15
Survival-weighted health profile for long-term survivors of acute myelogenous leukemia Chiun Hsu 1,2 , Jung-Der Wang 2,3 , Jing-Shiang Hwang 4 , Hwei-Fang Tien 2 , Shueh-Mei Chang 5 , Ann-Lii Cheng 1,2 , Yao-Chang Chen 1,6 , & Jih-Luh Tang 1,2 1 Department of Oncology; 2 Department of Internal Medicine, National Taiwan University Hospital (E-mail: [email protected]); 3 Institute of Occupational Medicine and Industrial Hygiene, National Taiwan University, College of Public Health; 4 Institute of Statistical Science, Academia Sinica; 5 Department of Nursing; 6 Department of Laboratory Medicine, National Taiwan University Hospital, Taiwan Accepted in revised form 30 July 2002 Abstract The objective of this study was to use survival-weighted psychometric scores (SWPS) to construct a ‘health profile’ for long-term survivors of acute myelogenous leukemia (AML). The study cohort included all patients who had been diagnosed and treated in our institution from 1985 to 1999 and achieved complete remission after standard chemotherapy (n ¼ 259). One hundred and four patients were interviewed by the European Organization for the Research and Treatment of Cancer (EORTC)-QLQ-C30 questionnaire and the brief form of World Health Organization quality of life questionnaire (WHOQOL-BREF) to estimate the quality of life (QOL) function of the cohort. Forty-one patients underwent bone marrow transplan- tation (BMT) as consolidation or salvage therapy; 63 received chemotherapy alone. SWPS for every functioning domain and symptom item was obtained by direct integration of the mean QOL function with the survival function of the cohort. A Monte Carlo method was used to extrapolate the life-long SWPS beyond the follow-up limit. The mean scores of EORTC-QLQ-C30 and WHOQOL-BREF did not differ significantly between patients who received BMT or those who received chemotherapy only (p > 0.01). In mean SWPS, patients who received BMT had significantly (p < 0.01) better SWPS in all of the functioning domains and symptom items of EORTC-QLQ-C30 and all four domains of WHOQOL-BREF. However, when the life-long extrapolation of SWPS was made, these differences diminished in global health and several symptom items of EORTC-QLQ-C30 as well as in the social and environmental domains of WHOQOL-BREF. Patients’ perspective on QOL may be domain-specific and may evolve over time. SWPS may be useful to evaluate the efficacy of different treatment strategies for AML. Confirmation of the relative merit of BMT vs. chemotherapy alone from prospective studies is needed. Key words: Acute myelogenous leukemia, Bone marrow transplantation, Chemotherapy, Quality of life, Survival-weighted psychometric score Abbreviations: AML – acute myelogenous leukemia; BMT – bone marrow transplantation; EORTC – the European Organization for Research and Treatment of Cancer; QOL – quality of life; Q-TWiST – quality- adjusted survival time without symptoms of disease and toxicity of treatment; SWPS – survival-weighted psychometric score; WHO – the World Health Organization Introduction Quality of life (QOL) has been increasingly rec- ognized as an important endpoint in the evalua- tion of efficacy of cancer therapy [1, 2]. It is especially informative when alternative treatment options are available and when a trade-off between survival benefit and treatment-related toxicity Quality of Life Research 12: 503–517, 2003. Ó 2003 Kluwer Academic Publishers. Printed in the Netherlands. 503

Survival-weighted health profile for long-term survivors of ... · Survival-weighted health profile for long-term survivors of acute myelogenous leukemia Chiun Hsu1,2, Jung-Der

  • Upload
    others

  • View
    0

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Survival-weighted health profile for long-term survivors of ... · Survival-weighted health profile for long-term survivors of acute myelogenous leukemia Chiun Hsu1,2, Jung-Der

Survival-weighted health profile for long-term survivors of acute myelogenous

leukemia

Chiun Hsu1,2, Jung-Der Wang2,3, Jing-Shiang Hwang4, Hwei-Fang Tien2, Shueh-Mei Chang5, Ann-LiiCheng1,2, Yao-Chang Chen1,6, & Jih-Luh Tang1,21Department of Oncology; 2Department of Internal Medicine, National Taiwan University Hospital (E-mail:[email protected]); 3Institute of Occupational Medicine and Industrial Hygiene, National TaiwanUniversity, College of Public Health; 4Institute of Statistical Science, Academia Sinica; 5Department ofNursing; 6Department of Laboratory Medicine, National Taiwan University Hospital, Taiwan

Accepted in revised form 30 July 2002

Abstract

The objective of this study was to use survival-weighted psychometric scores (SWPS) to construct a ‘healthprofile’ for long-term survivors of acute myelogenous leukemia (AML). The study cohort included allpatients who had been diagnosed and treated in our institution from 1985 to 1999 and achieved completeremission after standard chemotherapy (n ¼ 259). One hundred and four patients were interviewed by theEuropean Organization for the Research and Treatment of Cancer (EORTC)-QLQ-C30 questionnaire andthe brief form of World Health Organization quality of life questionnaire (WHOQOL-BREF) to estimatethe quality of life (QOL) function of the cohort. Forty-one patients underwent bone marrow transplan-tation (BMT) as consolidation or salvage therapy; 63 received chemotherapy alone. SWPS for everyfunctioning domain and symptom item was obtained by direct integration of the mean QOL function withthe survival function of the cohort. A Monte Carlo method was used to extrapolate the life-long SWPSbeyond the follow-up limit. The mean scores of EORTC-QLQ-C30 and WHOQOL-BREF did not differsignificantly between patients who received BMT or those who received chemotherapy only (p > 0.01). Inmean SWPS, patients who received BMT had significantly (p < 0.01) better SWPS in all of the functioningdomains and symptom items of EORTC-QLQ-C30 and all four domains of WHOQOL-BREF. However,when the life-long extrapolation of SWPS was made, these differences diminished in global health andseveral symptom items of EORTC-QLQ-C30 as well as in the social and environmental domains ofWHOQOL-BREF. Patients’ perspective on QOL may be domain-specific and may evolve over time. SWPSmay be useful to evaluate the efficacy of different treatment strategies for AML. Confirmation of therelative merit of BMT vs. chemotherapy alone from prospective studies is needed.

