European PsychometricValidation of the CONTILIFE 1 : A Quality of Life Questionnaire for Urinary Incontinence

Transcription

1 European Urology European Urology 43 (2003) European PsychometricValidation of the CONTILIFE 1 : A Quality of Life Questionnaire for Urinary Incontinence Gérard Amarenco a, Benoit Arnould b,*, Paulo Carita c,1, François Haab d, Jean-Jacques Labat e, François Richard f a Hôpital Rotschild, Service de Rééducation Neurologique et d Explorations Périnéales, 33, boulevard de Picpus, Paris Cedex 12, France b MAPI Values, 27, rue de la Villette, Lyon, France c Sanofi-Synthélabo, 31, avenue Paul Vaillant Couturier, Bagneux, France d Hôpital Tenon, Service Urologie, 4, rue de la Chine, Paris Cedex 20, France e Centre Médical Mauvoisins, Service de Rééducation Fonctionnelle, 25, rue Mauvoisin, Nantes, France f Groupe Hospitalier Pitiè-Salpétrière, Service Urologie, 47-83, boulevard de l Hôpital, Paris Cedex 13, France Accepted 3 January 2003 Abstract Objective: To validate 5 language versions (French, Dutch, German, English and Danish) of CONTILIFE 1,a urinary incontinence-specific measure of Quality of Life (QoL). Methods: CONTILIFE 1 was administered to Belgian, Danish, English, French, German, and Dutch women with genuine stress incontinence. The internal consistency, the construct and clinical validity, and the responsiveness to change over time were assessed using standardised procedures. Results: 505 women filled in the CONTILIFE 1 at baseline. Quality of completion was good. The mean number of missing data varied between 0.33 (1.17%) and 0.94 (3.36%) according to the country. Internal consistency was good for all dimensions except for the Effort Activities dimension. Construct validity was good for the Danish, French and pooled sample, and acceptable for English, German and Dutch samples. The distribution of QoL scores differed according to the number of urinary leaks (p < 0:007 ). 477 women completed the questionnaire twice with a 28-day interval. Responsiveness to clinical improvement was good (jesj 0:50) except for the Sexuality and Well-Being dimensions. Conclusion: In this international European study, CONTILIFE 1 matched the psychometric requirements to be used as a valid, reliable, and sensitive measure of the QoL of the patients. The Daily Activities, Effort Activities, Self- Image and Emotional Consequences dimensions are the most relevant and valid dimensions for purposes of comparison. # 2003 Elsevier Science B.V. All rights reserved. Keywords: Quality of life (QoL); Urinary incontinence (UI); Psychometrics; France; Belgium; Germany; The Netherlands; Denmark; United Kingdom 1. Introduction Urinary incontinence (UI) is a significant health problem in modern society. All over the world, UI affects more than 50 million people, and women in * Corresponding author. Tel. þ ; Fax: þ addresses: barnould@mapi.fr (B. Arnould), paulo.carita@sanofi-synthelabo.com (P. Carita). 1 Tel. þ ; Fax: þ particular. Population studies have demonstrated that approximately 10% of women suffer from UI. The prevalence of UI increases with age [1]. Treatment for UI depends on the type of incontinence. Nevertheless, efficacy is usually measured as a decrease in the number of incontinence episodes or in weight of pads. These traditional objective measures are limited in that they fail to take the patients perception into account. Since the 1990s, the importance of using Quality of Life (QoL) tools to evaluate /03/$ see front matter # 2003 Elsevier Science B.V. All rights reserved. doi: /s (03)00054-x

