 0. 0001 40 60 252 37 (30 44) 80 100 103 43 (33 52) Role functioning 15 0 196 27 (21 33) 0. 0055 50 193 31 (24 38) 100 120 46 (37 55) Cognitive functioning 3 0 50 94 20 (14 27) 0. 0010 60 90 225 32 (26 38) 100 202 39 (31 48) Global quality of life 22 0 40 187 25 (20 30) 0. 0068 41 65 182 38 (30 46) 66 100 133 36 (26 45) Fatigue 10 0 65 192 39 (32 46) 0. 0018 66 100 322 27 (20 34) Pain 8 0 65 274 38 (30 46) 0. 0003 66 100 242 23 (18 27)
psychosocial factors may influence the prognosis of cancer patients (Spiegel et al, 1989). Selection of cut-off points. Age was treated as a continuous variable in the multivariate models and as a dichotomous variable with a cut-off point corresponding to the median age (67 years) in the univariate analysis. Cut-off points for the remaining variables were chosen prior to data analysis (Altman et al, 1994). b-2 microglobulin was divided into three equally sized categories. Skeletal status was used as categorized in the protocol (see above). For W.H.O. performance status and the functioning quality-of-life scores, we formed three categories, in an attempt to obtain three equally sized groups with meaningful cut-off points. For fatigue and pain we used two categories, approximately corresponding to no or a little versus quite a bit or very much of the symptom in question. Sequential numbers were used for coding of the categories; the high-risk category of each variable was coded 0. The number of missing values for each variable is listed in Tables I and III. A new variable was created by computing the sum of the variables b-2 microglobulin and physical functioning, with categories and coding as described above. Statistical analysis. For univariate analysis, Kaplan-Meier estimates were calculated. Survival functions were compared by means of the log rank test. For variables with three ordered categories, the log rank test for trend was used. Variables with P value < 0. 15 in the univariate analysis were included in a multivariate Cox regression model; the likelihood-ratio statistic based on the maximum partial likelihood estimates was used as the criterion for variable removal. The assumption of proportional hazards for two or Quality of Life and Prognosis in Multiple Myeloma 31 more groups was checked by the inspection of log-minus-log survival plots, and by fitting the model with a variable-bytime interaction term (time-dependent Cox model). The landmark analysis was performed according to Anderson et al (1983). The selection of landmark (12 months) was made before the prognostic factor study, on the basis that almost 100% of the patients who eventually responded had done so after 12 months of chemotherapy. Since a large number of variables were assessed for prognostic significance, 99% confidence intervals were used for relative risks. All deaths were treated as events in the survival analyses. The statistical software SPSS for Windows release 6.1.2 was used throughout. RESULTS Survival from the start of therapy The results of Kaplan-Meier survival analysis of the potential prognostic factors are presented in Table I. There were large differences in survival between the three b-2 microglobulin levels. The extent of skeletal disease was also an important predictor of survival. The younger patient group had a survival advantage of about 10 months. W.H.O. performance status had a significant impact of survival. In particular, the category 3 4 carried a poor prognosis. Among the various quality-of-life functioning scores, physical functioning as well as role and cognitive functioning were significantly associated with survival, with a gradient across the three categories. Global quality of life, fatigue and pain were also statistically significant predictors. Emotional and social Table II. Multivariate analysis of variables predicting survival from the start of treatment, in 468 patients with 266 events. For categorical variables, the relative risk compared to the best risk category of the variable is presented. Overall P value Relative Variable for variable Category risk 99% CI b-2 microglobulin (mg/l) < 0. 0001 0. 1 3. 7 1 3. 8 6. 2 1. 24 0. 82 1. 87 6. 3 61. 0 2. 40 1. 62 3. 54 Skeletal disease (X-ray) 0 0017 Normal 1 Limited 1. 64 0. 99 2. 71 Extensive 2. 00 1. 17 3. 29 Age (years) 0. 006 Continuous 1. 02* 1. 00 1. 04 W.H.O. performance status 0. 0014 0 1 1 2 1. 03 0. 69 1. 44 3 4 1. 55 1. 02 2. 36 Physical functioning 0. 0013 0 20 1. 67 1. 05 2. 68 40 60 1. 05 0. 67 1. 64 80 100 1 Cognitive functioning 0. 029 0 50 1. 58 1. 02 2. 44 60 90 1. 17 0. 82 1. 67 100 1 * Relative risk for death, 2% increase per year. Examined in a model with b-2 microglobulin, skeletal disease and age.
