INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Keywords: genetics; obesity; sleepiness; snoring; twins

Two cardinal symptoms of sleep apnea are snoring and excessive daytime sleepiness (EDS) (1). Although objective measurements of these symptoms can be made in the laboratory, patient presentations of these conditions often involve self- report, which may have only limited relationships to objective measures. Despite the frequent lack of correspondence between self-reports and objective assessments of sleep apnea, a better understanding of how patients present symptoms would still provide important information and may be useful in clinical practice (4). In addition, the widespread inclusion of sleep apnea symptoms in more recent epidemiological surveys necessitates a better understanding of the factors that determine the variation and covariation of self-reported symptoms of snoring and EDS (5).

In this article, we analyze self-reports of snoring and sleepiness in a large population sample of elderly male twins. Because heritability has been previously documented for sleep apnea (6), genetic modeling of primary symptoms could further elucidate the relative contributions of genetic and nongenetic influences to intermediate phenotypes related to the expression of this syndrome. To take further advantage of a large, well-characterized, nonclinical population for whom physiologic measures of sleep-disordered breathing were not yet available, we investigated in this study the relative contributions of genetic and nongenetic (environmental) influences to variation and covariation in sleep apnea symptoms. In addition, we explored the genetic overlap of snoring and sleepiness with obesitya major risk factor for sleep apnea known to be largely under genetic control. Specifically, we addressed the following questions:

1. What are the contributions of genetic and nongenetic (environmental) influences to self-reports of snoring and EDS in a large sample of older male twins?

2. Is the relationship between snoring and sleepiness due to common genetic or common nongenetic (environmental) influences?

3. Do genetic influences involved in obesity contribute to the genetic variance in snoring and daytime sleepiness?

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects in the present study are male twin pairs from the National Academy of Sciences-National Research Council (NAS-NRC) World War II Twin Registry, surveyed in 1998-1999 for general health and a number of health behaviors. The methods used to construct this twin panel have been described in previous publications (9, 10). Data analyzed in the present study are from 1560 intact twin pairs (818 monozygotic [MZ] and 742 dizygotic [DZ] pairs).

Snoring was determined by responses to three questions: (1) How often do you snore in any way? (2) How often do you snore loudly and disruptively? (3) How often do you hold your breath during sleep? Questions were based on the time frame of the previous 6 mo and could be answered with any of the following responses: "never," "just a few times," "sometimes," "fairly often," or "don't know." These questions have been used extensively in studies of clinical and nonclinical populations and have been validated polysomnographically (11). Daytime sleepiness was assessed with the Epworth Sleepiness Scale (ESS) (12), included in the most recent health survey of the NAS-NRC Twin Registry.

Maximum likelihood model fitting (13) was used to decompose the observed phenotypic variation into genetic and environmental components of variance. The genetic variance is assumed to be due to multiple additive (A) genetic influences, and the environmental variance may be due to environmental factors shared by twins reared in the same family environment (C) and to nonshared environmental factors (E). Their influences on the phenotype are given by parameters h, c, and e, which are equivalent to the standardized regression coefficients of the phenotype on the latent factors A, C, and E, respectively. The amount of variance due to each source is the square of the corresponding parameter.

Multivariate genetic analysis was used for modeling the phenotypic correlation between pairs of variables (e.g., snoring and obesity). Figure 1 shows the model for one member of a twin pair. A_c and E_c are the genetic and environmental influences common to obesity and snoring, and A_s and E_s are the genetic and environmental influences specific to snoring and independent of obesity. The common effects of A_c and E_c on snoring and obesity are represented by parameters h_c and h_c' , and e_c and e_c' , respectively. To test the significance of the common genetic and environmental factors, parameters h_c and h_c' and e_c and e_c' were constrained to zero and to test the significance of specific genetic and environmental factors, parameters h_s and e_s were constrained to zero. Submodels that were nested in each other were compared by hierarchical ² tests, using Mx software (14). The genetic correlation r_g was calculated from the path coefficients in Figure 1, using the formula r_g = h_ch_c' / (h_c' ² + h_s² )^1/2. All the bivariate genetic analyses were conducted within a structural equation framework and should be viewed as exploratory in nature.

