Original article. G. Apolone, 1 A. Filiberti, 2 S. Cifani, 1 R. Ruggiata 1 & P. Mosconi 1

Transcription

1 Annals of Oncology 9: , Kluwer Academic Publishers. Printed in the Netherlands. Original article Evaluation of the EORTC QLQ-C30 questionnaire: A comparison with SF-36 Health Survey in a cohort of Italian long-survival cancer patients G. Apolone, 1 A. Filiberti, S. Cifani, 1 R. Ruggiata 1 & P. Mosconi 1 1{ Laboratorioper la Ricerca Clinica Oncologica, fstituto di Ricerche Farmacologiche Mario Negri'; Istiluto Europeo di Oncologia, Milano, Italy Summary Background: Despite the large amount of data available about the EORTC QLQ-C30 questionnaire, there have been very few studies focussed on long-survival cancer patients, and no data are available on its performance in the Italian setting. Patients and methods: Within the framework of a project aimed at evaluating the characteristics of available HR-QOL questionnaires in the Italian language, the EORTC QLQ-C30 questionnaire together with the Short Form 36-item Health Survey (SF-36) were mail-administered to a sample of patients previously recruited in two large multicenter randomized clinical trials on early breast and colon cancers. The properties of the questionnaire were evaluated using standard psychometric techniques and correlation analyses with demographic and clinical independent variables. Results: In the sample of patients who sent back the questionnaires under evaluation, the EORTC QLQ-C30 showed satisfactory acceptability (response rate = 64% and very low Introduction Disease-free and overall survival, tumor response, and toxicity have been the traditional end-points for evaluating treatment effects in clinical trials of cancer therapy. The past decade witnessed a growing interest in including and assessing the impact of the disease and its treatment on the patients' quality of life (QOL), and by now several clinical trial organizations have introduced QOL assessment as a standard component of new trials [1, ]. The term quality of life and, more specifically, Health Related Quality of Life (HR-QOL), refers to a multidimensional concept which encompasses perceptions of both negative and positive aspects of at least the four dimensions of physical, emotional, social and cognitive function [-5]. In addition to the clinical research setting, applications of HR-QOL data are suggested in several areas such as alerting of physicians and nurses to patients' common concerns, informing patients of common reactions to their cancer, aiding patients in medical decision-making, developing a training program for health professionals, and designing psychosocial interventions in oncology [6]. HR-QOL is also considered relevant in the economic appraisal of cancer prevalence of missing at item and scale level), and the psychometric analyses confirmed the multi-dimensional conceptualisation in terms of convergent and discriminant validity. Moreover, EORTC QLQ-C30 scales showed substantial correlation with the homologous SF-36 scales. Few socio-demographic (age, gender, schooling) and clinical (type of cancer disease) variables were associated with HR-QOL. Breast cancer patients reported, on average, worse physical health-related scores, but after adjustment for age and education, most of the differences disappeared. Conclusions: Thesefindingsconfirm the validity and robustness of the EORTC QLQ-C30 in this sample of long-survival Italian cancer patients. Further ad hoc validation studies are required to evaluate its significance in these particular patients. Key words: EORTC QLQ-C30, health outcome assessment, health-related quality of life, long-survival cancer patients, questionnaires, SF-36 therapy [7], and may be of prognostic importance for survival [8, 9]. With respect to cancer patients, there are many instruments for evaluating HR-QOL, ranging from generic ones designed for use across a wide range of chronic diseases such as the SF-36 Health Survey [10], to specific measures such as FLIC [11], CARES [1], FACT [13], RSCL[14]. In 1986, the EORTC Study Group on Quality of Life started a research program addressed to the development of instruments for the assessment of HR-QOL in international oncology clinical trials. This working group developed the EORTC QLQ-C36, which has been further refined, and designated EORTC QLQ-C30 [15]. The EORTC QLQ-C30 is a 30-item multidimensional questionnaire, whose psychometric properties have been tested in several studies [16-19], designed for heterogeneous groups of cancer patients. In the original EORTC field study the QLQ-C30 was tested in 313 patients suffering from advanced lung cancer. Of these patients, only 13 Italian lung cancer patients completed the instrument after the therapy [15]. Although a recent review of 14 published EORTC QLQ-C30 studies with a total of,945 patients (range 6-537) showed its possi-

