Abstract
To investigate the association between sleep apnea syndrome (SAS) and automobile accidents, and to evaluate potential underlying mechanisms, we prospectively recruited 60 consecutive patients with SAS (apnea–hypopnea index, 58 ± 3 h− 1) and 60 healthy control subjects, matched for sex and age. The number of automobile accidents during the past 3 yr was obtained from participants and insurance companies. We quantified the degree of daytime sleepiness (Epworth scale), anxiety and depression (Beck tests), and we assessed the level of vigilance (PVT 192) and driving performance (Steer-Clear). Patients had more accidents than control subjects (OR: 2.3; 95% CI: 0.97 to 5.33) and were more likely to have had more than one accident (OR: 5.2; 95% CI: 1.07 to 25.29, p < 0.05). These differences persisted after stratification for km/yr, age, and alcohol consumption. Patients were more somnolent, anxious, and depressed than control subjects (p < 0.01), and they had a lower level of vigilance and poorer driving performance (p < 0.01). Yet, we did not find any correlation between the degree of daytime sleepiness, anxiety, depression, the number of respiratory events, nocturnal hypoxemia, level of vigilance, or driving simulator performance and the risk of automobile accidents among SAS patients. In conclusion, patients with SAS have an increased risk of automobile accidents. None of the clinical or physiological markers commonly used to define disease severity appear able to discriminate those patients at higher risk of having an automobile accident. Barbé F, Pericás J, Muñoz A, Findley L, Antó JM, Agustı́ AGN, de Lluc Joan M. Automobile accidents in patients with sleep apnea syndrome: an epidemiological and mechanistic study.
It is widely assumed that the risk of automobile accidents is increased in patients with sleep apnea syndrome (SAS) (1, 2), particularly in those with severe disease (3, 4). However, as recently pointed out by Pakola and coworkers (5) most studies included a small number of subjects (3, 4, 6) and were often poorly controlled. On the other hand, large studies were based upon retrospective, self-reported questionnaires that were not later confirmed in independent data banks (police, insurance companies) (7-11); this can introduce a significant bias (12). Finally, some studies included patients whose diagnosis was not based upon a full polysomnography (7, 9), and potentially important confounding variables such as alcohol intake, drug usage, and number of kilometers driven per year were often not taken into account (3, 4, 6, 7). Likewise, no previous study has investigated in these patients the relationship between the risk of an automobile accident and multiple theoretical risk factors, i.e, disease severity, daytime vigilance, driving simulator performance, and measurements of anxiety and depression. Yet, a better delineation of the relationship between SAS and automobile accidents is an extremely relevant issue because of its social, sanitary, and economic implications (11, 13-17).
We designed a retrospective controlled study to compare the incidence of automobile accidents in a cohort of patients with severe SAS and in a control group of subjects individually matched for sex and age. Data on automobile accidents were obtained both from the subjects and the insurance companies, after obtaining appropriate signed consent. Finally, we investigated in the laboratory several potential mechanisms that may contribute to increase the risk of automobile accidents in these patients.
We prospectively considered potential candidates for inclusion in the study all those subjects attended by the sleep unit of our institution during 1996, who showed more than 20 apneas–hypopneas per hour of sleep during a full, supervised, standard polysomnographic study (Ultrasom Nicolette, Madison, WI). Inclusion criteria also required: (1) a valid driving license and (2) being a permanent resident in our community (Mallorca, Spain). Candidates were excluded if they were drug abusers, had a psychiatric disorder, were shift-workers, have been diagnosed with epilepsy, narcolepsy and/or periodic leg movements disease. Accordingly, we screened 88 candidates and excluded 28 of them. Nineteen subjects did not have a driving license, three were shift-workers, two had a psychiatric disorder, one had epilepsy, one suffered from periodic leg movements, one lived outside Mallorca most of the time, and one refused to participate. This selection process yielded 60 patients (59 males) who were finally included in the study. Simultaneously, we studied 61 healthy volunteers individually matched for age (± 5 yr) and sex, who served as controls. They were nonmedical workers or visitors of our hospital, who were referred to us by other faculty members of our institution not involved in the study. Patient relatives were specifically excluded. Control subjects were considered adequate for the purpose of the study if they fulfilled the same inclusion and exclusion criteria shown above for patients except for the presence of SAS. This was excluded on the basis of the clinical history (18) and, if necessary, after a full polysomnography. The latter was indicated in two individuals: SAS was excluded in one and confirmed in the other, who was therefore excluded from the study. Accordingly, a total of 60 individuals (59 males) served as control subjects. All participating subjects were fully informed of the nature of the study, and gave their written consent. The protocol was approved by the ethics committee of our institution.