2 392 G. Amarenco et al. / European Urology 43 (2003) patient perception of disease (subjective measures) has been underlined [2]. Generic and UI-specific measures have been used to assess the QoL effects of UI [3]. The CONTILIFE 1 was designed for women suffering from UI of any type (urge, stress and mixed UI). Developed and tested in France [4], it contains 28 items. Seven dimension scores and one global score are generated. It has since then been translated into 4 other languages (British English, Danish, Dutch, and German) [5]. As the aim of the translation procedure was to produce a tool of conceptual rather than mere linguistic equivalence [6], a standardised translation procedure [7] was used to obtain the other language versions of the CONTI- LIFE 1. This article describes the psychometric properties of the 5 language versions of CONTILIFE 1 : British English, Danish, Dutch, French and German. 2. Patients and methods 2.1. Study plan The psychometric properties of the 5 language versions of CONTILIFE 1 were evaluated during a European multicentre clinical study assessing the efficacy and safety of 3 doses of a new drug in women suffering from genuine stress incontinence. This study was conducted in accordance with the principles set out in the latest version of the Declaration of Helsinki. It was a randomised, double-blind, placebo-controlled study conducted in 4 parallel groups. At selection visit (D 14), patients received a self-evaluation diary. At inclusion (D0) and end-of study (D28) visit, patients completed the CONTILIFE 1 questionnaire and data from the diary were collected. The diary was completed during the week prior to each visit and enabled the number of urinary leaks per week (day and night) to be calculated Patients Only women, aged between 18 and 75 years, in good health, suffering from mild to moderate (grades I II) [8,9] genuine stress incontinence with at least 3 urinary leaks per week and 24-hour pad test values between 8 g (incl.) and 100 g (excl.), were included in the study. In all, 505 women, from 6 countries and speaking 5 languages, were recruited in 80 centres: 51 were from Belgium, 76 from Denmark, 147 from France, 48 from Germany, 95 from The Netherlands and 88 from the UK. The Belgian patients were categorised as French or Dutch speaking. All the included women signed an informed consent form CONTILIFE 1 questionnaire The original CONTILIFE 1 instrument was developed in France. It contains 28 items, each having a 5- or 6-point Likerttype response scale [10]. CONTILIFE 1 scores were generated following the instructions of the authors giving a global score (27 items) and 1 score per dimension (Daily Activities, Effort Activities, Self-Image, Emotional Consequences, Sexuality and Well-Being). Scores ranged from 0 (high QoL) to 100 (low QoL). The English version of CONTILIFE 1 is appended. Translation and cultural adaptation of the French version into British English, Danish, Dutch, and German versions was carried out using a standardised translation procedure [4]. The methodology included (a) recruitment of a QoL specialist in each target country, (b) production of 2 independent forward translations by 2 professional translators, who were native speakers of the target language, (c) creation of a reconciled version, (d) backward translation by a professional translator who was a native speaker of the source language, (e) production of a second reconciled version, (f) cognitive debriefing to assess clarity, appropriateness of wording and acceptability of the translated questionnaire, and (g) international harmonisation Measurement properties and statistics Analyses were carried out with SAS (6.12), MAP (1.0), Bigstep (2.76), and SPAD (4.03) software. The significance level for all tests was set at 5%, except for the analysis of clinical validity, where the significance level was adjusted using Bonferroni s method [11]. A p-value lower than was considered to be significant. Baseline sociodemographic and clinical data were described per country and for all countries together. The comparison of QoL scores according to the country were performed using a Kruskal Wallis test. Acceptability of the CONTILIFE 1 was analysed by describing the number of missing data per patient and assessment, the number of missing data per item, and the number of patients with a non-assessable QoL questionnaire (50% of missing data). Psychometric properties were tested using both standardised procedures and instrument review criteria developed by the Scientific Advisory Committee of the Medical Outcomes Trust [12]. The questionnaire structure was assessed on baseline data using Principal Component Analysis (PCA) with Varimax rotation and Multitrait Analysis (MA) per language group and for all groups together. Based on item-scale correlations, this analysis was used to confirm the a priori scale structure of the questionnaire. Itemconvergent validity was satisfactorily obtained when items were significantly correlated with their own dimension (Pearson s correlation coefficient >0.40) and item-divergent validity was satisfactorily obtained when items were more strongly correlated with their own dimension than with the other dimensions [13]. Results are expressed as percentage scaling success. The internal consistency reliability was assessed using Cronbach s alpha coefficient [14] per country and for all countries together. This coefficient indicates the proportion of variance due to the true score as compared to that due to measurement error, for each scale. The minimum recommended value is 0.7. The floor and ceiling effects (percentage of patients with the lowest and the highest scores, respectively) were determined. A percentage greater than 20% was considered high. Clinical validity explores the logical relations that may exist between the questionnaire and clinical measures. It was assessed by Spearman s correlation coefficients [15]. The clinical parameter adopted was the number of urinary leaks. The hypothesis was that the poorer the health status (the greater the number of urinary leaks), the lower the patient s QoL (the higher the CONTILIFE 1 scores). The observed frequency in urinary leakages recorded ranged from 0 to patients reported less than 3 urinary leakages during the week preceding baseline. However, all patients were included in the analysis. Patients were split into 4 categories: [0; 7], ]7; 13], ]13; 21] and >21 urinary leaks per week. The effect of age (4 classes were defined according to quartiles and median) and menopausal status (yes/no) on the QoL scores was also assessed. Responsiveness to change over time (sensitivity) was assessed by means of effect sizes (ES) for scores [16]. The aim was to evaluate the ability of the questionnaire scores to change when changes were observed in clinical status. Effect sizes were