Key words: Acute myelogenous leukemia, Bone marrow transplantation, Chemotherapy, Quality of life,Survival-weighted psychometric score

Abbreviations: AML – acute myelogenous leukemia; BMT – bone marrow transplantation; EORTC – theEuropean Organization for Research and Treatment of Cancer; QOL – quality of life; Q-TWiST – quality-adjusted survival time without symptoms of disease and toxicity of treatment; SWPS – survival-weightedpsychometric score; WHO – the World Health Organization

Introduction

Quality of life (QOL) has been increasingly rec-ognized as an important endpoint in the evalua-

tion of efficacy of cancer therapy [1, 2]. It isespecially informative when alternative treatmentoptions are available and when a trade-off betweensurvival benefit and treatment-related toxicity

Quality of Life Research 12: 503–517, 2003.� 2003 Kluwer Academic Publishers. Printed in the Netherlands.

503

Page 2: Survival-weighted health profile for long-term survivors of ... · Survival-weighted health profile for long-term survivors of acute myelogenous leukemia Chiun Hsu1,2, Jung-Der

needs to be considered. Incorporation of QOLassessment into clinical decision making will thustake into account both the quality and quantity ofsurvival.

Attempts at quantifying and comparing thechange of QOL within clinical trials have beenmade by estimating the quality-adjusted survivaltime. The Quality-adjusted Time Without Symp-toms of disease and Toxicity of treatment (Q-TWiST), first proposed by Glasziou et al. [3], hasbeen used in clinical trials for patients with severalsolid tumors [4, 5] as well as hematological ma-lignancies [6, 7]. According to the Q-TWiST de-sign, the overall survival time is partitioned intoseveral health states based on remission status andtreatment-related toxicity. Each health state is as-signed a utility weight, and the Q-TWiST is cal-culated by summing the utility-weighted durationof each health state. A sensitivity analysis isneeded to provide a range of outcomes to facilitatedecision-making. However, since QOL is a multi-dimensional concept involving the patients’ phy-sical, emotional, and social functioning, a singleutility value, as used in Q-TWiST analysis, maynot reflect all the QOL dimensions that maybe variably affected by the disease and the treat-ment.

An alternative approach is direct integration ofthe patient cohort’s survival function with theQOL function, the latter can be estimated by across-sectional survey of the surviving patients inthe cohort, as proposed by Hwang et al. [8]. Sincemost QOL questionnaires have a multi-dimen-sional construct, the integration of the psycho-metric scores from different dimensions with thesurvival function will generate a survival-weighted‘health profile’ of the patient cohort. Furthermore,the calculation of the survival-weighted psycho-metric scores (SWPS) can be extrapolated beyondthe follow-up limit of the patient cohort to pro-duce life-long estimation of QOL changes [9, 10].Therefore, comparison of projected life-long cu-mulative change of the psychometric scores, i.e.,the ‘survival-weighted’ psychometric scores, maybe feasible and useful in the evaluation of efficacyof different cancer therapies.

High-dose chemotherapy plus allogeneic bonemarrow transplantation (BMT) has an establishedrole in the treatment of acute myelogenous leuke-mia (AML) [11]. It has curative potential for pa-

tients with disease relapse. Besides, BMT is oftenrecommended as consolidation treatment afterinduction chemotherapy for young patients whohave a human leukocyte antigen (HLA)-compati-ble sibling donor [12, 13] and is associated with alower rate of disease relapse and longer disease-free survival [14–16]. However, many studiesindicated that BMT patients may suffer fromprolonged physical as well as psychosocial prob-lems [17–19]. Therefore, the choice betweendifferent treatment strategies will depend onboth the potential survival benefit and the pa-tient’s long-term subjective QOL assessment [20,21].

We herein report the results of SWPS estimationof long-term survivors of adult AML patients. Themain purpose of this study is to explore the fea-sibility of using SWPS as an endpoint to comparethe efficacy of BMT and conventional chemo-therapy.

Patients and methods

Patients

Between January 1985 and December 1999, a totalof 465 adult primary AML patients were diag-nosed and treated at our institution. Two hundredfifty nine of the 347 patients (74.6%) who hadreceived standard induction chemotherapyachieved complete remission and these patientsconstituted the study cohort. The cohort consistedof 137 men and 122 women, with a median ageat diagnosis of 35.7 years (range: 14.1–77.7 years). As of December 2000, 134 pa-tients died during follow-up and 18 lost to follow-up.

The design of the current study is to estimate theQOL function of the study cohort by a cross-sec-tional interview of the surviving patients in thatcohort (see below). Contact in person on the pa-tients’ visit to the clinic or by telephone was madefor all the 107 surviving patients to explain thepurposes and process of this study. Three patientsrefused further contact. Written informed consentwas obtained from the other 104 patients and thesepatients form the basis of the questionnaire inter-view.

504

Page 3: Survival-weighted health profile for long-term survivors of ... · Survival-weighted health profile for long-term survivors of acute myelogenous leukemia Chiun Hsu1,2, Jung-Der

Questionnaire interview

Two QOL questionnaires were used in this study.One is the European Organization for Researchand Treatment of Cancer (EORTC) Quality ofLife Questionnaire-C30 version 3.0 (QLQ-C30 v3.0), a questionnaire designed specifically for can-cer patients [22]. The 30 items in EORTC-QLQ-C30 were grouped into several functioning do-mains (global health, physical, emotional, role,cognitive, and social functioning) and symptomitems that are related to the underlying cancers ortheir treatment. The Taiwanese Chinese version ofQLQ-C30 v 3.0 was verified by EORTC afterstandard forward–backward translation processand field test (Dr Wei-Chiu Chie, College of PublicHealth, National Taiwan University). The EO-RTC-QLQ-C30 questionnaire uses a Likert-typescale, an ordinal one, to grade patients’ judgmenton different aspects of QOL. The ‘scalability’of EORTC-QLQ-C30 has been studied by usinga non-parametric Mokken scaling model [23,24]. The result indicated that the scalability ofEORTC scales was quite satisfactory except forthose of role functioning and cognitive function-ing.