32 Finn Wisløff and Martin Hjorth Fig 1. Survival functions calculated from the start of therapy for patient categories defined by b-2 microglobulin (B2m) (A), physical functioning score (PF) (B), and W.H.O. performance status (WHO) (C), at the means of the other covariates in Table II.
Quality of Life and Prognosis in Multiple Myeloma 33 Fig 2. Survival functions (Kaplan-Meier estimates) calculated from the start of therapy for the three categories of the variable created by computing the sum of the b-2 microglobulin and physical functioning values. P < 0 : 0001, log rank test for trend. functioning were not related to survival (P values 0. 35 and 0. 15, respectively). All variables with a P value < 0. 15 in the univariate analysis were included in the multivariate analysis. 468/ 524 patients who took part in the quality-of-life study had complete data on all variables and were available for this analysis. b-2 microglobulin, extent of skeletal disease and age were the only somatic variables with independent Table III. Univariate analysis of variables predicting for survival beyond 12 months. No. with Median survival, Variable missing value Category n months (95% CI) P value Response status 0 Response* 271 Not reached < 0. 0001 No response 113 16 (10 22) W.H.O. performance status 15 0 1 238 Not reached < 0. 0001 2 102 26 (18 34) 3 4 29 7 (1 13) Age (years) 0 67 214 37 (34 41) 0. 0454 > 67 190 26 (18 34) Physical functioning 22 0 20 39 9 (4 15) < 0. 0001 40 60 186 34 (27 40) 80 100 137 Not reached Role functioning 36 0 71 19 (11 27) 0. 0002 50 147 35 (30 40) 100 130 37 (not calculable) Global quality of life 22 0 40 55 20 (5 35) 0. 0044 41 65 123 34 (26 42) 66 100 177 42 (not calculable) Fatigue 35 0 65 283 37 (33 40) 0. 0235 66 100 66 26 (14 37) Pain 35 0 65 289 37 (33 40) 0. 0054 66 100 60 20 (7 33) * Minor, partial or complete response.
34 Finn Wisløff and Martin Hjorth Fig 3. Survival functions estimated from the 12 months landmark for patient categories defined by response to therapy (A), physical functioning (PF) (B), and W.H.O. performance status (WHO) (C), at the means of the other covariates in Table IV.
Quality of Life and Prognosis in Multiple Myeloma Table IV. Multivariate analysis of variables predicting for survival beyond 12 months, in 347 patients with 144 events. Relative risk compared to the category with best risk is presented. 35 Overall P value Relative Variable for variable Category risk 99% CI Response status < 0. 0001 Response* 1 No response 2. 32 1. 44 3. 74 Physical functioning < 0. 0001 0 20 5. 63 2. 76 11. 49 40 60 1. 66 0. 97 2. 83 80 100 1 W.H.O. performance status < 0. 0001 0 1 1 2 1. 60 0. 97 2. 64 3 4 5. 42 2. 83 10. 39 * Minor, partial or complete response. Examined in a model without physical functioning. prognostic significance (Table II). There was a moderately high correlation between W.H.O. performance status and physical functioning (Spearman correlation coefficient 0 : 54). When entered into the multivariate model together, both were significant and independent predictors, with P values 0. 03 and 0. 01 respectively. However, when examined separately with the somatic variables, much smaller P values (0. 001 for both) were found. This discrepancy is probably due to collinearity. It seems clear that W.H.O. performance status and physical functioning have prognostic importance that is independent of the somatic variables, and that these two variables have a similar importance. From the data on relative risk, it appears that the high risk was mainly associated with W.H.O. performance status 3 4 and physical functioning score 0 20. The cognitive functioning score also seemed to have independent prognostic significance, although lower than physical functioning. The impact of b-2 microglobulin, W.H.O. performance status and physical functioning score on survival is evident from Fig 1, which shows estimated survival functions for each category of these variables at the means of the other covariates in the model. The plots confirm that the patients with high b-2 microglobulin, W.H.O. performance status 3 4 or low physical functioning are at high risk. There is little difference between the middle and low risk categories. By coding the three categories of b-2 microglobulin and physical functioning 0, 1 and 2, a summation variable with the possible values 0 4 was formed. The