3 G. Amarenco et al. / European Urology 43 (2003) considered moderate when their absolute values were >0.50 and large when >0.8 [16]. This analysis was performed on pooled data of patients with an assessable QoL questionnaire on D0 and D28. Patients were considered to be improved if the relative difference in the number of urinary leaks between baseline and D28 was greater than 50%, minimally improved if between 50% and 25%, and stable if lower than 25%. 3. Results 3.1. Quality of completion of the CONTILIFE 1 At baseline, 505 patients filled in the questionnaire. The mean number of missing data per country varied between 0:33 1:45 in The Netherlands (1.17% missing items per questionnaire) and 0:94 1:96 in Belgium (3.36% missing per questionnaire). One French patient was excluded from further analysis due to having more than 50% missing data. On D28, 482 patients were still present, and filled in the CONTI- LIFE 1. The mean number of missing data varied between 0:24 0:71 (The Netherlands) and 0:73 1:56 (Denmark). In all, 477 patients filled in the QoL questionnaire on both D0 and D28, from which 18 were excluded from the longitudinal analysis (responsiveness) because the change in the number of urinary leakages was missing. Item 25 (concerning sexual intercourse) was the item with the highest number of missing data in all countries (however, at baseline, the number of missing data for item 25 ranged from 6% (in The Netherlands) to 20% (in Denmark), which can be considered acceptable for an item about sexuality) and at each assessment Baseline characteristics of patients The majority of patients were of Caucasian ethnic origin (98.6%).Their baseline characteristics are presented in Table 1. The QoL scores differed between countries (Table 2): all the scores except the Sexuality score were lower in The Netherlands group than in the other groups, indicating a better QoL Reliability and construct validity of CONTILIFE 1 To confirm the scaling hypothesis and scoring, a PCA with Varimax rotation was performed first on the whole population and then per language group. The number of factors was limited to 6, which was the number of dimensions in the CONTILIFE 1. In the whole population, the percentage of variance explained by these 6 factors represented 64.3% of the total variance. Most of the items were grouped according to their dimension. The Daily Activities and Effort Activities dimensions were grouped on the same factor, while the Well-Being item was correlated with the factor which included most of the Emotional Consequences items (data not shown). Results in the Danish and French language groups were good similar to those obtained with the pooled data. In the English, German and Dutch language groups, results were slightly poorer (data not shown). Internal consistency reliability was satisfactory in the whole population: all the Cronbach s alpha coefficients were over 0.70 ranging from 0.71 (Effort Activities) to 0.94 (Global Index). In the English and Danish language groups, results were similar: Cronbach s alpha coefficients varied between 0.77 (English, Effort Activities) and 0.94 (English, Global Index). In the French, Dutch and German groups, results were satisfactory except for the Effort Activities dimension (Cronbach s alpha coefficients: 0.69, 0.56 and 0.63 in each group, respectively). The percentages of patients with the lowest (floor effect) or highest (ceiling effect) Table 1 Sociodemographic and clinical baseline characteristics in the 504 women suffering from genuine stress incontinence Patients characteristics Belgium (n ¼ 51) Germany (n ¼ 48) Denmark (n ¼ 76) France (n ¼ 146) UK (n ¼ 88) The Netherlands (n ¼ 95) Total (n ¼ 504) Age (years): mean S.D Weight (kg): mean S.D Height (cm): mean S.D Number of urinary leaks: mean S.D Menopausal (%) Number of pregnancies (%) >

