Lindberg Dr Department of Medical Sciences, Respiratory Medicine and Allergology, Akademiska sjukhuset, SE-751 85 Uppsala, Sweden. Tel.: 46 18 66 40 61; fax: 46 18 66 28 19.
Snoring is a major sign of obstructive sleep apnoea syndrome. Despite the frequent number of studies based on subjective reports of snoring, self-reported snoring has hardly been validated at all. In some previous epidemiological studies, a significant association between snoring and cardiovascular morbidity and mortality was found only below the age of 50–60 y. This study was performed to investigate whether this is due to a decrease in the validity of reported snoring with increasing age. In a population-based study, 2668 men aged 40–79 y answered a questionnaire including questions on snoring. Those who reported loud and disturbing snoring often or very often were regarded as habitual snorers. Without taking account of reported snoring, an age-stratified sample of these men was selected and their snoring was measured using a microphone for 1 night. Significant snoring was defined as recorded snoring sounds for ≥ 10% of the night. The participants were divided into younger (age 40–59, mean ± SD: 51.8 ± 4.6 y, n=132) and older (age 60–79, 67.7 ± 5.4 y, n=99) age groups. When analysing the validity of reported snoring, no significant differences were found between the younger and older age groups in terms of specificity [younger: 82% (95% CI 74–90%), older: 88% (81–95%)] or sensitivity [younger: 40% (26–54%), older: 35% (17–53%)]. These data indicate that, in men aged 40–79 y, the validity of reported snoring is similar in different age groups. The lack of an association between reported snoring and cardiovascular disease at higher ages can, therefore, not be explained by a decrease in the validity of reported snoring.
Snoring is a major sign of obstructive sleep apnoea syndrome (OSAS). In epidemiological studies, the most frequently used method for estimating snoring is questionnaires. However, even though everyone knows what snoring is, there is still some uncertainty about how to define it and there is no standard and uniformly accepted technique for the objective measurement of it ( Hoffstein 1996). As there is no ‘golden standard’ for objective measurements, the validation of these questionnaires remains a problem. In most of the validated questionnaires described in the literature, the validation did not deal with objectively measured snoring sounds but with their usefulness as a predictor of sleep apnoea ( Bliwise et al. 1991 ; Kump et al. 1994 ).
In some previous epidemiological studies designed to investigate the relationship between reported snoring and hypertension, significant associations have been found in younger men but not in older ones ( Koskenvuo et al. 1985 ; Lindberg et al. 1998a ). Furthermore, we recently reported that the combination of reported snoring and excessive daytime sleepiness is associated with an increase in mortality rate and cardiovascular mortality but, again, the effects appear to be age dependent ( Lindberg et al. 1998b ). In men aged 60 y and over, no association with mortality was found. A change in the validity of sleep questionnaires with age could be a possible explanation for this age-dependent relationship and this analysis was carried out to study whether this is the case.
In a population-based study, 2668 men aged 40–79 y participated in a 10-year follow-up in 1994 and answered a postal questionnaire including questions about snoring ( Lindberg et al. 1998a ). An age-stratified sample of 231 men from this study population underwent a single-night sleep recording with the main purpose of studying snoring and sleep-disordered breathing among hypertensives and controls. Owing to the skew age distribution in the population ( Lindberg et al. 1998c ) they were divided into four 10-year strata to increase the relative number of older men. Within each 10-year stratum, subjects with reported hypertension were selected randomly together with a control person without hypertension, matched for age and body mass index. Inclusion in this study was thereby performed, regardless of the reported snoring status. If a selected subject was excluded or refused to participate, another was selected randomly from the same age stratum. Subjects who had moved abroad or to other parts of Sweden and those who were not expected to manage to use the device due to severe disease were excluded.
We planned to investigate 240 subjects; 60 men aged 40–49 y, 80 men aged 50–59 y, 60 aged 60–69 y and 40 men aged 70–79 y. When comparing validity between age groups, the population was divided into younger men aged 40–59 y and older men aged 60–79 y.
