2 Material & Methods
AGE AGE „DOLE AGE-„DOLE-BM
4. Discussion
4.3. Factors associated with chronotype
4.3.1. Statistical models
The circadian clock is a self-sustained endogenous pacemaker regulated by the interaction of many known and more unknown genes (Roenneberg & Merrow, 2003). However, the clock is not a strict machine but is highly adaptive to entraining signals (chapter 1.3.; Roen- neberg et al, 2003b; Duffy & Wright, 2005). When asking people for their sleep-wake behav- iour (as the MCTQ does), one usually asks for an entrained rhythm under normal life condi- tions. It is, therefore, crucial to identify and quantify potential biological (or endogenous) and social (or environmental) influences that intra- and inter-individually have the potential to alter the phase of an entrained sleep-wake rhythm.
From previous results of this study it is hypothesized that chronotype (phase of entrainment, MSFSc) is independent of the ratio of sleep and wakefulness (α/ρ, SLD€). For both, an asso- ciation with AGE, GENDER, €DOLE, PHOTO, LAT, POR, and BMI was estimated using Multiple regression models. It has to be mentioned first that regression models do not fulfill all statistical assumptions (see chapter 3.4.2.2.). While deviations from normal distribution of
Discussion
with increasing age is a reason for concern. Especially the impact of the variable AGE can differ between conditions (Table 3.16.), ranging from 3.8% to 4.9% of total explained vari- ance of MSFSc. However, in this part of the analysis, it is not of prime interest to predict MSFScor SLD€using other variables but to I) identify potential influences and II) assess the impact of each influence on chronotype compared to all other influences included in the models. More interesting than absolute values of variance explained by potential influences is the ratio of explained variances and differences between Gender and Age groups.
No transformation has finally been applied because it didn’t improve the models in terms of normal distributions (Kolmogorov-Smirnov z-values still indicated a significant difference from normal distribution, Table 3.16a). Except for strong modifications (only Age >20 + MSFSc log10 + AGElog10), the order of variables, ranged by r2, remains the same (Table 3.16b). An even weaker effect of log-transformations or of a restriction to Age >20 is apparent for SLD€.
The first obvious observation is that MSFScand SLD€are differently associated with poten- tial influences. For the total sample population, AGE and GENDER show about the same association with MSFScand SLD€. LAT has been removed in both models, presumably be- cause the range of latitude was too low. Longitude might have been a better parameter to test. Within Germany, time of dawn and dusk can differ by 36 min (between Aachen and G–rlitz as western and eastern end, respectively). Depending on postal codes, longitude could be coded as continous variable to quantify a potential variability of MSFScand SLD€ due to a difference in the relation of sunrise and sunset to actual time. €DOLE, PHOTO, POR, and BMI, in contrast, are very differently associated with MSFScand SLD€. Further- more, there are differences between Work days & Free days, Females & Males, and Age ≤30 & Age >30 which will be discussed in detail.
4.3.2. AGE
A strong impact of AGE on MSFand SLDFand the weak impact of AGE on MSWand SLDW is not surprising. The great difference between work days and free days reflects sleep dept and recovery sleep (which leads to a delayed phase). Chronotype changes with age (Roen- neberg et al, 2004b; Dijk et al, 2000) and the impact of age is supressed on work days due to work schedules. On free days, the impact of age is the sum of actual chronotype and sleep dept because sleep dept depends on chronotype (Roenneberg et al, 2003a) and chronotype depends on age.
Sleep duration becomes shorter with increasing age (Fig.3.23.; Dijk et al, 2000) and, for most people, is restricted on work days, especially for later chronotypes (Roenneberg et al, 2003a). The different impact of AGE in Work day models and in Free day models again re- flects sleep dept and sleep recovery.
Discussion
The change of chronotype throughout life is different for females and males (Roenneberg et al, 2004b; Fig.1d, Appendix 1). Between 15 and 50 years of age, males are on average later types (have a later MSFSc) than females. The maximum difference is reached at about the age of 20 years, then decreasing constantly until fading out at about the age of 50. The range of average MSFScthroughout life is greater for males, thus age explains a higher amount of variance of MSFScfor males.
For both, phase of entrainment and sleep duration, results support prior observations on that issue (Roenneberg et al, 2003a; Roenneberg et al, 2004b).
4.3.3. GENDER
Females show an advanced MSFScand a longer SLD€compared to Males. The strength of association of GENDER with MSFand MSWis almost similar (Fig.3.25.). SLDFis very weakly associated with GENDER while the association with SLDWis relatively high. Thus, Females appear less sleep deprived on work days possibly because of an earlier sleep phase. Females can sleep earlier and reach a longer sleep duration before the alarm clock wakes them on work days.
The average difference of MSFScbetween Males and Females is higher for Age ≤ 30 than for Ages >30 (see 4.3.2.; Roenneberg et al, 2004b; Fig.1d, Appendix 1). The association of GENDER and MSFScis, therefore, higher for the younger group (Fig.3.4.5.). This difference is less pronounced for SLD€. The change of SLD€throughout life seems to be more similar between females and males than the change of MSFSc.
4.3.4. Average daily outside light exposure („DOLE), Photoperiod (PHOTO), and Lati-