4 394 G. Amarenco et al. / European Urology 43 (2003) Table 2 Quality of life scores at baseline in each country CONTILIFE 1 dimensions Belgium (n ¼ 51) Germany (n ¼ 48) Denmark (n ¼ 76) France (n ¼ 146) UK (n ¼ 88) The Netherlands (n ¼ 95) Total (n ¼ 504) Comparison between countries (p) a Daily Activities b b b Effort Activities c < Self-Image < Emotional Consequences c c c < Sexuality d e e e f c Well-Being c c c b < Global score f e g f f c < a Kruskal Wallis test. b Missing data ¼ 2. c Missing data ¼ 1. d Missing data ¼ 3. e Missing data ¼ 4. f Missing data ¼ 6. g Missing data ¼ 7. possible scores were low, indicating the good coverage of the underlying concepts, except for the Sexuality and Well-Being dimensions. Concerning the Sexuality dimension, a floor effect was observed: percentages of patients with the lowest possible scores varied between 32.9% (English) and 65.1% (German). Concerning the Well-Being dimension, a floor effect was observed in the Dutch language group (22%). Fig. 1. Distribution of CONTILIFE 1 scores for 468 women suffering from genuine stress incontinence according to the number of urinary leaks per week. Patients were split into 4 classes: [0; 7], ]7; 13], ]13; 21] and >21 urinary leaks per week. DA, Daily Activities; EA, Effort Activities; SI, Self-Image; EM, Emotional Consequences; SEX, Sexuality; WLB, Well-Being; Global, Global Index. Lower QoL scores indicated better QoL.

5 G. Amarenco et al. / European Urology 43 (2003) Fig. 2. Distribution of CONTILIFE 1 scores for 504 women suffering from genuine stress incontinence according to their age. Patients were split into 4 classes: 1st quartile (43 years); ]1st quartile, median] (]43; 51] years); ]median, 3rd quartile] (]51; 59] years); and >3rd quartile (>59 years). DA, EA, SI, EM, SEX, WLB and Global: see Fig. 1. In the whole population, convergent validity, examined by correlation of each item with its own dimension, was good except for 1 item of the Daily Activities dimension. Scaling success rate varied between 85.7% and 100%. The percentage of items correlating more strongly with their own dimension than with the other dimensions was good. The scaling success rate varied between 91.4% and 100%. Results per language group were good regarding the Daily Activities and Sexuality dimensions, and moderate regarding the Effort Activities, Self-Image and Emotional Consequences dimensions (data not shown) Clinical validity Clinical validity was satisfactory, the more impaired patients (the greater the number of urinary leaks) having the poorer QoL (the higher the CONTILIFE 1 scores). All the correlation coefficients were associated with high significance levels (p < 0:007 ). The largest differences were observed for the Daily Activities and Effort Activities scores (Fig. 1). There were no significant differences in the distribution of QoL scores according to age (except for the Self-Image and Sexuality dimensions) (Fig. 2) or menopausal status (Fig. 3) Responsiveness over time From baseline to D28, 177 women were considered improved, 83 minimally improved and 199 stable or worsened. The change was missing in 18 patients. In improved women, mean changes in QoL score were highly significantly negative, indicating QoL improvement. Effect sizes (absolute values) were greater than 0.5 except for the Emotional Consequences, Sexuality and Well-Being dimensions. For minimally improved women, the mean changes in QoL score were significantly negative, except for the Well-Being dimension (Fig. 4).