2 550 bilities across a wide array of cancers, stages and settings, no empirical data were reported of its applications in cancer survivors, and none of the examples are actually from Italy [0]. Because of this dearth of data, within the framework of an ongoing project sponsored by the Italian National Research Council (CNR) and the Italian Association for Research on Cancer (AIRC) aimed at assessing the performance of the available HR-QOL questionnaires in Italy, we carried out a cross-sectional survey to test the performance of the EORTC QLQ-C30 questionnaire in a large cohort of cancer survivors. Operationally, we administered the EORTC QLQ-C30 questionnaire together with two other standardized questionnaires, the Short Form 36-item Health Survey (SF-36) [10, 1-31] and the Gruppo Italiano per la Valutazione degli Interventi in Oncologia Questionnaire (GIVIO) [3, 33] to a sample of patients recruited in two large multicenter randomized clinical trials. The present analysis focuses on the EORTC QLQ- C30 questionnaire and evaluates its structure and external validity in terms of: - applicability and patient acceptance, by measuring the response rate and the questionnaire completeness; - psychometric performance, by evaluating how well the hypothesized grouping and scaling assumptions were satisfied in the present sample; - external convergent validity, by assessing the extent to which EORTC QLQ-C30 and SF-36 measures of the same concept correlate each other; - cross-sectional relationship with a selected list of clinical variables (known-group clinical validity). Patients and methods Patients and study design Three different self-administered questionnaires were mail-administered to a sample of patients previously recruited in two large multicenter randomized clinical trials on early breast and colon cancers in which 1,30 and 888 cases, respectively, were followed for more than five years after randomization. Objective, design and findings of the original studies have been reported in detail elsewhere [3-35]. Briefly, in the first trial, early-stage breast cancer patients were randomized after curative surgery to two different follow-up regimens to assess the effectiveness of intensive follow-up diagnostic testing. HR-QOL was a complementary end-point and it was longitudinally assessed using standardized ad hoc questionnaires. The second trial was an adjuvant study in which a regimen containing 5-fluorouracil (5-FU) and highdose folinic acid was compared to a no-treatment arm. In this trial a formal HR-QOL was also carried out. All patients randomized in the two original studies were initially considered (,08 cases). Those who had previously agreed to participate in the HR-QOL assessment had completed at least one questionnaire during the trial's follow-up and were still alive at the beginning of this project were included in this study (1.77 cases) and randomized to receive a booklet containing two of the three questionnaires under evaluation, together with a cover letter and a stamped envelope. The patients returned the booklets to the central research office (Istituto Mario Negri. Milano) where data were stored according to code numbers accessible only to researchers. If the patient decided to dis- Responders Non responders Refusals Dead Wrong address r ORIGINAL RCTs DATA BASES- GIVIO breast trial SITAC colon trial Potentially eligible population* 08 Patients actually contacted* and randomized to receive a booklet with questionnaires 177 Samples witb couples of questionnaires SF-36 and EORTC QLQ-C30 EORTC QLQ-C30 crvio and SF-36*" and CMO" (591) (588) (593) Alive and answering to at least one questionnaire during the trials follow-up Overall, an EORTC QLQ-C30 questionnaire was sent to 1181 patients, 604 of which answered Figure I. Study sampling strategy. continue participation, he/she was asked to send the blank questionnaire back and was not contacted further. If the patient did not return the questionnaire within 1 weeks, a new booklet was mailed. Figure 1 summarizes the study sampling strategy. Besides the patient self-reported information, additional data regarding socio-demographic and medical information were available from the data-base from the original trials. The HR-QOL questionnaires used in the study The EORTC QLQ-C30 questionnaire is a cancer-specific, self-administered, structured questionnaire designed for use in clinical trials [15-0] that contains 30 questions, 4 of which form nine multi-item scales representing various aspects, or dimensions, of HR-QOL: one global scale,fivefunctional scales (Physical, Role, Emotional, Cognitive and Social), and three symptom scales (Fatigue, Pain and Nausea). The remaining 6 items, not considered in this paper, are intended to be mono-item scales describing relevant cancer-oriented symptoms (dyspnea, insomnia, appetite, constipation, diarrhoea, financial difficulties). Questions are combined in scales according to