The other questionnaire is WHOQOL-BREF, ageneric questionnaire developed by the WorldHealth Organization. WHOQOL-BREF is a briefform of the original questionnaire WHOQOL-100and consists of 26 items that are grouped intophysical, psychological, social, and environmentaldomains [25, 26]. WHOQOL-BREF also uses aLikert-type, 5-point scale to grade the patients’response to QOL items. To ensure scalar equiva-lence among different language versions of thequestionnaire, a descriptor study was designed tospecify the ‘anchor points’ for the different types ofresponse scales (capacity, intensity, frequency,evaluation). Between the extreme ‘anchors’ of, say,‘never-always’ or ‘not at all-completely’, interme-diate descriptors that best correspond to the 25,50, and 75% points between the two anchors wereidentified [27]. We have done the descriptor studyaccording to WHOQOL guidelines for the Tai-wanese version of WHOQOL-BREF that was usedin the present study [28].

Pertinent demographic data were obtained byan accompanying questionnaire. Eighty-four pa-tients filled the questionnaires in the clinic while

waiting to see the physician. Twenty patients didnot come back to the hospital during the studyperiod for personal reasons and the questionnaireswere sent and retrieved by mail.

The clinical characteristics of the 104 inter-viewed patients were summarized in Table 1. As ofDecember 31, 2000, the median follow-up was65.4 months (range: 15.6–189.8 months) and54.4 months (range: 12.5–153.9 months) for pa-tients in the BMT and chemotherapy group, res-pectively (log rank test p ¼ 0.07). Thirty-sevenpatients received allogenic and four receivedautologous transplantation. The disease status atthe time of BMT was first complete remission (29patients), second complete remission (11 patients)and second relapse (1 patient).

The scoring procedure was according to EO-RTC and WHOQOL guidelines, respectively [29,30]. In EORTC-QLQ-C30, higher functioning(physical, emotional, role, cognitive, and social)scores indicate a better level of functioning, whilehigher symptom (fatigue, pain, nausea/vomiting,insomnia, appetite loss, constipation, diarrhea andfinancial difficulty) scores indicate a worse level ofsymptoms. In WHOQOL-BREF, a score trans-formation was done for the ‘negative phrasing’items; so higher scores always indicate better QOL.Each item scores was transformed to a 0–100score. The results of the scores between patientgroups were compared by ANOVA test.

Estimation of survival-weighted psychometric scores(SWPS)

The theoretical basis of SWPS estimation was de-scribed in details previously [8]. The basic conceptis to consider SWPS as a direct integration of thecohort’s QOL function and the survival function.Let E½qðtÞ� the mean utility value of patients whoare still alive at time t. This mean utility can beestimated by a random sample of a cross-sectionalsurvey on the currently surviving patients in thiscohort. The survival function of the whole cohort,denoted by S(t), was calculated by Kaplan–Meiermethod. The estimated SWPS may then be ex-pressed by the following formula:

EðSWPSÞ ¼Z 1

0

E½qðtÞ�SðtÞ dt

505

Page 4: Survival-weighted health profile for long-term survivors of ... · Survival-weighted health profile for long-term survivors of acute myelogenous leukemia Chiun Hsu1,2, Jung-Der

The expected SWPS is then the area under thequality-adjusted survival curve. The standard er-ror of the quality-adjusted survival time was esti-mated by a bootstrap method [31].

A Monte Carlo method was designed to projectthe estimation of SWPS beyond the follow-uplimits [9]. The basic assumption of this approachwas (1) a ratio W exists between SWPS curves ofthe study cohort, SWPSstudy(t), and the referencepopulation, SWPSref(t), and (2) the logit transformof this ratio W will become linear after a period offollow-up, usually after intensive diagnostic andtherapeutic intervention have been completed. Areference population with known hazard function,with age and sex matched with the study cohortwas generated from the vital statistics data ofTaiwan (Department of Health, Taiwan, ROC,1995). The utility of the reference populationwas set to be 1 (normal health). The relationship

between SWPSstudy(t) and SWPSref(t) was thencalculated by a linear regression of the logit ofSWPSstudy(t)/SWPSref(t). Finally extrapolationbeyond the follow-up limit was done by the re-gression model and the survival function of thereference population. The method was furtherapplied to all the psychometric scores with thesurvival rate at each duration-to-date as theweights (J.S. Hwang and J.D. Wang, submitted).

To use standardized psychometric scores torepresent different aspects of utility, the domainscores of WHOQOL-BREF and the functioningscores of EORTC-QLQ-C30 were divided by 100.The symptom scores (SS) of EORTC-QLQ-C30were transformed by the following formula:

Scoresymptom ¼ ð1� SSÞ=100This transformation results in a 0–1 score in

which 1 represents normal and 0 the worst health

Table 1. Clinical characteristics of the interviewed patients

BMT group N = 41 Chemotherapy group N = 63

Age O50 years

N = 45

All patients

N = 63

Male/female 22/19 18/27 28/35

Age at diagnosis (years) (median/range) 28.8/15.1–42.1* 33.3/16.9–49.8* 41.1/16.9–65.9

Age at interview (years) (median/range) 34.3/20–51.4** 39.2/19.8–58.4** 45.9/19.8–69.6

Disease status at interview

First CR 29 31 42

Second/third CR 12 12 19

Second relapse 0 2 2

Time from last chemotherapy to interview (years)

O2 11 18 25

>2 30 27 38

Education

Elementary school 1 1 8

High school 11 22 30

College/university and up 29 22 25

Employment at the time of interview

Paying job/self-employed/studying 33 32 38

Working at home/unemployed 8 13 25

Personal income (NT dollar a/month)

<10,000 11 13 25

10,000–29,999 10 9 12

30,000–59,999 11 16 19

P60,000 9 7 7

Abbreviation: BMT – bone marrow transplantation; CR – complete remission.a The exchange rate of US dollar: NT dollar is about 1:32.