6 396 G. Amarenco et al. / European Urology 43 (2003) Fig. 3. Distribution of CONTILIFE 1 scores for 504 women suffering from genuine stress incontinence according to their menopausal status (yes/no) DA, EA, SI, EM, SEX, WLB and Global: see Fig Discussion In its 1996 draft proposal for the standardisation of outcomes measures in clinical trials of continence care, the International Continence Society recommended inclusion of QoL data [17]. For this reason, the CON- TILIFE 1 questionnaire was recently designed for women suffering from any type of urinary incontinence. Moreover, as clinical trials for new pharmaceutical compounds tend to be increasingly more international, so that different language and culture versions of QoL questionnaires are needed, the French version of the CONTILIFE 1 was translated and culturally adapted into British English, Danish, Dutch, and German versions [5]. Whenever a QoL questionnaire is translated into a new language, assessment of the following psychometric properties are recommended: reliability, validity and responsiveness [18]. This psychometric evaluation is important in order to ascertain the credibility of the various versions of the questionnaire. The questionnaire can in fact be used as an assessment instrument for health and product outcomes in each country or to compare data obtained in the various countries. The main purpose of the present study was thus to evaluate the psychometric properties of the 5 language versions of CONTILIFE 1. Psychometric properties were tested using standardised procedures and instrument review criteria developed by the Scientific Advisory Committee of the Medical Outcomes Trust [12]. Results demonstrated that the QoL questionnaire was well accepted in all countries. The construct validity of the CONTILIFE 1 was good in the French, Danish and pooled samples (items were grouped on their dimensions) and acceptable in the Dutch, German and English samples. Item convergent and divergent validity were good. This good construct validity observed in the pooled sample indicates that score calculations for all dimensions can be made in pooled data during international studies. On the other hand, the construct validity in the study sample was not stable enough to allow direct comparison of scores between countries. The distribution of the CONTILIFE 1 scores was good, with low ceiling and floor effects, except for the Sexuality and Well-Being dimensions. Regarding the

7 G. Amarenco et al. / European Urology 43 (2003) Fig. 4. Distribution of change in CONTILIFE 1 scores for 459 women suffering from genuine stress incontinence between baseline and D28 according to change in the number of urinary leaks. Patients were split into 3 classes: improved (decrease of at least 50% in the number of urinary leaks under treatment); minimally improved (decrease between 25% and 50% in the number of urinary leaks), and stable (increase or decrease of less than 25% in the number of urinary leaks). DA, EA, SI, EM, SEX, WLB and Global: see Fig. 1. A decrease in QoL scores indicated an increase in QoL. Sexuality dimension, more than 50% of the women (except in the English samples) had the lowest possible score, which may explain at least partially the lack of responsiveness of this dimension. The internal consistency reliability was satisfactory. The study also demonstrated that clinical validity was excellent regarding the number of urinary leaks per week. Moreover no strong relationship was found regarding the age of the patients (except for the Self-Image and Sexuality dimensions) or their menopausal status. This confirms that CONTILIFE 1 is a specific measure of QoL as related to UI and not to ageing or menopause. Therefore, it is relevant and valid to use the CONTILIFE 1 scores as outcome measures in the treatment of UI. The responsiveness over time was good for all dimensions except Sexuality and Well-Being. The Daily Activities, Effort Activities, Self-image, and Emotional Consequences were the dimensions allowing the best discrimination between improved, minimally improved and stable patients. They are thus the most relevant and valid dimensions for the comparison of treatment efficacy in the study sample. 5. Conclusion The CONTILIFE 1 holds promise as an outcome measure in the treatment of stress urinary incontinence for large International studies. In the study sample from a European clinical study (France, Belgium, UK, The Netherlands, Germany, and Denmark) in the treatment of genuine stress incontinence, scores could be calculated on pooled data, and could be used as reliable and valid measures of the dimensions. The internal consistency reliability of the Effort Activities dimension