We used a standardized questionnaire to record data on occupation, the mean number of hours slept per day, alcohol intake, relevant previous medical conditions and prescribed drugs, and the mean number of kilometers driven per year. The number of automobile accidents during the previous 3 yr was obtained in all participants both from the subject and the data bases of the insurance company (or companies). In our community insurance companies are the only reliable objective source of the occurrence of all automobile accidents. According to previous publications (6), automobile accident was defined as those that resulted in property damage greater than $500 and/or personal injury. We used the Epworth scale (19) to quantify the degree of subjective sleepiness of the participating individuals, and we used the Beck questionnaires to measure depression and anxiety (20).
The level of vigilance was explored using the Psychometer Vigilance Test device (PVT 192; CWE, Inc., Ardmore, PA) (21). In brief, a red light flashed on the top of the device at random intervals (80 to 85 times) throughout the duration of the test (10 min). Subjects were asked to press a button as soon as they realize that the red light flashed. The period of time elapsed between these two events (reaction time [RT]) was recorded by the computer for later analysis. In each individual, results were expressed as mean RT and, also, as the slope of the relationship between the inverse of RT versus time. This slope measures reaction fatigue (21). A negative slope indicates an increase of the RT over time (i.e., more reaction fatigue).
The level of alertness while driving and the ability of the subject to perform a task requiring prolonged vigilance was investigated using a computer program (Steer-Clear; Findley Fabrizio, Charlottesville, VA) (22). Briefly, during 30 min the subject sits in front of the computer screen, where a small car runs through a long, linear road. Unexpectedly, and at random intervals, an object (steer) appears in the road. The individual is asked to press the space bar to avoid hitting the steer. We set the test such that 500 steers appeared during the 30 min of its duration. The number of hits was recorded by the computer every minute. In each subject, data were later analyzed and expressed as the percentage of hits through the test.
Results are shown as mean ± SEM. A two-tailed chi-square test (Fisher exact test) was used to assess the statistical significance of differences between patients and control subjects with respect to reported automobile accidents during the past 3 yr and the consumption of drugs with potential influence upon driving abilities. Odds ratio (OR) and 95% confidence intervals (95% CI) were calculated. We used an unconditional logistic regression to assess the effects of potential confounding variables. Variables associated with both the outcome and the disease were included in the logistic model for adjustment. Continuous variables were compared between groups using a two-tailed independent t test or a Mann-Whitney test whenever variances were not equal (Levene's test). A p value less than 0.05 was considered significant.
Mean age of the patients and matched control subjects was 47 ± 1 yr. As expected, body mass index was higher in patients than in controls (33 ± 0.8 versus 27 ± 0.8, p < 0.001). The mean apnea–hypopnea index in patients was 58 ± 3 per hour (range, 21 to 101). Mean alcohol consumption tended to be higher in patients (17 ± 4 versus 9 ± 2 g/d, p = 0.06); these differences were statistically significant for the alcohol intake during the weekend (31 ± 6 versus 18 ± 3 g/d, p < 0.05). Both groups consumed a similar amount of drugs with potential influence upon driving abilities (β-blockers, oral antidiabetics, angiotensin converting enzyme inhibitors, and antidepressant drugs) except for benzodiazepines, which were consumed more often in patients (n = 7) than in control subjects (n = 1) (p = 0.06). Patients and control subjects reported sleeping a similar number of hours per day (8.6 ± 0.2 versus 8.0 ± 0.2 h/d, p = 0.8). Daytime sleepiness, depression, and anxiety scores were higher in patients than in control subjects (Table 1). Patients had more errors than controls during the Steer-Clear test and showed a longer reaction time and a tendency to have more reaction fatigue (Table 1).