pre-defined rules. The SF-36 Health Survey (SF-36), one of the most widely used HR-QOL instruments in the US [10, 1-31], is a generic questionnaire that measures two major health concepts (Physical and Mental Health) with 36 questions and eight multi-item scales: Physical Functioning, Role Physical, Bodily Pain, General Health, Vitality, Social Functioning. Role- Emotional, Mental Health and Reported Health Perception. An additional one-item measurement of self-evaluated change in health status is also available. SF-36 was adopted in the present study as reference for its comprehensiveness, brevity and high standard of reliability and validity, and because it has been available in Italian since In addition, the launching of the International Quality of Life Project (IQOLA) in 1991 made possible the use of data from several comparable applications across the world, increasing the availability of information clarifying the meaning of scores describing each of the health concepts [5-7]. In Italy data

3 551 from a representative sample of the Italian population are also available, with almost 10,000 Italian applications [8-31]. Both the EORTC QLQ-C30 and SF-36 scores were assembled using the Likert method for summated ratings; the raw scores were then linearly transformed to 0- scales, with 0 and assigned to the lowest and highest possible values, respectively. Higher transformed scores indicate better health, while higher scores on symptom scales mean more intensive symptoms. in score is divided by the standard deviation of the entire sample. It has been suggested that an ES of 0. represents a small effect, 0.5 a moderate and 0.8 a large effect [43]. Given the large number of statistical tests carried out and the relatively small sample size of the groups compared, the P-values reported should be interpreted cautiously as indicative of a trend. Data analysis and statistics were performed with the SAS statistical package (SAS, Release 6.08, Cary, NC, USA). Statistical analysis According to the study design and the objective of this analysis, three different samples were analysed: 1) All of the original eligible patients, to evaluate the differences between responders and non-responders; ) The patients who returned the booklet containing the EORTC QLQ-C30 questionnaire, to assess their acceptability, the questionnaire's scaling assumptions and the known-group clinical validity (two-thirds of the participants); 3) The cases for which both EORTC QLQ-C30 and SF-36 are available, to evaluate the EORTC QLQ-C30 convergent validity (one-third of participants). Clinical and demographic variables were described using descriptive statistics such as mean, standard deviation and proportion. Patient acceptability was assessed using the response rate and the proportion of missing data at item and scale level. Grouping and scaling assumptions were assessed using the standard psychometric analyses described in the papers presenting the original questionnaires [15-17, 0, 36]. Briefly, as the questionnaire is based on a multi-dimensional conceptualization of health, the multi-trait analysis approach [37] was adopted to test whether the conceptualization in domains fits the data and whether the characteristics of the scales replicate the evaluation performed for the original questionnaire: in addition to convergence and divergence, internal consistency, (Cronbach's Alpha) was also evaluated [38, 39], Convergent validity, which is the most commonly used type of construct validation, focuses on the extent of correlation among several measures of the same concept. Usually, product-moment-correlations and factor analyses are used. In this paper, because the concurrent and independent measures of self-reported HR-QOL were available (i.e., the SF-36 questionnaire), convergent validity at scale level was also evaluated for substantial correlation between scales of the two questionnaires pertaining to the same construct (for example, the physical scales). On the other hand, low correlation was expected between scales pertaining to different concepts (for example, the EORTC QLQ-C30 Physical Functioning scale with the SF-36 Mental Health scale). A selected list of clinical variables describing the patients' sociodemographic characteristics, such as age, gender, schooling, marital and working status, as well as medical variables describing the nature and stage of the cancer disease were derived using information collected through standardized clinical forms in the original trials [3-35]. After coding in appropriate categories of the clinical variables and assembling the scales as indicated by authors and computing the HR-QOL mean scores for the entire population and for relevant subgroups, associations between clinical and HR-QOL variables were estimated using parametric and non-parametric analyses of variance (ANOVA and Wilcoxon). Analysis of the covariance (multiple