ing system, age-dependent changes were related with phase advance, reduction of circadian amplitude, and changes in the internal phase relationship between the sleep-wake cycle and the output of the circadian pace- maker (Duffy et al., 1998). Earlier morning awakening and decreased sleep consolidation in early morning are the major age-dependent changes in sleep (Foley et al., 1995). The hypothesis of the Harvard group was that an age-related advance in sleep-wake timing can- not be attributed to a shortening of the circadian period (Czeisler and Brown, 1999 Duffy et al., 2002) or to an advance phase (Duffy et al., 2002). These age- related changes appear to be due to an interaction between a reduction of the homeostatic drive for sleep and a reduced amplitude of the circadian signal pro- moting sleep at the end of the night (Dijk and Duffy, 1999), leading older people to wake up early relative to both clock time and internal circadian time. In the middle years of life, the phase advance of sleep-wake schedules and temperature circadian rhythms, the reduction in sleep consolidation, and theimpossibility of returning to sleep in the early morning have already been demonstrated (Carrier et al., 1997 Taillard et al., 1999). These age-related changes in sleep were not found to be associated to any modification in circa- dian amplitude or the entrained phase (Carrier et al., 2002). In young participants, morningness is associ- ated with a shorter intrinsic period (Duffy et al., 2002) and a longer phase wake time interval (entrained phase) (Duffy et al., 1999), which is not observed in older people (Duffy et al., 2002). On the other hand, older people rate themselves more morning types than young participants. The most frequently used morningness/ eveningness questionnaires (MEQ) are those devel- oped by Horne and Ostberg (1976) and by Smith et al. (1989), but they were validated in a population mostly comprising young students. They mainly use 4-choice items corresponding to a definite morning type, mod- erate morning type, moderate evening type, and defi- nite evening type. As age modifies sleep times, Horne and Ostberg already pointed out that a ���bedtime of 23:30 may be indicative of morning type within a stu- dent population, but might be more closely related to an evening type in the 40���60 age-group.��� Work demands also modify the natural preference of the individual (Mecacci and Zani, 1983 Adan and Almirall, 1992). While students have better opportu- nities than workers to follow their own circadian rhythms, workers tend to skew toward morningness more than students. Consequently, the distribution of chronotype is not normal and the number of evening types is reduced. Therefore, phase advance induced by age and/or work demand might play a major role when it comes to determining chronotype by the eveningness/ morningness score. In a previous study on more than 2000 adults of all ages (17���80 years), we found 40.2% morning-type participants, 48.9% neither-type partic- ipants, and 10.8% evening-type participants (Taillard et al., 1999). In 1165 middle-aged workers, there were 9.4% definitive morning type, 48% moderate morning type, 40.3% neither type, 2.2% moderate evening type, and 0.2% definitive evening type (Taillard et al., 2001). For these reasons, we postulate in this study that the skewed distribution of chronotype toward morning- ness is due to inappropriate classification. The aim of this study was therefore to adapt and standardize the MEQ of Horne and Ostberg to an active middle-aged population. Our study was divided into 4 steps : (a) to identify the principal com- ponents in the preferences in sleep and wake timing, (b) to identify groups of participants with similar sleep-wake habits, (c) to determine new cutoff scores to classify chronotypes, and (d) to evaluate this new cutoff by external validation. METHOD Five hundred sixty-six volunteers enrolled in the GAZELcohort were selected for this study. These par- ticipants were the good sleeper subgroup in our study of morbidity and chronotype(Taillard et al., 2001). The selection, methods, and characteristics of this sub- group have already been described. Briefly, they com- pleted the MEQ of Horne and Ostberg (19-item scale) (Horne and Ostberg, 1976), the Basic Nordic Sleep Questionnaire (BNSQ) (Partinen and Gislason, 1995), the Epworth Sleepiness Scale (Johns, 1991), the Lavigne scale (Lavigne and Montplaisir, 1994), and a sleep agenda. The inclusion criteria were not com- plaining of any sleep or mood disorder, not taking any hypnotic or psychotic drugs, not reporting poor sleep quality, not suffering from excessive daytime sleepi- ness, not suffering from periodic movement during sleep, not reporting sleep apnea during at least 5 days/week and not snoring loudly and intermittently for 3 to 5 days per week. Three hundred eighty-eight males and 178 females were selected. The mean age Taillard et al. / VALIDATION OF THE HORNE MEQ IN MIDDLE-AGED SAMPLE 77 at Vrije Universiteit 34820 on January 28, 2010 http://jbr.sagepub.com Downloaded from