* p = 0.004 compared with patients who received BMT.

** p = 0.04 compared with patients who received BMT.

506

Page 5: Survival-weighted health profile for long-term survivors of ... · Survival-weighted health profile for long-term survivors of acute myelogenous leukemia Chiun Hsu1,2, Jung-Der

status, respectively. We then applied the abovemethod to obtain SWPS from all the functioningand symptom items and extrapolate to assess life-long cumulative changes.

A computer program has been designed to cal-culate the estimated SWPS using S-PLUS 2000software (MathSoft Inc.) (available at http://www.stat.sinica.edu.tw/jshwang). Comparison ofSWPS between treatment groups was done by Zstatistics. Because of multiple comparisons, wetook p ¼ 0.01 as the threshold of statistical sig-nificance.

Results

Survival analysis of AML patients

The median survival of the 259 patients who hadachieved CR after induction chemotherapy was37.5 months (95% CI: 22–53.1 months). In ourinstitution high-dose chemotherapy and BMT wassuggested only for patients aged 50 years oryounger. Sixty-eight patients had received BMT,41 patients as consolidation therapy at first CRand 27 patients as salvage treatment after relapse.The 5-year overall survival was 60.5% (95% CI:47.8–73.2%) and 34.3% (95% CI: 24.9–43.6%) forpatients who received BMT and patients who re-ceived chemotherapy only (n ¼ 191) (Figure 1A,p ¼ 0.002). Similar pattern was seen for the pa-tients aged 50 years or less (n ¼ 203) (Figure 1B,p ¼ 0.002).

QOL analysis by questionnaire interview

The mean scores of EORTC-QLQ-C30 andWHOQOL-BREF were summarized in Table 2. InEORTC-QLQ-C30 scores, patients who had re-ceived BMT appeared to have less financial diffi-culty (p ¼ 0.03). This may not indicate truesignificant difference because of multiple compari-sons. Both groups appeared to have lower levelsof emotional and cognitive functioning. The mostprominent symptoms in both groups were fatigueand insomnia. In WHOQOL-BREF scores, nosignificant difference in any of the four domainswas found between patients who received BMT orchemotherapy alone. There were no significantdifference of the QOL scores related to patients’

age, sex and educational levels, as determined byANOVA test.

Because some QOL domains may change withtime, we stratified the patients by the length oftime between last chemotherapy and questionnaireinterview (<1 year, 25 patients; 1–2 years, 11 pa-tients; 2–5 years, 36 patients; >5 years, 32 pa-tients). A significant trend of QOL improvementover time was seen in physical, social functioningand financial difficulty (linear regressionp < 0.01), while borderline significance was foundin role functioning (p ¼ 0.02).

Estimation of survival-weighted psychometric scores

To make the comparison of SWPS between pa-tients who received BMT and those who receivedchemotherapy more clinically relevant, only theQOL data from patients aged 50 years or less atdiagnosis (n ¼ 45), i.e., patients who were poten-tially eligible for transplantation, were used forcalculation of SWPS. The estimation and extra-polation of SWPS were illustrated in Figure 2.Figure 2A and B represented the survival curvesand the QOL function curves for patients whoreceived BMT or chemotherapy, respectively. Thehatched area under the curves indicated the cu-mulated SWPS. Figure 2C and D demonstratedthe relationship of logit of W, which stands forSWPSstudy(t)/SWPSref(t), with time. Figure 2E andF showed comparison of SWPS curves betweenthe study cohort (2E, BMT patients; 2F, chemo-therapy patients) and the reference population.The relationship between the SWPS curves of thestudy cohort and of the reference population wasestablished by a linear regression of the logit of W.Extrapolation beyond the follow-up limit wasdone by the regression model and the survivalfunction of the reference population. The rela-tionship of SWPS curves between the study cohortand the reference population over time has asimilar pattern in all of the functioning domainsand the symptom items. The logistic transforma-tion of this ratio has an initial period of fluctuationfor 3–4 years and becomes linear thereafter (datanot shown). This supports our hypothesis in theextrapolation of SWPS estimation.

The estimated QOL curves for representativefunctioning domain and symptom items, and the

507

Page 6: Survival-weighted health profile for long-term survivors of ... · Survival-weighted health profile for long-term survivors of acute myelogenous leukemia Chiun Hsu1,2, Jung-Der

resulting quality-adjusted survival curves, areshown in Figure 3. There may be several patternsof QOL evolution over time, as illustrated by usingEORTC-QLQ-C30 results. Symptoms such asnausea, vomiting and appetite loss did not appearto cause significant long-term problems. The neg-ative impact of fatigue and insomnia, by contrast,

was more persistent. The social and role func-tioning levels generally improved over time, sug-gesting adjustment by the patients themselves.Improvement in global health, emotional andcognitive functioning, on the other hand, was notevident, indicating potential chronic toxicity andincomplete psychological adaptation. In compari-

Figure 1. Kaplan–Meier overall survival curves of the whole study cohort, stratified by consolidation therapy after first complete

remission. (A) all patients (n ¼ 259) (log rank test p ¼ 0.002); (B) patients aged 50 years or less (n ¼ 203) (p ¼ 0.002).

508

Page 7: Survival-weighted health profile for long-term survivors of ... · Survival-weighted health profile for long-term survivors of acute myelogenous leukemia Chiun Hsu1,2, Jung-Der

son, the QOL change of the WHOQOL-BREFdomains was not obvious.

Comparisons of mean and life-long extrapolatedSWPS were listed in Table 3. In mean SWPS, pa-tients who received BMT had significantly betterSWPS in all of the functioning domains andsymptom items of EORTC-QLQ-C30 and all do-main scores of WHOQOL-BREF. This is mostlikely due to the effect of ‘weighting’ by the sur-vival advantage of BMT in this series. However,these differences diminished when the life-longextrapolation was made in global health and sev-eral symptom items of EORTC-QLQ-C30 as wellas in the social and environmental domains ofWHOQOL-BREF. Comparisons of the quality-adjusted survival curves were shown in Figure 4.The curves for BMT patients lied above those for

chemotherapy patients in all functioning domainsand symptom items. This suggested that patientswho received BMT generally had a higher life-longcumulative SWPS than those who received che-motherapy in this series.

