Evaluation of effectiveness: Q-TWiST

Cancer Treatment Reviews (1993) 19 (Supplement A), 73-84

Evaluation of effectiveness: Q-TWiST

Richard D. Gelber”, Aron Goldhirscht and Bernard F. Cole*, for the International Breast Cancer Study Group

*Division of Biostatistics and Epidemiology, Dana-Farber Cancer Institute, 44 Binney Street, Boston, Massachusetts 02115, U.S.A.; tospedale Civico, Servizio Oncologico, 6900 Lugano, Switzerland

Summary

The effectiveness of cancer treatments is often expressed in terms of disease-free survival or overall survival relative risk reduction or odds ratios, and the quality of life effects are often assessed separately from survival. Such end points and summary measures may be inadequate, however, for comparing two treatments in terms of their palliative effects because there is a trade-off between treatment toxicity and increased disease-free interval. Furthermore, this trade-off may depend on individual patient preferences and prognostic situations. The goal of this paper is to describe a method for evaluating the effectiveness of cancer treatments in terms of palliation by simultaneously considering both quality and quantity of time following treatment so that therapeutic choice may be determined according to patient preferences on quality of life and prognostic situation. The method we present is an extension of the Quality-adjusted Time Without Symptoms and Toxicity (Q-TWIST) method for comparing treatment effectiveness in clinical trials of adjuvant therapies. We illustrate an application using data from the International Breast Cancer Study Group Trial V which compares two chemotherapy schedules with different toxicities.

Introduction

The evaluation of cancer treatments in terms of palliation is becoming increasingly important in clinical research. In particular, there is a need to develop methods for comparing the palliative effects of treatment options within randomized clinical trials. Such methods are especially useful in situations where a new treatment is not shown to prolong life significantly, but it may have an advantage to improve or maintain the quality of life of the patient. Unfortunately, such an advantage is difficult to measure and will depend on individual patient preferences as well as known toxicities and other effects on quality of life.

First attempts at assessing the impact of treatments on the quality of life were made by identifying and grading the side-effects of treatment. Subsequent efforts

0305-7372/93/19A0073+12 $08.00/O

73

@ 1993 Academic Press Limited

74 R. D. GELBER ET AL.

have been made to measure patients’ perceptions of the influence of treatment side- effects, and perceptions of symptoms of disease and treatment for relapse. This has led to the development of several instruments for assessing quality of life, and these have been reviewed for their attributes and value for eliciting patient perceptions (l-3). Further efforts focused on the integration of both quality and quantity of life into a single end point which may be used to make treatment comparisons. This led to the development of the Q-TWIST method (4-9). Q-TWIST stands for ‘Quality- adjusted Time Without Symptoms of disease and Toxicity of treatment’ and was originally designed to incorporate aspects of quality of life into adjuvant chemotherapy and endocrine therapy comparisons for the treatment of breast cancer. The reader is referred to the recent articles by Schumacher et a/. (10) and Cox et al. (11) for reviews and recommendations on various methodologies including Q-TWiST for assessing quality of life.

The purpose of this article is to describe how the Q-TWiST method can be used to compare treatments simultaneously in terms of palliation and quantity of life.

The Q-TWIST method

The Q-TWiST method makes treatment comparisons in terms of quality and quantity of life by penalizing treatments which have negative quality of life effects and rewarding those which increase survival and have other positive quality of life effects. As in an ordinary survival analysis, the focus of the method is on time, but rather than look at a single end point such as overall survival or disease-free survival, multiple end points corresponding to changes in quality of life are considered. Periods of time with the negative side-effects of treatment are weighted according to the severity of the side-effects. A weight of 0 indicates the period of time is as bad as death, and a weight of 1 indicates perfect health. Weights between 0 and 1 indicate degrees between these extremes. These weights are called utility coefficients. The Q-TWIST end point is obtained by adding the weighted periods of time.

The application of the method proceeds according to the following three steps:

(a) The first step is to define quality of life oriented survival end points that are relevant for the disease setting under study and highlight specific differences between the treatments being compared in terms of time. For example, in the case of adjuvant chemotherapy for resectable breast cancer, the time with toxicity (TOX) is represented by the period in which the patient is exposed to subjective side-effects of therapy; disease-free survival (DFS) is the time until disease recurrence or death, whichever occurs first; and overall survival (OS) is the time to death from any cause. Time spent living with overt metastatic disease or time in relapse (REL) represents all time after the diagnosis of systemic spread of the disease and is given by REL = OS-DFS. The definitions of TOX and REL reflect the fact that these periods of time have a negative impact on the overall quality of life of the patient. Furthermore, their definition is designed to emphasize the contrasting properties of the different treatments under study. Time without either symptoms of the disease and toxicity of treatment is denoted TWiST = DFS-TOX.