regression analysis) [40] and non-parametric stratified analysis (stratified Wilcoxon's test) [41] were adopted to statistically test the association between independent and dependent variables controlling for the presence of confounding variables. When appropriate (i.e., when assessing the magnitude of the difference of mean scores between groups to give an estimate of the capability of each scale to detect relevant differences) an estimate of the effect-size (ES) was adopted. Operationally, the magnitude of change was assessed using the method proposed by Kazis et al. [4], the mean change Results Six-hundred four of the 1,181 patients who were asked to complete a booklet containing the EORTC QLQ-C30 questionnaire actually sent it back. As 189 deaths occurred and it was impossible to contact 43 patients because of incorrect address, the actual response rate of the present survey was 64% (604 of 949). At the inception of this cohort (i.e., when eligible patients were contacted) the median time since randomization was 65 and months for colon and breast cancers, respectively. The population was mainly composed of females (83%), the mean age was 63 years, and 8% were older than 70 years. Most were married, living with their respective partners; retirees comprised 36.6% of all responders, of whom only 6% were still working. Sixty-seven percent were breast cancer patients, 90% of the population had a ECOG performance status = 0 (i.e., fully active, able to carry on all pre-disease activities without restriction), 47% of the breast cancer patients had stage < II, and 5.4% of the colon cancer population had Dukes' stage B. Table 1 shows the characteristics of the three groups analyzed in this paper. Besides the expected self-selection due to the usual socio-demographic factors, there were no relevant differences among the entire sample contacted (Table 1, Original Sample), the patients whose returned booklet contained the EORTC QLQ-C30 in combination with the other questionnaires (Table 1, EORTC QLQ-C30) and those returning a booklet with both the EORTC QLQ-C30 and SF-36 questionnaires (Table 1, EORTC QLQ-C30 and SF-36). The percentage of missing data at item level was generally low, ranging from 1.9% to 8.6%. The items associated with the highest percentage of missing data concern the Role and Emotional Functioning scales. The proportion of scales that were non-computable ranged from 1.6% to 4.6%, with the Nausea and Vomiting scale being the one with the highest proportion of non-computable scores. Tables and 3 refer to the psychometric characteristics of the EORTC QLQ-C30 questionnaire. Data confirmed that the multi-dimensional conceptualization in domains and scales satisfactorily fits the present data: all scales met the recommended psychometric standards. In almost all cases the within-scale correlation coefficients were quite homogeneous, with all but one within-scale coefficient being higher than (convergent validity); most of the higher item-scale correlations were found within scale rather than between scales (dis-

4 55 Table I. Socio-anagraphical and clinical characteristics of samples. Characteristics Original sample (No. 177) EORTC QLQ-C30 (No. 604) EORTC QLQ-C30 and SF-36 (No. 313) No. % No. % No. % Age Mean (SD) Range < 60 years years > 70 years Gender Male Female School years Mean (SD) Range < =13 Marital status Single Married Divorced or widow Living alone Yes No, with husband/wife No, with family/relatives/other Profession Retired Housewife, unemployed, student Employee Other Cancer Colorectal Breast Performance status (ECOG) 0 (fully active) > (3.5) ' (9.5) b (9.3) (3.4) C (9.6) (3.5) 0-0 a Colorectal: stage A (No. 5, 0.9%); stage B (No. 305, 5.4%); stage C (No. 69, 46.%); Breast: stage < II (No. 558, 46.9%); stage» II (No. 63, 53.1%). b Colorectal: stage A (No. 3, 1.5%); stage B (No. 108, 54.3%); stage C (No. 87, 43.7%); Breast: stage < II (No. 17, 53.5%); stage 5= II (No. 188, 46.4%). c Colorectal: stage A (No.,.1%); stage B (No. 50, 5.1%); stage C (No. 43, 44.8%); Breast: stage < II (No. 119, 54.8%); stage > II (No. 98, 47.%). criminant validity). The internal-consistency reliability indexes (Cronbach's Alpha) were greater than, the standard recommended for group comparison for seven of the nine hypothesized scales, with the Physical Functioning and Nausea and Vomiting scales the only ones for which such a standard was not reached. In Table 4 the correlation matrix of the EORTC QLQ-C30 scales is shown: in general, as expected, there was substantial correlation between scales pertaining to the same health domain, and low correlation was observed between scales pertaining to different domains. In addition, all but one of the scales (Nausea and Vomiting) showed substantial-large correlation with the Global Health Status scale that can be considered a direct measure of the respondents' personal evaluation of his/her health. Table 5 