Discussion

In this study we demonstrated a new method tocalculate SWPS for long-term survivors of patientswith AML. By direct integration of QOL datafrom the EORTC-QLQ-C30 and WHOQOL-BREF questionnaires with the survival function,this method offers a multi-dimensional healthprofile from the patients’ perspective. The resultssuggest that patients who received BMT may enjoy

Table 2. Mean scores of questionnaire interview

BMT group N = 41 Chemotherapy group N = 63

Age O50 years

N = 45

All patients

N = 63

EORTC-QLQ-C30 scoresa

Global health 68.3 ± 19.6 73.7 ± 19.9 74.7 ± 20.0

Functional scales

Physical 84.7 ± 15.2 84.4 ± 18.4 83.8 ± 17.7

Emotional 79.7 ± 18.6 78.7 ± 17.4 80.3 ± 19.3

Cognitive 75.2 ± 23.3 77.8 ± 20.7 75.9 ± 21.3

Role 85.8 ± 21.6 86.7 ± 21.5 86.8 ± 21.0

Social 82.9 ± 20.2 74.1 ± 27.2 75.7 ± 25.4

Symptom scalesb

Fatigue 29.3 ± 19.0 31.4 ± 23.1 29.6 ± 22.4

Pain 14.2 ± 16.5 18.1 ± 18.7 18.0 ± 19.5

Nausea/vomiting 5.7 ± 10.3 9.2 ± 16.1 9.2 ± 18.6

Symptom itemsb

Appetite loss 13.8 ± 19.7 15.6 ± 24.2 16.4 ± 27.4

Constipation 7.3 ± 19.0 11.1 ± 18.8 10.1 ± 18.6

Diarrhea 20.3 ± 20.9 16.3 ± 23.2 14.3 ± 21.3

Dyspnea 18.7 ± 26.9 11.8 ± 19.0 12.2 ± 21.8

Financial difficulty 13.8 ± 18.2 25.9 ± 28.3** 24.9 ± 29.9**

Insomnia 26.8 ± 27.1 29.6 ± 30.3 29.1 ± 31.4

WHOQOL-BREF

Domain 1 (physical) 65.8 ± 13.4 66.0 ± 11.9 65.9 ± 12.6

Domain 2 (psychological) 65.5 ± 14.8 64.4 ± 13.5 65.0 ± 14.4

Domain 3 (social) 61.3 ± 17.0 63.6 ± 15.9 63.7 ± 14.5

Domain 4 (environmental) 68.1 ± 13.1 66.2 ± 13.2 66.9 ± 12.7

* 0 – 100 standardized scores.aHigher scores designate better function level.bHigher scores designate worse symptoms.

** p = 0.03 compared with patients who received BMT.

509

Page 8: Survival-weighted health profile for long-term survivors of ... · Survival-weighted health profile for long-term survivors of acute myelogenous leukemia Chiun Hsu1,2, Jung-Der

Figure 2. Estimation and extrapolation of SWPS. Results from the global health domain of EORTC-QLQ-C30 was used in the

calculation. (A) and (B), the survival function curve (solid lines) and the QOL function curve (dashed lines) for patients who received

BMT or chemotherapy alone. The hatched area under the curves represents the cumulated SWPS. (C) and (D), the logit of W curve

over time for the two groups of patients. W represents the ratio of SWPS between the study cohort and the reference population.

(W ¼ SWPSstudy(t)/SWPSref(t)). (E) and (F), Extrapolation of SWPS curves beyond the follow-up limit. The upper curves in the two

graphs represent SWPS curves of the reference population, which equal to the overall survival curve because the QOL value was set to

be 1 (normal health). The lower curves represent the SWPS curves of the study cohort.

510

Page 9: Survival-weighted health profile for long-term survivors of ... · Survival-weighted health profile for long-term survivors of acute myelogenous leukemia Chiun Hsu1,2, Jung-Der

better SWPS than patients who received chemo-therapy alone. This is consistent with findingsfrom the Q-TWiST analysis of a clinical trial forpediatric AML [6].

The limited number of patients and the non-randomized allocation of treatment in the presentstudy make it impossible to reach a conclusion onthe relative benefit of BMT over conventionalchemotherapy in this setting, i.e., AML in firstcomplete remission. It is likely that in the presentcohort, the survival difference between patientstreated by BMT or chemotherapy alone may playa more important role than the potential QOLdifference in the final SWPS results. The true sur-vival advantage of BMT is still a matter of debatein the oncology community because randomizedclinical trials cannot demonstrate significant ben-

efit of BMT in overall survival [16, 20]. Our SWPSmodel can be directly combined with data fromrandomized trials of AML to compare the SWPSdifference of different treatment strategies, becausethe QOL function of the whole cohort can beobtained by a cross-sectional survey of the sur-viving patients. This is analogous to the develop-ment of the Q-TWiST analysis for quality-adjustedsurvival time.

Our approach can supplement the Q-TWiSTmethod in two aspects. First, since different facetsof QOL may be variably affected by the diseasecourse and treatment-related toxicity, a multidi-mensional assessment of psychometric scores forevery QOL facet may provide a more comprehen-sive and holistic view than a single utility value can.For example, acute treatment- or disease-related

Figure 3. Mean SWPS of AML patients who received BMT or chemotherapy alone. The QOL values used are (A) representative

functioning and items of EORTC-QLQ-C30 and (B) domain scores of WHOQOL-BREF. Solid line, overall survival function; dashed

line, mean QOL function. Hatched area indicated the estimated SWPS.