EVALUATION OF EFFECTIVENESS: Q-TWIST 75

(b) The second step in applying the method is to consider each treatment separately and to partition the overall survival time into the defined periods. Areas between the Kaplan-Meier curves (12) for TOX, DFS and OS represent average amounts of time spent in TOX, TWiST and REL. These are calculated up to a specific point in time determined according to the follow- up time of the study cohort. The survival curves for the end points can be plotted on the same graph to illustrate the partitioning according to treatment group. This is known as.a partitioned survival analysis (7).

(c) The third step is to compare the treatment regimens using the weighted sum of the average times calculated in step (b). The quality-adjusted comparison offers the opportunity to include utility coefficients which reflect the assumed value of different periods characterized by different circumstances (TOX and REL) relative to the time without symptoms of disease or toxic effects of treatment (TWIST). Figure 1 displays the different time periods according to assumed utility coefficients of 1 .O for TWiST and 0.5 for both TOX and REL. This example represents a scenario in which 1 month spent in TOX or REL is equivalent to half a month spent with the better quality of life which characterizes TWIST. Specifically, a patient with utility 0.5 for TOX would be willing to trade 1 month of TOX in return for only half a month of TWiST. Gains in average Q-TWiST may be plotted over time by restricting the analysis to yearly intervals leading up to median follow-up. We call such a curve the Q-TWIST gain function.

ln parhIkIr, if uTox and UREL denote the respective utility coefficients for TOX and REL then Q-TWiST is calculated by

Q-TWiST = &Ox xTOX+TWiST+uRELxREL

The utility coefficient for TWiST is 1 .O because it is a period of relatively perfect health. The other utility coefficients express the value of time relative to TWIST. The utility coefficients used in the calculation of Q-TWIST are not often known exactly. In this case, a threshold utility analysis (7) can be performed. This is a type of sensitivity analysis which displays the treatment comparison results for all combinations of the utility coefficients. A threshold line is computed by equating the Q-TWIST formula for the two treatments (i.e., setting Q-TWIST treatment difference equal to zero) and solving for UToX in terms of &EL. The resulting linear equation indicates the pairs of utility coefficients for which the two treatments have equal Q-TWIST. Confidence bands for the threshold line are similarly constructed

Cl-TWiST=uTOXxTOX+TWiST+uRELxREL

l.Oa

.g 0.5. TWiST TOX REL

o.o- I

Time

Figure 7. Components of Quality-adjusted Time Without Symptoms and Toxicity (Q-TWIST), illustrating the division of overall survival into TOX (subjective toxic effects), TWIST, and REL (relapse), and the

weighting of these time periods using utility coefficients u,,, and uREL


by setting the end points of the treatment difference confidence interval equal to zero. Such bands can be used to make statistical inferences. The threshold utility analysis can be displayed in a figure by drawing the threshold line and the 95% confidence bands through a square with uTox ranging from 0 to 1 along the vertical edge and &EL ranging from 0 to 1 along the horizontal edge. The threshold lines divide the utility coefficient square into regions which illustrate the preferred treatment in terms of Q-TWIST for each pair of coefficient values.

Separate Q-TWIST analyses can be performed on subgroups in order to evaluate costs and benefits of a treatment when there is a subgroup effect. Alternatively, covariates and prognostic factors can be incorporated with regression methods, allowing the inclusion of continuous covariates as well as discrete stratifying variables. In most cases, the entire sample of patients can be used to estimate one model, avoiding the problem of decreased sample sizes due to stratification. This has been done with proportional hazards regression models (13). In step (b), the curves for TOX, DFS and OS, are estimated according to patient profiles using the regression model instead of the product limit method. Threshold utility analyses in step (c) are performed for each of the patient profiles, allowing one to evaluate treatment effectiveness under a variety of prognostic situations. This is called a proportional hazards Q-TWIST analysis.