| Cases | Controls | p Value | ||||
|---|---|---|---|---|---|---|
| Sleepiness score (Epworth scale) | 12 ± 1 | 3 ± 0.3 | < 0.001 | |||
| Depression score (Beck test) | 8 ± 1 | 4 ± 1 | < 0.001 | |||
| Anxiety score (Beck test) | 9 ± 1 | 4 ± 1 | < 0.01 | |||
| Steer-Clear, % hits | 2 ± 0.5 | 0.4 ± 0.1 | < 0.01 | |||
| Reaction time, msec | 283 ± 6 | 262 ± 5 | < 0.01 | |||
| Reaction fatigue, msec−1/min | −0.04 ± 0.007 | −0.03 ± 0.004 | < 0.07 |
The percentage of patients with SAS who had at least one automobile accident in the past 3 yr was higher (33%) than in control subjects (18%) (OR: 2.3; 95% CI; 0.97 to 5.33, p = 0.06). Patients with SAS had a higher mean number of accidents during the previous 3 yr than control subjects (0.53 ± 0.1 versus 0.22 ± 0.06, p < 0.05) and were more likely to have had more than one accident (OR: 5.2; 95% CI: 1.07 to 25.29, p < 0.05). Figure 1 shows the distribution of the number of accidents in both groups. Patients drove more kilometers per year than control subjects (27,305 ± 2,905 versus 15,695 ± 1,580 km/yr, p < 0.01). After stratifying for potential confounding variables (including the number of kilometers driven per year), the increased proportion of accidents in patients with SAS did persist with a similar magnitude (Table 2). Potential confounders were also assessed using unconditional logistic regression and were found not to modify the reported association. After adjusting for the number of kilometers driven, the association between SAS and having had one or more accidents did increase slightly (OR: 2.6, 95% CI: 1.06 to 6.43, p < 0.05). Identical results were obtained when the number of kilometers driven was categorized in two or four levels.
| Cases (%) | Controls (%) | |||
|---|---|---|---|---|
| Age, yr | ||||
| < 47.5 | 40 | 20 | ||
| ⩾ 47.5 | 29 | 16 | ||
| Distance driven per year, km | ||||
| < 15.000 | 40 | 20 | ||
| ⩾ 15.000 | 32 | 16 | ||
| Alcohol intake, g/d | ||||
| < 9.5 | 43 | 23 | ||
| ⩾ 9.5 | 26 | 14 |
To assess the relationship between disease severity and the risk of automobile accidents we investigated whether there was an increased risk of accidents among patients with SAS according to their level of clinical impairment, objective polygraphic data, or driving performance (Figure 2). We found that neither clinical data (depression, anxiety, or referred daytime sleepiness) nor markers of disease severity (apnea–hypopnea index, time below 90% SaO2 at night, and mean nocturnal SaO2 ) were related to the number of accidents (Figure 2). By contrast, those patients with a worse reaction time or a higher reaction fatigue appear to have a higher number of accidents (Figure 2). Yet, differences failed to reach statistical significance. Performance in the computer simulator was not significantly related to the risk of automobile accidents (Figure 2).
Fig. 2. Occurrence of automobile accidents in patients with SAS according to the quartile distribution of several clinical and physiological markers of disease severity. In all panels, except on those related to mean SaO2 , time below 90% SaO2 , and reaction fatigue, a higher quartile indicates more abnormal response. No significant difference (one-way analysis of variance) was detected in any of these variables.
[More] [Minimize]The potential impact of SAS upon public health is currently under intense debate (13, 14, 16, 17). Among several potentially relevant issues, an important one is whether or not the risk of automobile accidents is increased in these patients (15, 17). The main findings of our study were that: (1) the occurrence of automobile accidents was indeed increased in patients with SAS; (2) this increased risk was independent of the number of kilometers driven, age of the patient, or average alcohol consumption; and (3) disease severity (evaluated by the apnea–hypopnea index, different levels of nighttime oxygenation, subjective daytime sleepiness, anxiety or depression level) was not related to the number of automobile accidents.