8 398 G. Amarenco et al. / European Urology 43 (2003) should be further assessed. Due to a better responsiveness to change over time in the study population, the Daily Activities, Effort Activities, Self-Image and Emotional Consequences were the most relevant and valid dimensions for the purpose of treatment efficacy comparison. Further validation work would be useful to confirm our findings with a broader coverage of countries and cultures. Furthermore, the CONTI- LIFE 1 was developed in both stress and urge incontinence: its encouraging psychometric performances in stress incontinence should be confirmed by a specific validation study in urge incontinence. References [1] Milsom I. The prevalence of urinary incontinence. Acta Obstet Gynecol Scand 2000;79: [2] Fitzpatrick R, Fletcher A, Gore S, Jones D, Spiegelhalter D, Cox D. Quality of life measures in health care. I: Applications and issues in assessment. Br Med J 1992;305: [3] Patrick DL, Martin ML, Bushnell DM, Marquis P, Andrejasich CM, Buesching DP. Cultural adaptation of a quality-of-life measure for urinary incontinence. Eur Urol 1999;36: [4] Girod I, McCarthy C, Marrel A, De La Loge C, Marquis P. Development and psychometric validation of a quality of life questionnaire in urinary incontinence. In: Abstracts presented to the 6th Annual Conference of ISOQOL, Barcelona, Spain, 3 6 November Qual Life Res 1999;7:658. [5] Conway K, Pouget C, Marquis P, Girod I, Mccarthy C. Linguistic validation of the urinary incontinence quality of life questionnaire. In: 4th Annual International Meeting, International Society for Pharmacoeconomics and Outcomes Research (ISPOR), Crystal City, USA, May Value Health 1999;3:203. [6] McKenna SP, Williamson T, Renck-Hooper U, Whalley D. The development of UK and Canadian English versions of the Incontinence Quality of Life Index (IqolI). J Outcomes Res 1997; 1:9 16. [7] Acquadro C, Jambon B, Ellis D, Marquis P. Language and translation issues. In: Spilker B, editor. Quality of life and pharmacoeconomics in clinical trials. Philadelphia: Lippincott-Raven; p [8] Sandvik H, Hunskaar S, Vanvik A, et al. Diagnostic classification of female urinary incontinence: an epidemiological survey corrected for validity. J Clin Epidemiol 1995;48(3): [9] Sandvik H, Hunskaar S, Seim A, et al. Validation of a severity index in female urinary incontinence and its implementation in an epidemiological survey. J Epidemiol Community Health 1993;47: [10] Streiner DL, Norman GR. Health measurement scales: a practical guide to their development and use. New York: Oxford University Press; [11] Zhang J, Quan H, Ng J, Stapanavage ME. Some statistical methods for multiple endpoints in clinical trials. Control Clin Trials 1997; 18: [12] Lohr KN, Aaronson NK, Alonso J. Evaluating quality-of-life and health status assessment instruments: development of scientific review criteria. Clin Ther 1996;18(5): [13] Campbell DT, Fiske JL. Convergent and discriminant validation by the multitrait multimethod matrix. Psychol Bull 1959;56: [14] Cronbach LJ. Coefficient alpha and the internal structure of tests. Psychometrica 1951;16: [15] Stewart AL, Hays RD, Ware J. Methods of validating MOS health measures. In: Stewart AL, Ware JE, editors. Measuring functioning and well-being: the medical outcomes study approach. Durham, NC: Duke University Press, p [16] Kazis LE, Anderson JJ, Meenan RF. Effects sizes for interpreting changes in health status. Med Care 1989;27:S [17] Kelleher CJ, Cardozo LD, Khullar V. A new questionnaire to assess the quality of life of urinary incontinent women. Br J Obstet Gynaecol 1997;104: [18] Hays RD, Anderson R, Revicki DA. Psychometric evaluation and interpretation of health related quality of life data. Oxford: Rapid Communications of Oxford Ltd.; 1995.

9 G. Amarenco et al. / European Urology 43 (2003)

10 400 G. Amarenco et al. / European Urology 43 (2003)

11 G. Amarenco et al. / European Urology 43 (2003)

12 402 G. Amarenco et al. / European Urology 43 (2003)

13 G. Amarenco et al. / European Urology 43 (2003)

14 404 G. Amarenco et al. / European Urology 43 (2003)