The questionnaire used in 1994 consisted of 71 questions, relating mainly to sleep disturbances and associated daytime symptoms, somatic diseases and medication. Snoring was estimated using the multiple-choice question ‘How often do you snore loudly and disturbingly?’. The five scores in the answer were (1) never, (2) rarely, (3) sometimes, (4) often, and (5) very often ( Lindberg et al. 1998a , b, c). Those who replied with scores of 4 or 5 were regarded as habitual snorers. When answering the questionnaire, the subjects were not aware that they might subsequently be invited to take part in this validity study.
Snoring and sleep-related respiration were monitored for 1 night in the subjects’ homes using the Eden-Tec equipment ( Redline et al. 1991 ). The portable device included measurements of nasal/oral air flow (thermistry), chest-wall impedance and oxygen saturation (finger pulse oximetry). The device further included sound recording using a microphone attached at the level of the upper right corner of the cricothyroid cartilage to record snoring sounds. Sampling was performed at 10 Hz and all sounds of > 90 dB were recorded. Immediately upon waking in the morning, the subject filled in a diary about estimated sleep time during the night, time of falling asleep, awakenings during the night and final awakening in the morning. The recordings were transferred to a computer and scored manually on the screen for respiratory events and estimated sleep time. Total sleep time was estimated by visual assessments of the overnight tracing in conjunction with the subject’s diary. If the estimated total sleep time (TST) was less than 4 h, a new recording was performed.
Automatic analysis of time with snoring sounds during the estimated sleep time was performed. The percentage of the night with snoring sounds was then calculated as the ‘total time with snoring/estimated total sleep time × 100’.
The computations were performed on a Macintosh Ilci computer using the Statistica 4.0 software package (StatSoft Inc, Tulsa, OK, USA). The results are presented as the means ± SD. The Spearman rank correlation test was used to test for comparison of continuous variables and variables on an ordinal scale.
All subjects gave their informed consent and the protocol was approved by the ethics committee at the Medical Faculty at Uppsala University.
Acceptable whole-night recordings were obtained from 233 men. Two of these men had not answered the question on snoring in the questionnaire and were therefore excluded from the following analysis. Of the 231 men, 132 were in the younger age group, aged 40–59 y (mean ± SD: 51.8 ± 4.6 y), and 99 were in the older group, aged 60–79 y (mean ± SD: 67.7 ± 5.4 y).
Reported and measured snoring
On the basis of the answers in the questionnaire, 52 men (22.5%) reported that they snored loudly and disturbingly often or very often and were therefore regarded as habitual snorers. The mean time interval between the questionnaire and sleep recording was 24.8 ± 5.7 mo (range 14–36 mo). During the study night, snoring sounds were recorded on average for 8.7% of the night (range: 0–59%). Snoring sounds for ≥ 10% of the estimated sleeping time were recorded in 71 men (30.5%).
Reproducibility of measured snoring
Duplicate whole-night recordings were performed in 17 men. Intervals between recordings ranged from 21 to 35 days (26.1 ± 4.1, mean ± SD). There was no suggestion of a ‘first-night’ effect for measured snoring as snoring sounds were recorded on average for 8.0 ± 11% during the first study and 7.9 ± 12.3% during the second (P=0.8). A high correlation was noted between percentage of night with snoring sounds for replicate studies (r=0.83).
Validity of snoring
In the whole study group, there was a significant association between the score given to the question about snoring in the questionnaire and the percentage of the night with measured snoring sounds (r=0.38, P < 0.0001) ( Fig. 1). When significant snoring was defined as recorded snoring sounds for ≥ 10% of the night, the question used here to investigate snoring had a high specificity (84%, 95% CI 79–90%) while the sensitivity was low (38%, 95% CI 27–49%).
When analysing the validity of reported snoring in the younger and older age groups separately, no significant differences were found between the groups in terms of specificity, sensitivity or predictive values (Table 1).