Q-TWIST in the palliative setting

The Q-TWiST method provides a means for comparing treatments in terms of both quantity and quality of life simultaneously. When comparing treatments in terms of palliation, the quality of life experienced is of utmost concern. If the treatments being compared differ on one or more of the quality of life oriented survival end points, then palliative effects should be evaluated with respect to these differences especially if neither treatment has a substantial overall survival advantage. For example, consider two treatments, A and B, which have similar effects on overall survival. Treatment A has more toxicity than treatment B, but it also delays disease recurrence. Effectiveness in terms of palliation depends on the trade-off between the negative quality of life effects of treatment toxicity and positive quality of life effects of delayed disease recurrence.

If one of the treatments has a demonstrated overall survival advantage, the issue of cost-benefit is less pertinent due to the relevant life-years gained. However, unless the benefit is substantial and the treatment appears preferable under all circumstances, its palliative effects should be thoroughly investigated in order to facilitate medical decision making in as many situations as possible. This means taking into account individual patient preferences as well as various prognostic situations.

International Breast Cancer Study Group Trial V: evaluation of a short course of chemotherapy vs. long duration chemotherapy using Q-TWiST

To illustrate the evaluation of treatment effectiveness using Q-TWIST, we applied the method to a randomized clinical trial of adjuvant chemotherapy for resectable


breast cancer. Trial V of the International Breast Cancer Study Group (IBCSG) investigated the effectiveness of 1 month of perioperative systemic treatment compared with a Long duration adjuvant therapy (6 or 7 months) in patients with node-positive breast cancer (9, 14). The Short duration therapy consisted of a combination chemotherapy with CMF (cyclophosphamide, methotrexate, fluorouracil) given entirely iv. on days 1 and 8 after surgery. The majority of the patients with perioperative treatment also received leucovorin. The Long duration treatment regimen consisted of CMFp (oral cyclophosphamide, methotrexate, fluorouracil, plus low dose daily prednisone) for 6 months either following the single perioperative course or initiated 3 to 5 weeks after surgery. Postmenopausal women also received tamoxifen for the 6-month duration of the Long duration chemotherapy. Thus, the Long duration treatment comprised six cycles of chemotherapy for the premenopausal women, and six cycles of chemoendocrine therapy for postmenopausal women, either with or without an initial 4-week course of perioperative chemotherapy.

A total of 1,229 patients were randomized to the two treatments. Of these, 413 patients were randomized to the Short duration treatment, and 816 were randomized to the Long duration treatment. Overall, 715 were premenopausal, and 514 patients were postmenopausal. The median follow-up for this analysis is 7 years.

Figure 2 shows the disease-free survival and overall survival comparisons of the Long duration treatment group vs. the Short duration group. Table 1 gives the 7-year percentages for disease-free survival and overall survival according to treatment group.

Partitioning overall survival

Figure 3 shows the partitioned survival analysis according to treatment group for the IBCSG Trial V. The larger area of TOX and the smaller area of REL are characteristics of the Long duration treatment in terms of time with reduced quality of life. The areas between the curves give the average amount of time spent in TOX, TWiST and REL as indicated.

Table 2 gives the average amounts of time in TOX, TWiST and REL up to 7 years from randomization derived from the partitioned survival analysis. The right-hand column of the Table shows the treatment comparison in terms of differences in the average amount of time patients spend in the various states comparing Long duration chemotherapy minus Short duration chemotherapy. The Q-TWIST calculation was made as an example attributing the utility coefficients of 0.5 to both TOX and REL. These values were arbitrarily selected to illustrate the method and do not represent specific values actually derived from an individual patient preference. Within 7 years, the amount of Q-TWIST gained by the Long duration treatment compared with the Short duration treatment was 5 months, an amount of time gained even after quality of life adjustments for toxic effects and disease relapse.

By restricting the Q-TWiST analysis to yearly intervals leading up to the 7-year analysis, we see how Q-TWIST gains for the Long duration treatment are accumulated over time. This is described by the Q-TWiST gain function shown in Figure 4. The solid line reflects the result for utility coefficients of 0.5 for both TOX and REL. Early in the course of the follow-up, the toxic effects of the Long duration treatment result in a loss in Q-TWIST compared with the Short duration treatment.