displays the correlation matrix between the EORTC QLQ-C30 and SF-36 scales. The matrix was organized by ordering each set of variables according to the hypothesized domain to which they are attributed: from left (physical) to right (psycho-social), with the scales measuring general HR-QOL issues (general health) in the middle. The global picture confirms a substantial convergent validity. Correlations between each pair of homologous scales (those on the diagonal) were in the expected directions and substantial (/ > ). In addition, the EORTC scales pertaining to the physical health (PF, Pain, Fatigue) had higher correlations with the SF-36 physical scales (PF, RP, BP) than with the mental scales (SF, RE, MH), and the opposite expected phenomenon (higher correlations with mental than with physical scales) was also observed. Table 6 shows the impact of a few selected sociodemographic and clinical variables and the EORTC scale scores in terms of group mean difference. In general, differences were in the expected direction. Gender, education and type of cancer were the variables that most affected the health profiles: males, the well-edu-

5 553 Table. EORTC QLQ-C30: Item-scale correlation matrix. Items Mean SD Physical (PF) Role (RF) Emotional (EF) Cognitive (CF) Social (SF) Global health status (QL) Fatigue (FA) Nausea and vomiting (NV) Pain (PA) For the analysis 550 cases were considered because 54 cases were omitted for missing data. Grey area indicates hypothesized item groupings and item-scale correlation corrected for overlap. Table 3. Psychometric properties of the EORTC QLQ-C30 questionnaire. Scales Physical (PF) Role (RF) Emotional (EF) Cognitive (CF) Social (SF) Global health status (QL) Fatigue (FA) Nausea and vomiting (NV) Pain (PA) No. of items Item internal consistency Item discriminant validity ~ Item convergent validity test c Item discriminant validity test d Reliability 5 For the analysis 550 cases were considered because 54 cases were omitted for missing data. a Range of correlations between items and hypothesized scale corrected for overlap. b Range of correlations between items and other scales. c Item convergent validity scaling succes (%) i.e. number of item-scale correlations greater than /total number of correlations (corrected for overlap). d Item discriminant validity scaling succes (%) i.e. number of correlations of items with own scales significantly higher than correlations with other scales/total number of correlations. c Internal-consistency reliability (Cronbach's Alpha) cated, and colon cancer patients reported, on average, better health. The most sensitive scales were those pertaining to physical health (Physical and Role Functioning), although the size of the effect detected was most often low-moderate. Table 7 analyses in depth the impact of the cancer type on HR-QOL scales. After adjusting for age and education, the differences between colon and breast cancers were reduced, approaching statistical significance only in the PFand PA scales.

6 554 Table 4. EORTC QLQ-C30 versus EORTC QLQ-C30: matrix of correlation. EORTC QLQ-C30 EORTC QLQ-C30 Physical Role Emotional Cognitive Social Global health status Fatigue Nausea and vomiting Pain Physical Role Emotional Cognitive Social Global health status Fatigue Nausea and vomiting Pain Discussion The past decade of scientific and medical publishing has witnessed an increasing interest in HR-QOL, yielding a growing number of available generic and specific instruments to measure the concept of HR-QOL. The amount of available data on HR-QOL in different disease conditions, cancer stages and populations has increased to such an extent that many authors strongly underscore a concern that instruments and methods are being inappropriately used. In addition, the pros and cons of generic versus specific instruments need to be well focused. This debate has yielded the approach of assembling multi-packet tools that, according to the aims of the study, are variously composed of generic, diseasespecific or symptom-oriented questionnaires. In the oncology setting, of the few self-completed questionnaires that can be considered valid, reliable, and sufficiently brief to be of practical use in the clinical research setting, the EORTC QLQ-C30 has been frequently applied in Europe in different patient populations. According to the published data, the attention of most researchers has been confined to the onset of the disease or the final phase of its natural history, with very few studies focused on long-survivor cancer patients, a group whose numbers are going to increase because of the concurrent effect of various factors such as higher incidence, earlier diagnosis and better prognosis. In addition to these considerations, the fact that after the first test in 13 lung cancer patients [15] no additional data from the Italian sample were published suggested the need to test its