511

Page 10: Survival-weighted health profile for long-term survivors of ... · Survival-weighted health profile for long-term survivors of acute myelogenous leukemia Chiun Hsu1,2, Jung-Der

symptoms, such as nausea/vomiting and dyspnea,may resolve during follow-up. On the other hand,patients may suffer from prolonged psychologicalas well as physical distresses that may or may notrecover even after long-term disease remission isachieved [17–19]. Other investigators also pointedout that multi-dimensional measures could bettercapture the QOL change after therapeutic inter-vention, thus increase the sensitivity and respon-siveness of QOL as an endpoint of treatmentefficacy [32]. Second, changes in patients’ judgmenton utility may occur, even in a time ‘withoutsymptoms of disease and toxicity of treatment’, asdefined in Q-TWiST. This is especially obvious inaspects in which patients’ adaptation may play animportant role, such as social and role functioning.Taken together, the comparison of ‘health profiles’consisting of SWPS may provide a new endpoint inclinical decision making.

It will be interesting to know which aspects ofQOL are most relevant to the patients’ generalperception of health status. A recent study hasdemonstrated that the changes of QOL scores overtime corresponded to the severity of diseases andtreatment courses in individual cancer patients.The QOL profile was especially informative onphysical performance, emotional distress and so-cial function [33]. For patients who received BMTthe most prominent long-term sequels include fa-tigue and emotional distress [19]. These enduringcomplications may disproportionately influencethe patients’ overall judgment on health status andutility.

It is arguable to use the psychometric scoresfrom Likert-type questionnaires such as EORTC-QLQ-C30 or WHOQOL-BREF to represent thepatients QOL values, because these scores are nottrue interval scores, and thus comparisons among

Figure 3. (Continued)

512

Page 11: Survival-weighted health profile for long-term survivors of ... · Survival-weighted health profile for long-term survivors of acute myelogenous leukemia Chiun Hsu1,2, Jung-Der

different items may be difficult. The descriptorstudy required by WHOQOL group, as describedabove, is aimed at solving the problem of compa-rability. As for EORTC-QLQ-C30, studies byother investigators have shown satisfactory ‘sca-lability’ of most of the functioning domains andsymptom items, although improvement in reli-ability may be needed for domains that relied onthe scores of only one to two items [23, 24]. To

improve the accuracy of SWPS comparison, futureQOL research may incorporate instruments usingtrue interval scales.

A second question of using these questionnairesis that the items may overlap with one another. Inthe development of EORTC-QLQ-C30, the vali-dity of the individual items was demonstrated byexamination of the convergent and discriminantvalidity and the inter-scale correlation. For example,

Figure 3. (Continued)

513

Page 12: Survival-weighted health profile for long-term survivors of ... · Survival-weighted health profile for long-term survivors of acute myelogenous leukemia Chiun Hsu1,2, Jung-Der

the items of physical functioning are correlatedwith those of social functioning and fatigue. Bycontrast, items that are conceptually not related,such as those of cognitive functioning and symp-toms of nausea or diarrhea, have been demon-strated to have low correlation. Therefore, itemsthat are clinically correlated with one anothershould behave similarly in the results of SWPSestimation, while items that are not clinically cor-related should differ from one another. In this re-gard, we suggest that the different ‘patterns’ ofQOL evolution, as shown in Figure 3, may trulyreflect the clinical situation and may help bothpatients and clinicians understand what mayhappen at different time points during the clinicalcourses. In comparison, the four domains of theWHOQOL-BREF, which are conceptually distinctfrom one another in their impact on QOL judg-ment, have been shown to have satisfactory dis-criminant validity and spillover, though notcompletely avoidable, should be minimal.

Other methods for extrapolation of survivaldata beyond follow-up limits have been proposed.In a cost effectiveness analysis of the data from theGlobal Utilization of Streptokinase and TissuePlasminogen Activator for Occluded CoronaryArteries (GUSTO) trial, the investigators assumedthat, no matter what thrombolytic therapy wasused, the patients’ long-term survival was similarwith that of chronic, stable coronary artery disease[34]. The 1-year survival data in GUSTO trial wasextended by 14 years based on the experience ofDuke Cardiovascular Disease Database. A Gom-pertz function was used for extrapolation to obtainlifetime survival curve. This approach is concep-tually similar with ours. Since a reference databaseis seldom available, we proposed to use the vitalstatistics data from the general population, whichshould be an informative reference population formost long-term follow-up studies of chronic dis-eases. A feedback check for the linearity assump-tion of logit of quality adjusted survival ratio

Table 3. Estimated SWPS

Mean SWPS (mean ± SE) Projected life-long SWPS (mean ± SE)

BMT Chemotherapy* p-Value BMT Chemotherapy* p-Value

EORTC-QLQ-C30

Global health 83.1 ± 9.8 52.0 ± 6.4 0.005 183.2 ± 28.7 121.2 ± 18.9 0.04

Functional scales

Physical 103.9 ± 10.2 60.4 ± 6.4 <0.001 237.3 ± 30.0 147.3 ± 19.1 0.006

Emotional 96.3 ± 10.6 55.4 ± 4.9 <0.001 213.8 ± 30.4 125.4 ± 13.9 0.004

Cognitive 89.7 ± 8.7 55.4 ± 6.4 <0.001 191.9 ± 24.4 127.6 ± 18.4 0.02

Role 105.9 ± 10.6 62.7 ± 7 <0.001 248.1 ± 30.5 156.4 ± 21.1 0.007

Social 102.1 ± 10.6 53.0 ± 5.9 <0.001 238.6 ± 30.0 132.1 ± 18.2 0.002

Symptom scales

Fatigue 86.7 ± 9.1 49.1 ± 5.4 <0.001 197.3 ± 27.2 113.7 ± 17.3 0.003

Pain 105.2 ± 10.3 59.1 ± 5.9 <0.001 241.3 ± 27.9 141.3 ± 18.0 0.002

Nausea/vomiting 114.4 ± 12.1 64.8 ± 7.1 <0.001 261.8 ± 32.4 155.8 ± 21.1 0.002

Symptom items

Appetite loss 104.6 ± 12.5 60.7 ± 7.0 0.001 246.7 ± 35.8 147.3 ± 20.8 0.01

Constipation 113.3 ± 11.5 63.7 ± 7.4 <0.001 261.1 ± 32.6 154.0 ± 21.7 0.002

Diarrhea 96.3 ± 11.1 60.8 ± 6.6 0.003 221.2 ± 29.6 151.1 ± 20.0 0.02

Dyspnea 99.9 ± 11.0 63.2 ± 6.1 0.002 224.7 ± 30.6 152.8 ± 18.0 0.02

Financial difficulty 106.5 ± 10.8 54.3 ± 6.6 <0.001 242.7 ± 32.4 134.2 ± 18.6 0.003