The assumption of increase risk of automobile accidents in patients with SAS is mostly based on studies with small numbers of subjects which were often poorly controlled (5). In our study, we included a large group of patients with severe SAS (in whom a higher occurrence of automobile accidents can be expected [3, 4]), and we compared them with a control group of healthy individuals matched for sex and age. We obtained information on automobile accidents both from the patient and the insurance companies, and we considered in our analysis the effects of several potentially confounding factors. We found (Figure 1) that the occurrence of automobile accidents in patients with SAS was increased with respect to control subjects (OR: 2.3; 95% CI: 0.97 to 5.33). Also patients were more likely to have had more than one accident (OR: 5.2; 95% CI: 1.07 to 25.29, p < 0.05).
We found that patients with SAS were more somnolent, anxious and depressed, and performed poorly in several of the tests assessed in the laboratory than control subjects (Table 1). However, neither clinical variables (anxiety, depression, or daytime sleepiness) nor several indices of severity (such as the apnea–hypopnea index or the degree of nocturnal hypoxemia) were related to the number of accidents in patients with SAS (Figure 2). This observation is at variance with the suggestion that automobile accidents are more common in patients with severe disease (3, 4). Yet, other studies using different techniques to explore daytime sleepiness (multiple sleep latency test) also failed to report a significant association with the risk of automobile accidents (11). This lack of association between several indices of disease severity and the risk of automobile accidents admits, at least, two potential explanations. First, although the variables we used to define disease severity are commonly used for this purpose (23), they may lack discriminant capacity to detect differences between groups. Second, the mechanisms linking SAS and increased risk of automobile accidents may not have been explored directly enough in our study (2). Only the assessment of reaction time (and reaction fatigue) showed a trend to discriminate patients according to the number of accidents (Figure 2). Although differences between groups failed to reach statistical significance, patients at the worse end of the quartile distribution of the mean reaction time and reaction fatigue showed a clearly increased risk (Figure 2). It is likely that, if the number of patients included in the study would have been greater, differences would have reached statistical significance. Because of the potential ability of this test to identify patients at risk of an automobile accident, future studies will have to explore this possibility further.
An unexpected but interesting finding was to observe that patients with SAS drove a significantly higher mean number of kilometers per year than control subjects. It is unlikely that a selection bias can explain this observation because an identical proportion of patients (70%) and control subjects (68%) came from areas surrounding our hospital. Further, this observation is in keeping with the data reported recently in preliminary form from a case-control study (24). Two potential explanations should be considered. On the one hand because of the repetitive and tedious nature of driving, it is possible that those patients with SAS who drive a significant number of kilometers per year become aware of their disease more often (and, perhaps, earlier) than those who did not. On the other hand, the possibility that driving a significant number of kilometers per year may by itself constitute a risk factor for the development of SAS should also be considered. We accept that this is unlikely because, currently, it is very difficult to offer any meaningful mechanism underlying this possibility. However, at least on theoretical grounds, it should be considered in any rigorous analysis of our findings. In this context, it is also interesting to note that other investigators have shown that the incidence of SAS among commercial long-haul truck drivers is higher than in the general population (25).
Our study design has several limitations which may have influenced our results. First, we have assessed the car accidents during the 3 yr preceding the diagnosis of SAS. However, we believe that the chronic nature of this entity allows us to assume that the disease preceded the accident. Second, because our control subjects were not a random sample from the general population, we cannot totally exclude that their rate of traffic accidents was an unbiased estimate of the rate of accidents in the population from which our subjects with SAS were selected. However, through individual matching we tried to improve that comparability between patients and control subjects.
In summary, our study shows that the risk of automobile accidents is increased in patients with SAS. Yet, none of the clinical or physiological markers commonly used to define disease severity appear able to discriminate those patients at higher risk of having an automobile accident.
The authors thank the ongoing support of Dr. J. C. Gonzalez-Luque (Dirección General de Tráfico) and the helpful comments of Dr. B. Togores and Dr. J. Sauleda (Servei Pneumologia, Hospital Son Dureta). They would also like to acknowledge the secretarial assistance of Mr. Ll. Collell.