performance in a large group of Italian cancer patients. Thus, in order to assess its performance in terms of construct and external validity we carried out a cross-sectional survey in a well-characterized sample of cancer patients assumed to be free, or to have a very low prevalence, of events related to the primary disease. Our findings showed that the validity of the EORTC QLQ-C30, when used in long-survival cancer patients, is quite satisfactory. The response rate (64%) was satisfactory, if we consider the study design used and the fact that patients included in the present survey were contacted several years after the completion of the two original studies. Although the present level of participation is somewhat lower than those reported in some surveys carried out in northern Europe [0], our findings can be considered comparable to those achieved in Italy in other similar 591

7 555 Table 5. EORTC QLQ-C30 versus SF-36: correlation matrix. SF-36 EORTC QLQ-C30 Physical Role Pain Global health status Fatigue Social Emotional Cognitive Physical Role - physical Bodily pain General health Vitality Social Role - emotional Mental health Note: Nausea and Vomiting scale (NV) was excluded because it does not match any SF-36 scale. Table 6. EORTC QLQ-C30: mean group differences according to a selected list of patients variables Patients variables PF RF EF CF SF QL FA a NV a PA a Age: < 60 years vs. > 60 years Gender: male vs. female School years: > 5 years vs. < 5 years Marital status: married vs. unmarried Living alone: yes vs. no Working status: employed vs. unemployed Cancer: colorectal vs. breast Overall mean scores (SD) a bc +7.39" +8.6 b ' c bc +9.3 l bc (3.3) b ' c C C 83.1 (.) b ' c (1.4) (1.) Higher scores correspond to worse health (high level of symptomatology/problems). b ES = moderate-large effect (> 0.5). c P < (17.3) C C 70. (1.3) " (0.9) bc (9.7) c C 18.5 (.8) experiences in cancer patients [44]. Moreover, the low incidence of missing data at item and scale level indicates a good acceptability of the questionnaire. Psychometric findings also confirm its multi-dimensional conceptualization, as the scales met all of the recommended standards in terms of convergent and discriminant validity. The internal consistency was also very high, as seven of the nine scales conformed to the minimal psychometric standards for reliability (Cronbach's Alpha coefficient greater than ), with the exception of the Physical Functioning and Nausea and Vomiting scales (0.64 and 0.58, respectively). The relatively poor performance of these two scales can be justified in light of the nature of some of the items which compose the scale, exploring the limitations encountered by patients in basic daily activities, together with the type of sample, namely, a relatively healthy population with no active cancer-related treatments involved. It is also worthwhile to consider, however, that values of Cronbach's Alpha ranging between and are generally accepted for group comparison [46] and that according to some authors [47] these standards can even be considered too stringent. Moreover, a similar Cronbach's Alpha value (0.68)

8 556 Table 7. Adjusted scales means scores according to cancer types. EORTC QLQ-C30 Physical Role Emotional Cognitive Social Global health status Fatigue Nausea and vomiting Pain Adjusted mean Breast cancer Colorectal cancer Scores were adjusted for age and schooling. P-values Analysis covariance Wilcoxon stratified for PF scale has already been documented in the original paper presenting the EORTC questionnaire [15]. Last but not least, the same failure (i.e., a poor performance of the homologous physical scale) was also documented in the SF-36 questionnaire. This suggests that the failure we documented in the convergent/discriminant tests and the low value of the internal consistency reliability should be attributed to the long-term survivorship status and not to the questionnaire conceptualization or translation. The documented substantial convergent validity with SF-36 scales supports the hypothesis that the EORTC QLQ-C30 questionnaire, although designed for cancer patients who had recently undergone active treatment, is also able to measure the patient perception of health status several years after their exposure to active interventions. These findings not only support the validity of the cancer-specific questionnaire when compared to the generic one, but also suggest the comparability of these two tools when used in long-survival cancer patients. This finding is in accord with a recent study [45] that showed a very satisfactory comparability between the two questionnaires, specifically for the Physical Functioning and the symptom scales in a hospitalized cancer population suffering from haematological malignancies and various other types of solid tumors. In this sample, demographic and clinical variables affected the EORTC QLQ-C30 