survival was compared for three categories of this variable (0 1, 2 and 3 4) with approximately equal numbers of patients. Median survival for these three groups was 17 (95% CI 13 22), 29 (23 36) and 49 (CI not calculable) months (P < 0 : 0001), and Fig 2 indicates a linear gradient across the categories. Inspection of the cumulative survival curves presented in Fig 1 and the corresponding log-minus-log plots (not shown) suggested that the hazards in the categories of W.H.O. performance status and physical functioning were not proportional, i.e. that the prognostic significance of these variables decreased with time. To test that hypothesis, a variable-by-time interaction was introduced in the regression model. Statistics for this time-dependent Cox model showed that there was a significant interaction between time and W.H.O. performance index (P ¼ 0 : 002) as well as physical functioning (P ¼ 0 : 01). There was no statistically significant interaction between time and b-2 microglobulin. Survival from 12 months In order to study the prognostic significance of quality-of-life scores obtained during treatment, a landmark analysis at 12 months was performed. Univariate analysis (Table III) showed that having obtained a response conferred a significant survival benefit, whereas the effect of age was only marginally significant. The various categories of W.H.O. performance status as well as physical and role functioning, global quality of life, fatigue and pain differed significantly in survival. At this time, cognitive functioning no longer had any prognostic importance. As before therapy, the level of emotional and social functioning had no impact on survival. Multivariate analysis (performed with 347/384 surviving patients) showed that response status, physical functioning and W.H.O. performance status were independent predictor variables (Table IV, Fig 3). A linear gradient across the three categories of physical functioning and W.H.O. performance status was now apparent. W.H.O. performance status was excluded from the Cox model when entered along with physical functioning Unlike the situation prior to therapy, a time-dependent Cox regression analysis did not show any significant interaction between time and any of the covariates. DISCUSSION In multiple myeloma, the disease itself as well as its treatment has a pronounced influence on the patients quality of life. In addition to an adequate follow-up of the biological course of the disease, it is important to assess the patients subjective well-being. The present data show that measurement of health-related quality of life prior to and during treatment
36 Finn Wisløff and Martin Hjorth also contributes important prognostic information. These findings provide indirect support for the validity of the EORTC QLQ-C30 questionnaire for myeloma patients. Before the start of chemotherapy, physical functioning, role functioning, cognitive functioning, global quality of life, and the symptom scales for fatigue and pain were predictors for survival in a univariate Kaplan-Meier model. There was no evidence that emotional and social functioning, as measured by the EORTC QLQ-C30, had any prognostic importance. It has been recognized for several years that performance status, as judged by the physician, incorporates powerful prognostic information in cancer patients (Sorensen & Badsberg, 1990). This is confirmed for multiple myeloma in our study. In multivariate analysis, W.H.O. performance status and physical functioning were powerful predictors for survival from the start of therapy, independent of the somatic variables b-2 microglobulin, age, and extent of skeletal disease. Physical functioning and performance status could be included in the same model, indicating that they are independent prognostic factors. However, as expected, these two variables are correlated, since both are estimates of the patient s physical capacity by the patient himself and by the physician, respectively. When analysed with the somatic variables in separate models, their prognostic significance increased considerably. Cognitive functioning also had independent predictive significance, but far less than physical functioning. To evaluate the prognostic significance of quality-of-life measurements made in the course of the disease, we performed a landmark analysis for patients who had