View larger version (12K): [in this window] [in a new window]   Figure 1.   Bivariate genetic model. Ac and Ec are genetic and nonshared environmental influences common to obesity and snoring; As and Es are genetic and nonshared environmental influences specific to snoring. The effects of Ac and Ec are represented by parameters hc, h'c , and ea' , ec' , respectively, whereas specific As and Es effects are represented by parameters hs and es.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

In this sample of older male twins, 26% were habitual snorers, 18% reported excessive daytime sleepiness (ESS  11), and 29% were obese (body mass index [BMI]  28). For the sample as a whole, the correlations were as follows: between BMI and ESS scores, phenotypic correlation (r_ph) = 0.18, p < 0.001; between BMI and snoring, r_ph = 0.13, p < 0.00; and between snoring and ESS scores, r_ph = 0.12, p < 0.001. A high percentage of subjects, between 21 and 27% of the cohort, answered questions about snoring with a "don't know" response. We conducted a comparison of those reporting low and high levels of snoring with "don't know" respondents and found that compared with nonsnorers "don't know" respondents were older and more obese (higher BMI). We previously reported similar findings suggesting that "don't know" responses may confer a high likelihood of having sleep apnea (15) and decided for the purpose of this analysis to include the "don't know" respondents in the snoring category.

Before proceeding with the genetic analyses, we examined the within-pair and cross-twin cross-trait correlations for each zygosity group. These provide a means of determining whether genetic or environmental parameters are likely to explain the variation and covariation within and among traits. Table 1 summarizes the pairwise correlations, with MZ correlations presented below the diagonal and DZ correlations presented above the diagonal. The within-pair correlations for obesity, snoring, and sleepiness reveal that the MZ within-pair correlations (0.6494, 0.2370, and 0.4118) are greater than the corresponding DZ correlations (0.2923, 0.0925, and 0.1664), suggesting a contribution of genetic influences to individual differences in self-reports of snoring, sleepiness, and obesity in this cohort. There is a significant relationship between phenotypes within the same individual (underlined data in Table 1) and because the cross-twin, cross-trait correlations are higher for MZ twins than for DZ twins (off-diagonal italic data in Table 1) the notion of genetic commonality is also supported by these data.

The next step in the analysis was to fit the bivariate genetic model of Figure 1 to the correlation data summarized in Table 1. Table 2 presents estimates of genetic and environmental correlations calculated from the path coefficients of the best models fitting the bivariate correlations in Table 1. Also listed in Table 2 are maximum-likelihood estimates of (1) the phenotypic correlation (r_ph), (2) the proportions of the phenotypic correlation due to genetic and environmental influences (hrh and ere), (3) the genetic and environmental correlation (r_g and r_e), and (4) the proportions of common and specific genetic variance shared and not shared with obesity. To calculate the proportion of the phenotypic correlation due to genetic influences, we multiplied the genetic correlation by the square root of heritabilities (e.g., hr_gh' = 70% is the square root of the heritability of obesity times the genetic correlation r_g times the square root of the heritability of snoring).

From Table 2, we see that both the genetic correlation (r_g = 0.25) and the environmental correlation (r_e = 0.10) were statistically significant, suggesting common genetic and environmental influences between obesity and sleepiness. In contrast, the phenotypic association between obesity and snoring and between snoring and sleepiness was mediated entirely by common genetic and not common environmental influences. Interestingly, in spite of the significant genetic overlap in the phenotypic association between sleepiness and obesity, genes in common with obesity accounted for a mere 6% of the genetic variance in sleepiness. Similarly, only 30% of the total genetic variance in snoring was accounted for by genes in common with obesity. Individual estimates of heritability were as follows: for obesity, 64%; for sleepiness, 40%; and for snoring, 23%. Table 3 lists the individual components of variance with corresponding 95% confidence intervals (CIs) for heritability estimates obtained from univariate genetic analyses. When "don't know" responses were included in the snoring category, the heritability of snoring was 20% with a corresponding 95% confidence interval of 12 to 31%. Similarly, by using different cutoff points in the definition of EDS and obesity (e.g., EDS  14 and BMI  30), we observed minor changes in estimates of heritability.