Insomnia 89.5 ± 9.1 50.1 ± 5.7 <0.001 208.6 ± 26.9 121.0 ± 19.6 0.002

WHOQOL-BREF

Domain 1 (physical) 80.5 ± 8.9 46.4 ± 4.6 <0.001 185.7 ± 26.0 111.1 ± 13.7 0.007

Domain 2 (psychological) 77.9 ± 9.0 44.8 ± 4.0 <0.001 171.8 ± 28.3 103.2 ± 11.5 0.01

Domain 3 (social) 74.2 ± 7.8 47.5 ± 4.7 0.002 164.2 ± 24.3 119.1 ± 13.9 0.06

Domain 4 (environmental) 80.5 ± 9.2 46.5 ± 5.4 <0.001 176.4 ± 26.4 112.5 ± 15.4 0.02

*Only patients aged 50 years or less at diagnosis were included for analysis.

514

Page 13: Survival-weighted health profile for long-term survivors of ... · Survival-weighted health profile for long-term survivors of acute myelogenous leukemia Chiun Hsu1,2, Jung-Der

(qascstudy(t)/qascref(t)) is necessary to verify thevalidity of such an extrapolation [9]. Our data fitwell with such a validity check-up.

There are several limitations in this study. Thefirst is the limited patient number in the QOL in-terview that prevents stratification by potentiallyimportant disease- or treatment-related factors,such as leukemia classification and types oftransplantation. Besides, our previous simulationstudy suggested that a sample size of 100 or morein each group may be needed to reduce the bias oflong-term SWPS estimation to less than 5–7% [9].Therefore, more patients are needed to improvethe sensitivity and validity of the SWPS as an

endpoint for evaluation of clinical efficacy. Thesecond is the potential selection bias on assignmentof transplantation in the study cohort. In thisstudy patients who received BMT were signifi-cantly younger at diagnosis than those who re-ceived chemotherapy. In addition, some patientshad disease relapse while waiting for BMT. Thesepatients were included in the chemotherapy groupand thus resulted in bias favoring BMT in thesurvival comparison. Nevertheless, application ofthis new method on existing cohorts from ran-domized prospective trials will prevent these biasesand produce more credible inference of SWPScomparison.

Figure 4. Comparison of life-long extrapolation of quality-adjusted survival time (E(SWPS)) between patients who received trans-

plantation or chemotherapy. Hatched area denoted the difference of E(SWPS) between the two groups.

515

Page 14: Survival-weighted health profile for long-term survivors of ... · Survival-weighted health profile for long-term survivors of acute myelogenous leukemia Chiun Hsu1,2, Jung-Der

In conclusion, the SWPS may be useful toevaluate the efficacy of different treatment strate-gies for AML and to facilitate clinical decision-making. Patients’ perspective on QOL may bedomain-specific and may evolve over time. Con-firmation from prospective studies is needed.

Acknowledgements

This work was supported by a grant (NSC-89-2314-B-002-475-M56) from the National ScienceCounsel, Taiwan, ROC. The authors thank MsMing-Ling Lin for her secretarial help and MsShu-Yu Chen, Pey-Ing Chen and I-Ching Hou fortheir help in questionnaire interview.

References

1. Lee SJ, Earle CC, Weeks JC. Outcome research in oncol-

ogy: History, conceptual framework, and trends in the lite-

rature. J Natl Cancer Inst 2000; 92: 195–204.

2. Gotay CC, Muraoka MY. Quality of life in long-term

survivors of adult-onset cancers. J Natl Cancer Inst 1998;

90: 656–667.

3. Glasziou PP, Simes RJ, Gelber RD. Quality adjusted sur-

vival analysis. Stat Med 1990; 9: 1259–1276.

4. Cole BF, Gelber RD, Kirkwood JM, Goldhirsch A, Bar-

ylak E, Borden E. Quality-of-life adjusted survival analysis

of interferon alfa-2b adjuvant treatment of high-risk

resected cutaneous melanoma: An Eastern Cooperative

Oncology Group study. J Clin Oncol 1996; 14: 2666–

2673.

5. Goldhirsch A, Gelber RD, Simes RJ, Glasziou P, Coates

AS. Costs and benefits of adjuvant therapy in breast cancer:

A quality-adjusted survival analysis. J Clin Oncol 1989; 7:

36–44.

6. Parsons SK, Gelber S, Cole BF, et al. Quality-adjusted

survival after treatment for acute myeloid leukemia

in childhood: A Q-TWiST analysis of the pediatric on-

cology group study 8821. J Clin Oncol 1999; 17: 2144–

2152.

7. Mounier N, Haioun C, Cole BF, et al. Quality of life-ad-

justed survival analysis of high-dose therapy with autolo-

gous bone marrow transplantation versus sequential

chemotherapy for patients with aggressive lymphoma in

first complete remission. Blood 2000; 95: 3687–3692.

8. Hwang JS, Tsauo JY, Wang JD. Estimation of expected

quality adjusted survival by cross-sectional survey. Stat

Med 1996; 15: 93–102.

9. Hwang JS, Wang JD. Monte Carlo estimation of extrapo-

lation of quality-adjusted survival for follow-up studies.

Stat Med 1999; 18: 1627–1640.