(a) 100

60

s

40

(b) IOO

80

60

8

40

20

I I I I I I I I 2 3 4 5 6

Years from randomization

I I I I I I I 0 I 2 3 4 5 6 7


Figure 2. Disease-free survival (a) and overall survival (b) for 1,229 patients with node-positive breast cancer in International Breast Cancer Study Group (IBCSG) Trial V at 7 years of median follow-up.

-, Long duration; -----, Short duration.

This is because the advantages of the Long duration treatment (i.e., increased DFS and OS) do not appear until later on in time. As the benefits are realized with follow-up, the Q-TWiST gain function begins to increase and will continue to increase provided the DFS curves for the two treatments remain separated. The shaded region in Figure 4 illustrates the range of results for the Q-TWIST gain function obtained for different utility coefficient values for TOX and REL between 0 and 1.


Table 1. Seven-year disease-free survival (DFS) and overall survival (OS) percentages according to treatment for 1,229 patients with node- positive breast cancer in International Breast Cancer Study Group Trial V

Chemotherapy treatment 7-year DFS % (SE.) 7-year OS % (S.E.)

Long duration 51 (1.8) 63 (1.8) Short duration 33 (2.5) 50 (2.7)

Log-rank test 2-sided p-value < 0.0001 0.0002

Threshold utility analysis

Clearly, the results of a Q-TWiST analysis depend on the values of the utility coefficients. This is especially true in the palliative setting where treatments have similar effects on overall survival. If the utility coefficients are unknown then a threshold utility analysis should be performed to investigate how the treatment comparison depends on the coefficient values. Figure 5 gives the threshold utility analysis for the IBCSG Trial V data at 7 years. The solid threshold line corresponds to values of &Ox and UREL for which the treatments have equal Q-TWiST. The Long duration treatment has greater Q-TWiST for pairs of utility coefficients which fall above the threshold line, while the Short duration treatment has greater Q-TWIST for pairs of values which fall below the threshold line. The dashed lines give upper and lower 95% confidence bands for the threshold line. The results show that a significant Q-TWIST gain was achieved for a large range of choices for the utility coefficients.

The graph in Figure 5 allows one to determine the treatment preference given a pair of utility coefficient values. For example, for a patient with utility coefficient values of UTox = UREL = 0.5, the Long duration treatment is significantly better and, thus, is the preferred treatment in terms of Q-TWIST. On the other hand, for a patient with Utility Coefficient Values Of uTex= 0.1 and &EL= 0.9, for whom the disutility of toxic effects is great while the disutility of relapse is minimal, the gain in Q-TWiST at 7 years is not significant. In this case, the treatment preference in terms of Q-TWIST is not conclusive. It is important to note that the threshold utility analysis does not indicate the distribution of the utility coefficients for the population. In other words, the threshold line does not tell us how many patients prefer one treatment over the other. This question must be addressed with additional research.

Incorporating prognostic factors

The Q-TWiST method has recently been extended to incorporate prognostic factors using a Cox proportional hazards regression model (13). We performed a proportional hazards Q-TWiST analysis of the IBCSG Trial V data using prognostic factors of: tumor size, age, tumor grade, and the number of lymph nodes involved. These were all significant covariates in the model. Other factors, such as estrogen receptor status, were not included because they were not statistically significant. Threshold utility analyses at 7 years based on the model are presented in Figure 6 for two patient profiles. The two profiles correspond to a good prognosis and a poor prognosis group of 45-year-old patients. The range of utility coefficients


a-”

100

80

60

40

20

0

TWiST

0 I 2 3 4 5 6 7 f


(b) IOO

80

0 I 2 3 4 5 6 7 z Years from rondomization

Figure 3. Partitioned survival for the Long duration treatment (a) and for the Short duration treatment (b) for 1,229 patients in IBCSG Trial V at 7 years of median follow-up. In each graph, the area under the overall survival curve (OS) is partitioned by the survival curves for disease-free survival (DFS) and time with treatment toxicity (TOX). The areas between the survival curves give the average time spent in

TOX, TWiST and REL as indicated.

favoring the more toxicity-intensive Long duration chemotherapy is larger for the poor prognosis group compared with the good prognosis group. This is the case even though relative effectiveness is similar for both patient profiles; i.e., the same percentage reduction in the risk of an event is achieved for good and poor prognosis groups. Figure 6 illustrates how, from a patient’s point of view, the poor prognosis


Table 2. Average months of time within 7 years according to quality of life end point for 1,229 patients in International Breast Cancer Study Group Trial V

End point

Chemotherapy treatment Difference

Long duration Short duration (Long-Short) 95% C.I.