In summary, the biometric analyses of these data suggest a significant contribution of genetic influences to self-reports of snoring and daytime sleepiness in the elderly. However, because we based our results on questionnaire data, and genetically influenced traits are expected to be stable over time, we further tested the genetic hypothesis by administering a follow-up survey to a subset of 360 twins from this cohort. Average time of follow-up between the baseline and repeat survey was 8 wk. By using the two-time survey data, we were able to test subject consistency on responses to snoring, daytime sleepiness, and self-reports of height and weight. We found 66% agreement (p < 0.001) in subject responses to snoring loudly and disruptively. The test-retest correlation for the full ESS scale was r = 0.73, and test-retest correlations for height and weight were r = 0.96 and r = 0.97, respectively (both p < 0.001).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

In this study, we used the twin design to determine the extent to which individual differences in self-reports of snoring and daytime sleepiness are due to genetic or nongenetic (environmental) influences. Our results indicate that snoring and excessive daytime sleepiness are independently influenced by genetic factors, and their co-occurrence in the same individuals may be largely due to common genetic influences. We also investigated the nature of the association between obesity and subject self-reports of snoring and daytime sleepiness and found evidence of a significant genetic correlation. The estimated genetic variance, however, of snoring and sleepiness could not be explained entirely by genes involved in obesity, suggesting that many other genetic influences specific to snoring and sleepiness may be involved.

Four epidemiologic studies (two of which were twin studies) have previously studied the familial aggregation of sleep apnea symptoms (16). Kaprio and coworkers (16), in their study of more than 4,000 Finnish twins, found that concordance for snoring was greater between MZ twins than between DZ twins, suggesting a role for inheritance. Ferini-Strambi and coworkers (19), studying 776 twin pairs, showed stronger associations between recognized risk factors and snoring in DZ twins discordant for snoring than in MZ discordant pairs, which can also be interpreted as evidence of a predisposing genetic factor for snoring. In addition to snoring, Ferini-Strambi and coworkers (19) and Heath and coworkers (20) have investigated the answers to a single question dealing with excessive daytime sleepiness, and McCarren and coworkers (21) investigated the heritability of waking up from sleep unrefreshed. These later studies generally corroborate a moderate contribution of genetic influences to self-reports on items dealing with excessive daytime sleepiness. The present study, however, is the first to estimate the heritability of excessive daytime sleepiness by using a threshold value derived from the ESS.

Consistent with previous genetic studies of sleep apnea symptoms (19, 20), we found in this study that the majority of the variation in self-report symptoms was due to nongenetic (environmental) influences not shared between cotwin brothers. It is important, however, to recognize that measurement error is also included in the proportion of nongenetic (environmental) component of variance.

In the multivariate analysis, snoring and sleepiness were significantly correlated with obesity, and the sources of this phenotypic correlation could be attributed primarily to genetic influences in common with obesity. Obesity did show a nongenetic (environmental) correlation with sleepiness, accounting for 28% of the phenotypic correlation, in contrast to the correlation between snoring and sleepiness that was entirely due to common genetic and not common environmental influences.

Because of differences in designs and analytic measures, the present results are difficult to compare directly with those of previous investigations. Redline and coworkers (18) estimated the degree of familial aggregation of sleep apnea symptoms and related risk factors, including age, sex, BMI, and selected cardiovascular risk factors (smoking, alcohol, physical activity), and found that the familial aggregation of sleep apnea symptoms was not fully explained by the familial clustering of risk factors. Previous MZ cotwin analyses examining the association between snoring and sleepiness and between snoring and BMI also found weak associations suggesting the presence of genetic interactive effects (22). In the present study, the contributions of obesity genes to the genetic variation in snoring and sleepiness were relatively small, suggesting that individual differences in symptoms commonly associated with sleep apnea are only partly explained by genes associated with obesity. In the case of snoring, other genetic influences could be operating through factors affecting upper airway caliber (e.g., craniofacial structure), whereas for sleepiness, the genetic component may involve sleep schedules and nocturnal timing of rapid eye movement (REM) sleep (23, 24).

In summary, these data support the hypothesis that inheritance of sleep apnea symptoms is likely to be polygenic, involving aspects of physiology determined by multiple genetic and nongenetic (environmental) factors. However, as with any self-report measure, the data analyzed in the present study have some fundamental limitations. First, both ESS scores and self-reports of snoring are subject to situational factors. Second, reported correlations between measures of subjective daytime sleepiness (including the ESS) and more objective measures of EDS, such as the Multiple Sleep Latency Test (MSLT), are often low (25, 26). It therefore remains to be seen whether objective measures such as the MSLT or the Maintenance of Wakefulness Test (MWT) reveal comparable levels of heritability and similar configurations of common genetic and nongenetic variance with obesity. Given, however, reported estimates of heritability for a host of other psychological symptoms (27, 28), one could contend that subjective ratings of sleep apnea might be particularly affected by subtle response biases. Thus, studies using objective measurements of snoring and daytime sleepiness are needed before making a definitive statement regarding the heritability of sleep apnea.