Table 1. Sensitivity, specificity and predictive values for reported habitual snoring by age group
The main finding in this study is that there were no significant differences in the validity of snoring between men aged 40–59 y and those aged 60–79 y. The decrease in the association between reported snoring and cardiovascular consequences by age reported previously cannot therefore be explained by a decrease in the validity of reported snoring.
In this study the question used for snoring had a high specificity, compared with snoring measured by a microphone during 1 night, while the sensitivity was low. This is in accordance with the results of Ulfberg et al. (1996 ) who compared reported snoring with snoring measured by the Apnolog sleep apnoea registration equipment. However, measuring and quantifying the snoring sound is complicated due to the lack of standardized and accepted methods. Techniques for monitoring snoring that are described in the literature vary a great deal both when it comes to were to place the microphone and also how to process and analyse the signal ( Hoffstein 1996). There are also data indicating that the severity of snoring may be overestimated during polysomnographic recordings in hospital compared with recordings performed in the home environment ( Sériès et al. 1993 ). Hoffstein et al. compared both the technologist’s and the patient’s subjective impression of snoring during one night with the snoring sound recorded during nocturnal polysomnography in 613 unselected sleep clinic patients. They found poor agreement between the technologist’s and the patient’s rated snoring intensity while the technologist’s impression of snoring correlated to the number of snores per hour of sleep and to a lesser degree also to the mean nocturnal sound intensity ( Hoffstein et al. 1994 ). However, when counting snoring sounds manually the agreement between two listeners in judging snoring severity of the same patients is also only moderate ( Hoffstein et al. 1996b ), indicating that there are also discrepancies between individuals in the subjective perception of snoring.
Also, the procedure used here for validation is associated with many weaknesses. One important disadvantage is that polysomnography was not performed and total sleep time could therefore only be estimated. However, our main purpose was to compare the validity in the older and younger age groups. Identical criteria were used to estimate the sleep time for the two age groups and the person who performed the visual assessment was not informed about the age of the subject. The mean interval between answering the questionnaire and performing the recording was almost 2 y and this may lead to an underestimation of the actual correlation between reported and recorded snoring. There is, however, no reason to assume that changes due to the time delay differ between the older and younger age groups.
No ‘true’ cut-off point for significant snoring exists. The definition of significant snoring as recorded snoring sounds for ≥ 10% of the night could therefore be criticized. However, even though the calculations were performed with different definitions of snoring (≥ 15%, ≥ 20% of the night), the validity did not differ significantly between the younger and older age groups (not shown). Even though a cut-off point was used to compare the validity of reported snoring between age groups, this does not imply that measuring snoring is better than the subject’s own reports. One advantage with reported snoring is that it represents an integrated statement over time which cannot be achieved by whole-night measurements.
In studies designed to investigate the associations between snoring or sleep-disordered breathing and cardiovascular disease at different age groups the associations has been found to be age dependent with risk decreasing at older ages ( Koskenvuo et al. 1985 ; Lavie et al. 1995 ; Lindberg et al. 1998a , b). The results of the present analysis indicate that this is not due to a decrease in validity of snoring by age. In the studies cited above, BMI was the only anthropometric variable adjusted for, while neck circumference has been found to be better as a predictor for OSAS ( Hoffstein and Mateika 1992) and waist circumference for cardiovascular disease ( Larsson et al. 1984 ). The reliability of the results that there is an age dependence in the health consequences of snoring would obviously be affected if central obesity was more pronounced in younger than in older obese men. As far as we know, however, there is as yet no objective evidence to support such an argument.
We conclude that, in men aged 40–79 y, the validity of reported snoring is similar in different age groups. The decrease in the association between reported snoring and cardiovascular disease cannot, therefore, be explained by a decrease in the validity of snoring.
This study was supported financially by the Swedish Heart Lung Foundation and the Uppsala County Association Against Heart and Lung Diseases.