TOX 6 1 5 5- to 5+ TWiST 54 47 6 3to10 REL 9 16 -7 -9to-5 Q-TWIST (uTox = uREL = 0.5) 61 56 5 3 to 8 OS 69 64 5 2 to 8 DFS 59 48 11 8to15

0 I 2 3 4 5 6 7

Years

Figure 4. The solid dark curve gives the average months of Q-TWIST (u,,, = uREL = 0.5) gained for the Long duration treatment compared with the Short duration treatment in IBCSG Trial V as a function of years from randomization (Q-TWiST gain function). The shaded region surrounding the solid curve

shows the ranges for the Q-TWiST gain function as the utility coefficients vary between 0 and 1.

group has the potential to gain more Q-TWIST in the short term (within 7 years), thus increasing the rationale for Long duration chemotherapy.

Conclusions

The evaluation of effectiveness in terms of palliation will become increasingly important in cancer clinical trials. Until a cure is found, new treatments should be evaluated not only for a survival effect but also for possible palliative advantages. The Q-TWiST method is directly applicable and well suited for this because treatments are evaluated simultaneously in terms of quantity and quality of life. It is especially useful in comparing cancer treatments which have similar effects on overall survival, but different effects on quality of life. In particular, Q-TWIST is


1 Longer duration <lg. better

i /

/ #’

/ f

,’ a’

/ a’

,/ .*

Figure 5. Threshold utility analysis for 1,229 patients in IBCSG Trial V. The vertical axis shows the value of UT,,. and the horizontal axis shows the value of uREL. Both u,cx and uREL range between 0 and 1, where the value 1 indicates that the time is worth the same as TWIST, while the value 0 indicates that the time is worth nothing. The solid line is the threshold (based on values of u,,x and uREL) for which the treatments have equal Q-TWIST. The dashed line shows the 95% confidence band for the threshold. The region denoted by ‘Longer duration Sig. better’ indicates the values of utility coefficients for which average Q-TWIST at 7 years after randomization was statistically significantly greater for the Long

duration chemotherapy treatment compared with the Short duration chemotherapy treatment.

Good prognosis

Longer duration

(b)

Longer durotion Sig. better

Figure 6. Threshold utility analyses for two patient profiles based on the proportional hazards model for 1,229 patients in IBCSG Trial V. Threshold diagrams are shown for 45-year-old patients in a good prognostic situation (a) and in a poor prognostic situation (b). The regions denoted by ‘Longer duration Sig. better’ indicate the values of utility coefficients for which average Q-TWiST at 7 years after randomization was statistically significantly greater for the Long duration chemotherapy treatment

compared with the Short duration chemotherapy treatment.


useful for evaluating palliative cancer treatments that involve toxicity. The negative quality of life associated with toxicity is weighed against future quality of life advantages. It is especially important to perform this type of analysis in situations where the administration of a toxic treatment occurs when the patient is free of the disease due to some other treatment (e.g., surgery for breast cancer). The Trial V example illustrates these situations and shows how the Q-TWIST methodology addresses the problems of evaluating palliative effectiveness.

The main advantage of the Q-TWiST method is that it incorporates time into the analysis of quality of life. This can be very important in cancer clinical trials, because the quality of life experienced depends on the amount of time spent with toxicity of the treatment and the time spent with metastatic disease. These are directly affected by the treatment. Other quality of life measures, which do not account for time, would only indirectly reflect benefits of delayed disease recurrence, for example. Furthermore, Q-TWiST does not aggregate quality of life results for an entire population, instead allowing individual patients and physicians to determine the preferred treatment according to individual preferences. This advantage is derived from a threshold utility analysis which gives the preferred treatment according to all combinations of the utility coefficients. In addition, prognostic factors can be included in the analysis by regression methods, allowing the prediction of treatment effects according to different prognostic situations.

Further research must be undertaken, however, to provide some guidance on appropriate values for the utility coefficients. This would assist patients and physicians who are contemplating treatment alternatives. In addition, other methods for assessing quality of life should be investigated. For example, Brunner (15) suggests a point system which incorporates both quality and quantity of life. Four points are awarded for each month following treatment, and points are deducted according to the severity of treatment toxicities and changes in disease status. This would allow patients in a trial to contribute directly to an index of both quality and quantity of life. Daily experiences could be kept in a pocket-sized diary which could be evaluated at regular follow-up visits jointly by the patient and the treating physician.