Supported in part by Dir. General de Trafico, ABEMAR, and Carburos Metálicos SA.
| 1. | Findley L. J., Levinson M. P., Bonnie R. J.Driving performance and automobile accidents in patients with sleep apnea. Clin. Chest Med131992427435 |
| 2. | Strohl K. P., Bonnie R. J., Findley L., Fletcher E. C., Getsy J., Kryger M. H., Millman R., Novak R., Roth T., Walsleben J., Williams A.Sleep apnea, sleepiness, and driving risk. Am. J. Respir. Crit. Care Med.150199414631473 |
| 3. | Findley L., Fabrizio M., Thommi G., Suratt P. M.Severity of sleep apnea and automobile crashes. N. Engl. J. Med3201989868869 |
| 4. | Findley L. J., Weill J. W., Jabour E. R.Drivers with untreated sleep apnea. Arch. Intern. Med151199114511452 |
| 5. | Pakola S. J., Dinges D. F., Pack A. I.Driving and sleepiness: review of regulations and guidelines for commercial and noncommercial drivers with sleep apnea and narcolepsy. Sleep181995787796 |
| 6. | Findley L. J., Unverzagt M. E., Suratt P. M.Automobile accidents involving patients with obstructive sleep apnea. Am. Rev. Respir. Dis1381988337340 |
| 7. | George C., Nickerson P., Hangly P.Sleep apnoea patients have more automobile accidents. Lancet85561987447 |
| 8. | Cassel W., Ploch T., Becker C., Dugnus D., Peter J. H., Von Wichert P.Risk of traffic accidents in patients with sleep-disordered breathing: reduction with nasal CPAP. Eur. Respir. J9199626062611 |
| 9. | Haraldsson P., Carenfelt C., Tingvall C.Sleep apnea syndrome symptoms and automobile driving in a general population. J. Clin. Epidemiol451992821825 |
| 10. | Wu H., Yan-Go F.Self-reported automobile accidents involving patients with obstructive sleep apnea. Neurology46199612541257 |
| 11. | Aldrich M. S.Automobile accidents in patients with sleep disorders. Sleep121989487494 |
| 12. | Findley L., Buckner R., Guchu R., Findley J., Suratt P.Self reporting of automobile accidents underestimates the occurrence of serious accidents in subjects with sleep apnea (abstract). Am. J. Respir. Crit. Care Med1491994A54 |
| 13. | Phillipson E. A.Sleep apnea—a major public health problem. N. Engl. J. Med328199312711273 |
| 14. | Young T., Palta M., Dempsey J., Skatrud J., Weber S., Badr S.The occurrence of sleep-disordered breathing among middle-aged adults. N. Engl. J. Med328199312301235 |
| 15. | Leger D.The cost of sleep-related accidents: a report for the national commission on sleep disorders research. Sleep1719948493 |
| 16. | Stradling J.Sleep apnoea and the misuse of evidence-based medicine. Lancet3491997201202 |
| 17. | Wright J., Johns R., Watt I., Melville A., Sheldon T.Health effects of obstructive sleep apnoea and the effectiveness of continuous positive airways pressure: a systematic review of the research evidence. B.M.J3141997851860 |
| 18. | Kapuniai L. E., Andrew D. J., Crowell D. H., Pearce J. W.Identifying sleep apnea from self-reports. Sleep111988430436 |
| 19. | Johns M. W.A new method for measuring daytime sleepiness: the Epworth sleepiness scale. Sleep141991540545 |
| 20. | Beck, A. T. 1967. Depression: Clinical Experimental and Theoretical Aspects. Harper & Row, New York. |
| 21. | Dinges, D. F., and N. B. Kribbs. 1991. Performing while sleepy: effects of experimentally induced sleepiness. In T. H. Monk, editor. Sleep, Sleepiness and Performance. John Wiley, Chichester. 97–128. |
| 22. | Findley L., Fabricio M., Knight H., Norcross B. B., Laforte A. J., Suratt P.Driving simulator performance in patients with sleep apnea. Am. Rev. Respir. Dis1401989529530 |
| 23. | Stradling J. R.Sleep studies for sleep related breathing disorders. Sleep11992265273 |
| 24. | Cordero J., Teran J., Jimenez A., Gomez, Garcia I., Quintana I., Carpizo R., Golpe R., Garcia M., Jimenez A., Sanz, Cifrian J.Obstructive sleep apnea syndrome (OSAS) and traffic accidents: a case-control study (abstract). Eur. Respir. J101997A227 |
| 25. | Stoohs R. A., Bingham L., Itoi A., Guilleminault C., Dement W. C.Sleep and sleep-disordered breathing in commercial long-haul truck drivers. Chest107199512751282 |