10. Tsauo JY, Hwang JS, Chiu WT, Hung CC, Wang JD.

Estimation of expected utility gained from the helmet law

Figure 4. (Continued)

516

Page 15: Survival-weighted health profile for long-term survivors of ... · Survival-weighted health profile for long-term survivors of acute myelogenous leukemia Chiun Hsu1,2, Jung-Der

in Taiwan by quality-adjusted survival time. Accident Anal

Prev 1999; 31: 253–263.

11. Armitage JO. Bone marrow transplantation. N Engl J Med

1994; 330: 827–838.

12. Appelbaum FR, Fisher LD, Thomas ED, et al. Chemo-

therapy vs. marrow transplantation for adults with acute

nonlymphocytic leukemia: A five-year follow-up. Blood

1988; 72: 179–184.

13. Reiffers J, Gaspard MH, Maraninchi D, et al. Comparison

of allogeneic or autologous bone marrow transplantation

and chemotherapy in patients with acute myeloid leuka-

emia in first remission: A prospective controlled trial. Br J

Haematol 1989; 72: 57–63.

14. Zittoun RA, Mandelli F, Willemze R, et al. Autologous

or allogeneic bone marrow transplantation compared with

intensive chemotherapy in acute myelogenous leukemia.

N Engl J Med 1995; 332: 217–223.

15. Harousseau JL, Cahn JY, Pignon B, et al. Comparison of

autologous bone marrow transplantation and intensive

chemotherapy as postremission therapy in adult acute

myeloid leukemia. Blood 1997; 90: 2978–2986.

16. Cassileth PA, Harrington DP, Appelbaum FR, et al. Che-

motherapy compared with autologous or allogeneic bone

marrow transplantation in the management of acute mye-

loid leukemia in first remission. N Engl J Med 1998; 339:

1649–1656.

17. Syrjala KL, Chapko MK, Vitaliano PP, Cummings C,

Sullivan KM. Recovery after allogeneic marrow trans-

plantation: Prospective study of predictors of long-term

physical and psychosocial functioning. Bone Marrow

Transplant 1993; 11: 319–327.

18. McQuellon RP, Russel GB, Rambo TD, et al. Quality of

life and psychological distress of bone marrow transplant

recipients: The ‘time trajectory’ to recovery over the first

year. Bone Marrow Transplant 1998; 21: 477–486.

19. Neitzert CS, Ritvo P, Dancey J, Weiser K, Murray C,

Avery J. The psychosocial impact of bone marrow trans-

plantation: A review of the literature. Bone Marrow

Transplant 1998; 22: 409–422.

20. Burnett AK. Transplantation in first remission of acute

myeloid leukemia. N Eng J Med 1998; 339: 1698–1700.

21. Zittoun R, Suciu S, Watson M, et al. Quality of life in

patients with acute myelogenous leukemia in prolonged

first complete remission after bone marrow transplantation

(allogeneic or autologous) or chermotherapy: A cross

sectional study of the EORTC-GIMEMA AML 8A trial.

Bone Marrow Transplant 1997; 20: 307–315.

22. Aaronson NK, Ahmedzai S, Bergman B, et al. The Euro-

pean Organization for Research and Treatment of Cancer

QLQ-C30: A quality-of-life instrument for use in interna-

tional clinical trials in oncology. J Natl Cancer Inst 1993;

85: 365–376.

23. Ringdal GI, Ringdal K. Testing the EORTC Quality of

Life Questionnaire on cancer patients with heterogeneous

diagnoses. Qual Life Res 1993; 2: 129–140.

24. Ringdal K, Ringdal KI, Kaasa S, et al. Assessing the con-

sistency of psychometric properties of the HRQoL scales

within the EORTC QLQ-C30 across populations by means

of the Mokken scaling method. Qual Life Res 1999; 8: 25–

43.

25. The WHOQOL group. The World Health Organization

quality of life assessment (WHOQOL): Development and

general psychometric properties. Soc Sci Med 1998; 46:

1569–1585.

26. The WHOQOL group. Development of the World Health

Organization WHOQOL-BREF quality of life assessment.

Psychol Med 1998; 28: 551–558.

27. Bullinger M. Ensuring international equivalence of quality

of life measures: Problems and approaches to solutions. In:

Orley J, Kuyken W (eds), Quality of Life Assessment:

International Perspectives, Heidelberg: Springer-Verlag,

1994.

28. Lin MR, Yao KP, Hwang JS, Wang JD. Scale descriptor

selection for Taiwan-version of questionnaire of World

Health Organization quality of life. Chin J Pub Health

(Taipei) 1999; 18: 262–270 (Chinese).

29. Fayers PM, Aaronson NK, Bjordal K, Curran D, Groen-

vold M. EORTC QLQ-C30: Scoring manual. 2nd edn.

Brussel, Belgium: European Organization for Research

and Treatment of Cancer Quality of Life Study Group,

1999.

30. World Health Organization. WHOQOL-BREF: Introduc-

tion, Administration, Scoring and Generic Version of the

Assessment-field Trial Version. Geneva, 1996.

31. Efron B, Tibshirani RJ. An Introduction to the Bootstrap.

Chapman and Hall, New York, 1993.

32. Mangione CM, Goldman L, Orav EJ, et al. Health related

quality of life after elective surgery. J Gen Intern Med 1997;

12: 686–697.

33. Velikova G, Wright P, Smith AB, et al. Self-reported

quality of life of individual cancer patients: Concordance of

results with disease course and medical records. J Clin

Oncol 2001; 19: 2064–2073.

34. Mark DB, Hlatky MA, Califf RM, et al. Cost effectiveness

of thrombolytic therapy with tissue plasminogen activator

as compared with streptokinase for acute myocardial in-

farction. N Eng J Med 1995; 332: 1418–1424.

Address for correspondence: Jih-Luh Tang, Department of In-

ternal Medicine and Department of Oncology, National Tai-

wan University Hospital, No. 7, Chung-Shan South Road,

Taipei 100, Taiwan

Phone: þ886-2-23123456 ext. 5420; Fax: þ886-2-23816716

E-mail: [email protected]

517