survived the first 12 months. At this time, almost 100% of the patients who responded to chemotherapy had achieved their response. Hence, response status was compared to the quality-of-life indices and to performance status for prognostic importance. Univariate analysis showed significantly better survival for patients who achieved at least a minor response to chemotherapy compared to those who did not, confirming the importance of obtaining a response in multiple myeloma (Bladé et al, 1994). However, multivariate analysis suggested that physical functioning and W.H.O. performance status after 12 months of therapy were at least as important for survival beyond that time as having obtained a response. Although the relative risks and the survival plots indicate that performance status and physical functioning were of equal importance, W.H.O. performance status was excluded from the Cox model when entered along with physical functioning, indicating that the latter was more important. Cognitive functioning reported at the 12 months landmark did not predict for survival beyond this point. Inspection of the survival curves and the corresponding log-minus-log plots (not shown) indicated that the prognostic importance of quality of life and performance status recorded at diagnosis was greatest during the first 1 2 years of follow-up. Time-dependent Cox regression analysis confirmed that the prognostic importance of physical functioning and W.H.O. performance status was time-dependent. However, at the 12 months landmark analysis there was no significant interaction between time and these variables. Our results indicate that the prognostic significance of both these indices was even greater at 12 months than before the start of treatment. Physical functioning emerges as a powerful prognostic factor in multiple myeloma. The high-risk category (score 0 20) corresponds to very poor physical functioning. The middle category (40 60) includes the mean score at diagnosis, which was 47. 9, whereas patients in the lowrisk category (80 100) have a physical functioning comparable to a healthy reference population (M. J. Hjermstad, unpublished data). One of the controversial aspects of prognostic factor studies is the determination of cut-off points for categorical variables. Simon & Altman (1994) warned against the calculation of statistical significance for all possible cut-off points and selecting the one with the smallest significance level, stating that the use of such data-derived cut-off points leads to biased P values and regression coefficients. In the present study we selected the cut-off points before the analysis, trying to form two or three categories for the quality-of-life variables whilst taking into consideration the desirability of equal group sizes as well as meaningful cut-off points. The same cut-off points were used for the 12 months landmark analysis. In spite of our efforts to obtain an unbiased analysis, it is reasonable to assume that our model fits the present sample better than the entire population of multiple myeloma patients. For that reason, we intend to repeat the analysis of these prognostic factors in our on-going trial (NMSG 5/94) comparing highdose chemotherapy with peripheral blood stem cell infusion to conventional chemotherapy, with the same variables and cut-off points. The analysis showed that the relationship between the three-category variables b-2 microglobulin, W.H.O. performance status and physical functioning, and survival from the start of therapy was not linear. The worst score category of these variables before therapy was associated with a considerably increased hazard, whereas the middle and best score categories had approximately similar risks. However, a summation variable for b-2 microglobulin and physical functioning could be created that divided the patients into three equally sized groups with widely different survival. The 12 months landmark analysis suggested a linear gradient across the three categories for the primary variables b-2 microglobulin and physical functioning at this time. Prognostic assessment may be an aid in counselling individual patients, selecting appropriate treatment, stratification in clinical trials and in the comparison of trial results. In multiple myeloma, a great number of parameters have been evaluated for prognostic significance (for review see Kyle, 1994). The varying results that have been published probably reflect different patient selection and the widespread use of data-derived cut-off points. In multivariate analysis, plasma cell labelling index, soluble IL-6 receptor, b-2 microglobulin and performance status seem to be independent predictors for survival (Kyle, 1994). Of these, only b-2 microglobulin and performance status are widely available. We believe