was 51.2 �� 3.2 years (range: 44���58 years). According to occupational status, 3.7% had a low grade, 42.6% had a middle grade, and 53.7% had a high grade. The sam- ple consisted of 9.2% definitive morning type, 52.9% moderate morning type, 36.6% neither type, 2.1% moderate evening type, and 0.1% definitive evening type. A multiple correspondence analysis (MCA) was first performed to construct principal components. MCA is a descriptive multivariate method that limits the contribution of each of the questions of the MEQ to itself. This method makes it possible to derive a set of coordinate values summarizing the participants and the items in the questionnaire. Each individual is rep- resented in a multidimensional space based on the response to each item in the MEQ, where the distance between 2 points reflects their similarity: the shorter the distance, the greater the similarity. To observe dis- tances between points, they are projected on a plane of 2 orthogonal factors. These 1st factors (components) are chosen as the most representative of the global variance of the cloud. Only the gravity center of items in the MEQ is placed on the graph, thus allowing rapid identification of the different subgroups (chronotypes) and their morningness-eveningness profile. The questions that best contribute to the axis are those that have their categories spread right along the continuum and show a pattern of gradation. It may be particularly beneficial to ignore categories with a poor statistical power (small size), because they can distort results. Choices of items with a size lower than 2% (11 participants) were eliminated and were randomly assigned to another choice. Then, after the calculation of principal components, the eliminated choices were projected on the factorial plane. They were considered in the analysis as ���illustrative��� choices. In this way, the outliers are not eliminated from the analysis and the statistical power of the study is high. A hierarchical clustering analysis (HCA) was then performed to determine the most plausible number of groups (chronotypologies) and to identify a signifi- cant choice of items in the MEQ that would constitute homogeneous groups of participants. The purpose of cluster analysis is to place participants into groups suggested by the data, not defined a priori, so that par- ticipants in a given cluster tend to be somewhat simi- lar to each other. Participants with contrasted charac- teristics appear in different clusters. Statistical analysis for the external validation was performed with ANOVA followed by the Student- Newman-Keuls post hoc test for sleep log parameters and the Kruskal-Wallis (KW) test for BNSQ. MCA and HCA were carried out with the SPAD software package. ANOVA and KW tests were per- formed using the SPSS software package. RESULTS In general, few participants definitely classified themselves as evening type on extreme choices of items of the MEQ (i.e., bedtime after 0145 h). Fewer than 10 participants got up after 0945 h (item 1). Fewer than 5 volunteers went to bed before 2100 and after 0145 h (item 2). Seven found that getting up in the morning was notat all easy (item 4, choice 1), and 4 did not at all feel alert during the 1st half hour after getting up in the morning (item 5, choice 1). Fewer than 10 felt tired after 0100 h (item 10). Only 11 felt that they could best perform intellectual and physical activity after 1900 h (item 11, choice 4 and item 15, choice 4). Five stopped working between 0400 and 0800 h (item 17). Two volunteers felt best between 2200 and 0400 h (item 18). These 13 choices were eliminated owing to their size of 11 participants (size 2%). In the group, the preferential waking time and bed- time was 0748 �� 0.50 h and 2254 �� 0.48 h, respectively. The time when participants felt tired and in need of sleep was 2233 �� 1.17 h. The total score had a mean value of 59.6 �� 3.5. The distribution was not normal (Kolmogorov-Smirnov, p 0.0005). Skewness and flat- ness (Kurtosis) values were ���.53 and 0.36, indicating no serious deviation from normality. Identification of Preferences in Sleep and Wake Timing The 1st 3 principal factors (components) of the mul- tiple correspondence analysis were kept. The percent- age of the variance explained by the 1st component accounted for 32%, with 9% for the 2nd component, and 5% for the 3rd (total for the 3 components: 46%). For the 1st component, the most contributive item (13.2% of the variance explained) was morningness and eveningness assessed by item 19. (One hears about ���morning��� and ���evening��� types of people. Which one of these types do you consider yourself to be?) Evening type had negative values, whereas morning type had positive values. The other items that best defined axis 1 were, in decreasing order of importance, 11 (You wish to be at your peak perfor- 78 JOURNAL OF BIOLOGICAL RHYTHMS / February 2004 at Vrije Universiteit 34820 on January 28, 2010 http://jbr.sagepub.com Downloaded from