scores. Gender (male versus female), education (>5 years versus <5 years), and age (< 65 years versus > 65 years) were significantly associated with better HR-QOL, as others have previously documented for other generic and specific questionnaires [1, 3, 9]. It means that the questionnaire under evaluation was able to discriminate between groups known to differ in a given health concept because of a particular external factor. As a matter of fact, most of the differences we documented at the univariate analyses were small in size and although the scores of colon cancer patients were generally different from those of breast cancer patients, the size and statistical significance of such differences were reduced after adjustment for age and education imbalance. Comparison of our data with similar data derived from the literature shows that values from our sample are fully compatible with those reported in the King review [0], but lower than those reported in the Osoba paper [16]. Because we don't have information about the characteristics of the Osoba sample, we cannot explain this difference. However, knowing the factors that usually have an impact on such scales when the subjects are cancer- and active-treatment-free, and on the basis of observations in our sample, we believe the difference to be attributable to the difference in the study designs and to the impact of non-cancer or -treatment factors rather than to the validity and cross-cultural stability of the scale under discussion. In conclusion, our study shows that the EORTC QLQ-C30 questionnaire maintains its psychometric properties in the present Italian sample of patients who may be considered free of disease- and treatment-related events. This evidence should be considered important because it is the first independent application of the questionnaire in Italy. In addition, for the first time, to our knowledge, it was administered to long-survivor cancer patients together with the SF-36, one of the most frequently used generic questionnaires. Although further validation studies in which generic, cancer-specific and stage-tailored instruments are administered to different types of long-term cancer survivors are mandatory before conclusions can be drawn about its applicability in this particular population, these data contribute evidence about the psychometric validity and robustness of the questionnaire when used in cancer patients who are not on treatment or do not have active disease. Acknowledgements This research was supported by CNR-ACRO (grant no PF39) and by AIRC (Associazione Italiana per la Ricerca sul Cancro). References 1. Editorial. Quality of life in clinical trials. Lancet 1995; 346: 1-.. Moinpour C. Measuring quality of life: An emerging science. J Clin Oncol 1994; 5: Osoba D. Lessons learned from measuring health-related quality of life in Oncology. J Clin Oncol 1994; 1: Cella D. Quality of life: Concepts and definition. J Pain Symptom Manage 1994: 9: Testa M, Simonson D. Assessment of quality of life outcomes. N Engl J Med 1996: 334: Meyerowittz BE. Quality of life in breast cancer patients: The contribution of data to the care of patients. Eur J Cancer 1993; 9A: Uyl de Groot CA, Rutten F, Bonsel G. Measurement quality of life in economic appraisal of cancer treatment. Eur J Cancer 1994: 30A: Ganz PA. Lee JJ, Siau J. Quality of life assessment. An independent prognostic variable for survival in lung cancer. Cancer 1991:67: Coates A. Gebski V, Signorini D et al. Prognostic value of quality

9 557 of life scores during chemotherapy for advanced breast cancer. J Clin Oncol 199: 10: Ware J. Sherbourne CD. The MOS 36-item short form for health survey (SF-36). 1. Conceptual framework and item selection. Med Care 199; 30: Shipper H, Clinch J, McMurray J, Levitt M. Measuring the quality of life of cancer patients. The Functional Living Index-Cancer. Development and validation. J Clin Oncol 1984; : Ganz PA, Schag C, Lee JJ, Sim M. The CARES: A generic measure of health-related quality of life for patients with cancer. Qual Life Res 199; 1: Cella D, Tulsky DS, Gray G et al. The Functional Assessment of Cancer Therapy (FACT) Scale: Development and validation of the general measure. J Clin Oncol 1993; 11: De Haes JC, Raatgever JW, Van der Burg MEet al. Evaluation of the quality of life of patients with advanced ovarian cancer treated with combination chemotherapy. Monograph Ser EORTC 1987; 17: Aaronson N, Ahmedzai S, Bergman B et al. The European Organisation for Research and treatment of cancer QLQ-30: A quality of life instrument for use in international clinical trials in oncology. JNCI 1993; 85: Osoba D, Zee B, Pater J et al. Psychometric properties and responsiveness of the EORTC quality of life questionnaire (QLQ- 30) in patients with breast, ovarian and lung cancer. Qual Life Res 1994; 3: Kaasa S, Bjordal K, Aaronson N et al. The EORTC Core Quality of Life of Questionnaire (QLQ-30): Validity and reliability when analysed with patients treated with palliative radiotherapy. Eur J Cancer 1995; 31A: Wisloff F, Eika S, Hippe E et al. Measurement of health-related quality of life in multiple myeloma. Br J Hematol 1996; 9: Hjermstad MJ, Fossa SD, Bjordal K, Kaasa S. Test-retest study of the European Organisation for research and treatment of cancer core quality of life questionnaire. J Clin Oncol 1995; 13: King MT. The interpretation of scores from the EORTC quality of life questionnaire QLQ-C30. Qual Life Res 1996; 5: McHorneyCA,Ware JEJr, Raczek AE. The MOS 36-item shortform health survey (SF-36): II. Psychometric and clinical tests of vitality in measuring physical and mental health constructs. Med Care 1994; 31: McHorney CA, Ware JE Jr, Lu JFR, Sherbourne CD. The MOS 36-item short-form health survey (SF-36): III. Tests of data quality, scaling assumptions and reliability across diverse patients groups. Med Care 1994; 3: Ware JE Jr. SF-36 Health Survey. Manual and Interpretation Guide. Boston, MA: The Health Institute, New England Medical Center Ware JEJr. The SF-36 Health Survey. In Spilker B(ed): Quality of Life and Pharmacoeconomics in Clinical Trials. Philadelphia: Lippincott-Raven 1996; Aaronson NK, Acquadro C, Alonso J et al. International quality of life assessment (IQOLA) project. Qual Life Res 199; I: Ware JE Jr, Keller SD, Gandek B et al. Evaluating translations of health status questionnaires. Methods from the IQOLA Project. Intern J Tech Assessment 1995; 11: Ware JE Jr, Gandek BL, Keller SD and the IQOLA Project. Evaluating instruments used cross-nationally: Methods from the IQOLA Project. In Spilker B (ed): Quality of Life and Pharmacoeconomics in Clinical Trials. Philadelphia: Lippincott-Raven 1996; Apolone G, Cifani S, Liberati MC, Mosconi P. Questionario sullo stato di salute SF-36. Traduzione e validazione della versione italiana: Risultati del progetto IQOLA. MEDIC 1997; 5: Apolone G, Mosconi P, Ware JE Jr. Questionario sullo stato di salute SF-36. Manuale d'uso eguida all"interpretazione dei risultati. Milano: Guerini Editore & Associati Apolone G, Mosconi P. The Italian SF-36 Health Survey: Translation, validation and norming from a clinical epidemiology perspective. Submitted. 31. Bosisio M, Parma E, Apolone G, Bertolini G, on behalf of OMG Group. La valutazione dello stato di salute in pazienti anziani con bronchite cronica. Uno studio di esito nella Medicina Generale. Ricerca Pratica 1996; 1: The GIVIO Investigators. Impact of follow-up testing on survival and health related quality of life in breast cancer patients. A multicenter randomized controlled trial. JAMA 1994; 71: Hart S, Meyerowitz E, Apolone G et al. Quality of life among mastectomy patients using external breast prostheses. Tumori 1997; 83: IMPACT investigators. Efficacy of adjuvant fluorouracil and folicin acid in colon cancer. Lancet 1997; 345: Zaniboni A, Labianca R, Marsoni S et al. for the SITAC-01 Investigators. GIVIO-SITAC 01. Randomized trial of adjuvant 5-fluorouracil and folinic acid in patients with colon cancer: Long-term results and evaluation of indicators of health related quality of life. Cancer (in press). 36. Fayers P, Aaronson N, Bjordal K, Sullivan M. EORTC QLQ-C30 scoring manual. EORTC Quality of Life Study Goup Hays RD, Hayashi T, Carson S, Ware JE Jr. User's Guide for the Multitrait Analysis Program (MAP). Rand Corporation Cronbach LJ. Coefficient alpha and the internal structure of tests. Psychometrika 1951; 16: Bland JM, Altman DG. Cronbach's alpha. BMJ 1997; 314: Kleinbaum DG, Kupper LL, Muller KE. Applied Regression Analysis and Other Multivariable Methods. Second edition. Boston: PWS-Kent Publishing Company. 41. Koch GG, Edwards S. Clinical efficacy trials with categorical data. In Peace KG (ed): Biopharmaceutical Statistics for Drug Development. Basel: Marcel Dekker. 4. Kazis LE, Anderson JJ, Meenan RF. Effect sizes for interpreting changes in health status. Med Care 1989; 7: Cohen J. Statistical Power Analysis for the Behavioral Sciences. Hillsdale, NJ: Lawrence Erlbaum Associates Mosconi P, Torri V Cifani S et al. The multi-centre assessment of quality of life: The interdisciplinary group for cancer care evaluation (GIVIO) experience in Italy. Stat Med 1988; 17 (in press). 45. Porzsolt F, Wolp CP, Rist CE et al. Comparison of three instruments (QLQ-C30, SF-36, QWB-7) measuring health-related quality of life/quality of well-being. Psycho-Oncology 1996; 5: Nunnally JC. Psychometric Theory, second edition, New York, NY: McGraw Hill Hays RD, Anderson R, Revicki D. Psychometric considerations in evaluating health-related quality of life measures. Qual Life Res 1993: : Received 31 July 1997; accepted 10 February Correspondence to: Giovanni Apolone, MD Laboratorio per la Ricerca Clinica Oncologica Istituto di Ricerche Farmacologiche 'Mario Negri' Via Eritrea, Milano Italy apoloncairfmn.mnegri.it

10