that quality-of-life assessment represents a valuable addition to the known prognostic factors in multiple myeloma. ACKNOWLEDGMENTS We are grateful to Odd O. Aalen, professor in medical statistics, University of Oslo, and to Stein Kaasa, professor in palliative oncology, University of Trondheim, and member of the EORTC Study Group on Quality of Life, for valuable comments on this manuscript. The study was supported by the Schering-Plough Company and by the Norwegian Cancer Society. REFERENCES Aaronson, N.K., Ahmedzai, S., Bergman, B., Bullinger, M., Cull, A., Duez, N.J., Filiberti, A., Flechtner, S., Fleishman, S.B., de Haes, J.C.J.M., Kaasa, S., Klee, M., Osoba, D., Razavi, D., Rofe, P.B., Schraub, S., Sneeuw, K., Sullivan, M. & Takeda, F. (1993) The European Organization for Research and Treatment of Cancer QLQ-C30: a quality-of-life instrument for use in international clinical trials in oncology. Journal of the National Cancer Institute, 85, 365 376. Altman, D.G., Lausen, B., Sauerbrei, W. & Schumacher, M. (1994) Dangers of using optimal cutpoints in the evaluation of prognostic factors. Journal of the National Cancer Institute, 86, 829 835. Anderson J.R., Cain, Kevin C.C. & Gelber, R.D. (1983) Analysis of survival by tumor response. Journal of Clinical Oncology, 1, 710 719. Bladé, J., Lopez-Guillermo, A., Bosch, F., Cervantes, F., Reverter, J.-C., Montserrat, E. & Rozman, C. (1994) Impact of response to treatment on survival in multiple myeloma: results in a series of 243 patients. British Journal of Haematology, 88, 117 121. Coates, A., Gebski, V., Signorini, D., Murray, P., McNeil, D., Byrne, M. & Forbes, J.F. (1992) Prognostic value of quality-of-life scores during chemotherapy for advanced breast cancer. Journal of Clinical Oncology, 10, 1833 1838. Quality of Life and Prognosis in Multiple Myeloma 37 Fayers, P.M., Aaronson, N.K., Bjordal, K. & Sullivan, M. (1995) EORTC QLQ-C30 Scoring Manual. EORTC Study Group on Quality of Life, Brussels. Ganz, P.A., Lee, J.J. & Siau, J. (1991) Quality of life assessment: an independent prognostic variable for survival in lung cancer. Cancer, 67, 3131 3135. Kaasa, S., Mastekaasa, A. & Lund, E. (1989) Prognostic factors for patients with inoperable non-small cell lung cancer, limited disease. Radiotherapy and Oncology, 15, 235 242. Kyle, R.A. (1994) Why better prognostic factors for multiple myeloma are needed. Blood, 83, 1713 1716. Sorensen, J.B. & Badsberg, J.H. (1990) Prognostic factors in resected stage I and II adenocarcinoma of the lung. Journal of Thoracic and Cardiovascular Surgery, 99, 218 226. Simon, R. & Altman, D.G. (1994) Statistical aspects of prognostic factor studies in oncology. British Journal of Cancer, 69, 979 985. Spiegel, D., Bloom, J.R., Kraemer, H.C. & Gottheil, E. (1989) Effect of psychosocial treatment on survival of patients with metastatic breast cancer. Lancet, ii, 888 891. Tamburini, M., Brunella, C., Rosso, S. & Ventafridda, V. (1996) Prognostic value of quality of life scores in terminal cancer patients. Journal of Pain Symptom Management, 1, 32 41. The Nordic Myeloma Study Group (NMSG) (1996) Interferon-alfa- 2b added to melphalan prednisone for initial and maintenance treatment in multiple myeloma: a randomized, controlled trial. Annals of Internal Medicine, 124, 212 222. Weeks, J. (1992) Quality-of life assessment: performance status upstaged? (Editorial). Journal of Clinical Oncology, 10, 1827 1829. Wisløff, F., Eika, S., Hippe, E., Hjorth, M., Holmberg, E., Kaasa, S., Palva, I. & Westin, J. (1996a) Measurement of health-related quality of life in multiple myeloma. British Journal of Haematology, 92, 604 613. Wisløff, F., Hjorth, M., Kaasa, S. & Westin, J. (1996b) Effect of interferon on the health-related quality of life of multiple myeloma patients: results of a Nordic randomized trial comparing melphalan-prednisone to melphalan-prednisone þ a-interferon. British Journal of Haematology, 94, 324 332.