Acknowledgments

We thank the patients, physicians, nurses and data managers of the International Breast Cancer Study Group who contributed to the clinical trial described in this paper. Financial support for the clinical trial was provided by the Swiss Cancer League, the Cancer League of Ticino, the Ludwig Institute for Cancer Research, the Swedish Cancer Society, the Australia-New Zealand Breast Cancer Study Group, the Frontier Science and Technology Research Foundation, and the Swiss Group for Clinical and Epidemiological Cancer Research. Support for the method- ological development was provided by grant PBR-53 from the American Cancer Society and grant CA-0651 6 from the National Cancer Institute.

References

1. Maguire, P. & Selby, P. on behalf of the Medical Research Council’s Cancer Therapy Committee Working Party on Quality of Life (1989) Assessing quality of life in cancer patients. Br. J. Cancer 60: 437440.


2. Donovan, K., Sanson-Fisher, R. W. & Redman, S. (1989) Measuring quality of life in cancer patients. J. C/in. Oncol. 7: 959-968.

3. Moinpour, C. M., Feigl, P., Metch, B., Hayden, K. A., Meyskens, F. L. & Crowley, J. (1989) Quality of life end points in cancer clinical trials: Review and recommendations. J. Nat/ Cancer inst. 81: 485-495.

4. Gelber, Ft. D. & Goldhirsch, A. (1988) A new endpoint for the assessment of adjuvant therapy in postmenopausal women with operable breast cancer. J. C/in. Oncol. 4: 1772-I 779.

5. Gelber, Ft. D., Gelman, R. S. & Goldhirsch, A. (1989) A quality-of-life-oriented endpoint for comparing therapies. Biometrics 45: 781-795.

6. Goldhirsch, A., Gelber, R. D., Simes, R. J., Glasziou, P. & Coates, A. S. for the Ludwig Breast Cancer Study Group (1989) Costs and benefits of adjuvant therapy in breast cancer. A quality adjusted survival analysis. .I. C/in. Oncol. 7: 36-44.

7. Glasziou, P. P., Simes, R. J. & Gelber, R. D. (1990) Quality adjusted survival analysis. Stat. Med. 9: 1259-I 276.

8. Gelber, R. D., Goldhirsch, A. & Cavalli, F. for the International Breast Cancer Study Group (1991) Quality-of-life-adjusted evaluation of a randomized trial comparing adjuvant therapies for operable breast cancer. Ann. Intern. Med. 114: 621-628.

9. Gelber, R. D., Goldhirsch, A., Hurny, C., Bernhard, J. & Simes, R. J. for the International Breast Cancer Study Group (1992) Quality of life in clinical trials of adjuvant therapies. J. Nat. Cancer Inst. Monographs 11: 127-I 35.

10. Schumacher, M., Olschewski, M. & Schulgen, G. (1991) Assessment of quality of life in clinical trials. Stat. Med. 10: 1915-I 930.

11. Cox, D. R., Fikpatrick, R., Fletcher, A. E., Gore, S. M., Spiegelhalter, D. J. & Jones, D. R. (1992) Quality of life assessment: can we keep it simple? J. R. Stat. Sot., Pert A 155: 353-393.

12. Kaplan, E. L. & Meier, P. (1958) Nonparametric estimation from incomplete observations. J. Am. Stat. Assoc. 541 457-481.

13. Cole, B. F., Gelber, R. D. & Goldhirsch, A. for the International Breast Cancer Study Group (1992) Cox regression models for quality adjusted survival analysis. Stat. Med. (in press).

14. The Ludwig Breast Cancer Study Group (1988) Combination adjuvant chemotherapy for node- positive breast cancer: inadequacy of a single perioperative cycle. N. Engl. ./, Med. 319: 677-683.

15. Brunner, K. W. (1987) Evaluation criteria in comparative clinical trials in advanced breast cancer: a proposal for improvement. In: Cavalli, F., ed., Endocrine Therapy of&east Cancer//. Berlin, Springer- Verlag, pp. 47-51.

Documents

Evaluation of effectiveness: Q-TWiST