- Email: [email protected]
Validation of a posturographic approach to monitor sleepiness
Journal of Biomechanics 43 (2010) 3214–3216
Contents lists available at ScienceDirect
Journal of Biomechanics journal homepage: www.elsevier.com/locate/jbiomech www.JBiomech.com
Short communication
Validation of a posturographic approach to monitor sleepiness ¨ c,d Pia Forsman a,b,n, Anders Wallin c, Edward Hæggstrom a
Sleep and Performance Research Center, Washington State University Spokane, USA Finnish Institute of Occupational Health, Helsinki, Finland Department of Physics, University of Helsinki, Helsinki, Finland d Helsinki Institute of Physics, Helsinki, Finland b c
a r t i c l e in f o
a b s t r a c t
Article history: Accepted 11 August 2010
Sleepiness is a major risk factor in traffic- and occupational accidents. While sleepiness is a persistent concern, there is no convenient test to monitor impending levels of sleepiness. We show that force platform posturographic balance testing addresses this need because it estimates time awake (TA) accurately and precisely. Testing the TA is appropriate because TA drives the sleep homeostatic process, a component in sleepiness. With 12 subjects we evaluated the accuracy and precision of repeated estimates of TA. Our extended study design that allows evaluating the accuracy and precision of posturographic TA-estimates is new. First, we tested the subjects’ balance every 2 h during 36 h of sustained wakefulness. This comprised the subjects’ reference curves (balance as a function of known and increasing TA). Then, we tested the subjects’ balance once a day over one week. We also tested the subjects’ balance once a week over one month. Finally, to estimate the subjects’ TA, we equated the balance scores with the scores in their reference curves. The accuracy of the estimates was 86%, and the precision was 97%. The high accuracy and precision of the estimates obtained with this one-month protocol validates the method of posturographic monitoring of sleepiness. So far, force platform posturographic balance testing has generally been used for clinical purposes, to quantify balance control and musculoskeletal performance. Our main result is that we now validated that balance testing provides accurate and precise estimates of TA, and hence, also provides an approach towards an automated monitor of sleepiness. & 2010 Elsevier Ltd. All rights reserved.
Keywords: Extended wakefulness Time awake Work safety Circadian rhythm Balance Quantitative sleepiness test
1. Introduction Sleepiness is related to accidents because it increases the risk for micro sleeps and lowers performance (Philip, 2005). For example, sleepiness ensuing from extending time awake (TA) beyond 16 h impairs performance as much as the legally proscribed blood alcohol concentration of 0.05% (Dawson and Reid, 1997). Whereas alcohol causes 20% of all road accidents, sleepiness causes 10% (Philip et al., 2001). Notably sleepiness causes 50% of all occupational accidents (Leger, 1994; Webb, 1995). Sleepiness is a normal physiological state that exhibits two trait-like concurrent processes: the TA-dependent sleep homeostatic process, and the time of day-dependent circadian rhythm (Baehr et al., 2000; Van Dongen et al., 2004). The sleepiness-issue has attracted interest from policymakers, but the lack of a breathalyzer-like test precludes convenient monitoring of
n Corresponding author at: Sleep and Performance Research Center, Washington State University Spokane, 702 South Campus Facility, P.O. Box 1495, Spokane, WA 99210-1495, USA. Tel.: + 1 509 358 7756; fax: +1 509 358 7810. E-mail address: [email protected] (P. Forsman).
0021-9290/$ - see front matter & 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.jbiomech.2010.08.013
exposed professions. The current sleepiness tests are expensive, time-consuming, invasive, biased by motivation, or subjective (Guilleminault and Brooks, 2001). Therefore, there is a need for a test that addresses these issues. Force platform posturography is traditionally used in clinical examinations to quantify balance control and musculoskeletal performance (Toppila et al., 2006). However, monitoring balance during sustained wakefulness has shown that balance exhibits the homeostatic- and circadian processes associated with sleepiness (Nakano et al., 2001; Avni et al., 2006). In fact, the homeostatic TA-dependency explains about 60% of the variance in the balance scores (Forsman et al., 2007a). This finding means that balance is a marker of sleepiness and it advocates the development of force platform posturographic balance testing to estimate TA. Moreover, in contrast to existing sleepiness tests, balance testing is fast, simple, objective, and reliable. With 20 subjects Forsman et al. (2007a) developed a procedure to estimate TA. First, they scored the subjects’ balance every 2 h during 36 h of sustained wakefulness (this generated the subjects’ reference curves). On another occasion they scored the subjects’ balance once. Then, to estimate the subjects’ TA at this occasion, they equated the balance scores with those in the reference
P. Forsman et al. / Journal of Biomechanics 43 (2010) 3214–3216
20
3215
10
16
14 12 10 8
6 4
y = 0.88 x + 1.35 r 2 = 0.73, F =156, p < 0.0001 n = 84
2
Estimated hours awake
Estimated hours awake
18
8 6
4 2
y = 0.88 x + 0.92 r 2 = 0.69, F =72.1, p < 0.0001 n = 48
0
0
0
2
4
6
8 10 12 14 16 18 20
0
Hours awake
2
4
6
8
10
Hours awake
Fig. 1. Scatter plot and linear regression of estimated against actual TA. Left pane shows estimates in the Second experiment. Right pane shows estimates in the Third experiment. Significant correlations (p o0.0001) between the estimated- and actual TA were found in both experiments.
curves. Balance was scored as the feedback time (FT) of balance control. The FT reflects how the balance control allows the body to drift from its equilibrium. The FT increases with decreasing balance control (Collins et al., 1995). The results in Forsman et al. (2007a) are promising but rely on one estimate per subject. This prevents evaluating the precision of the method. Moreover, the 2007 paper does not report the accuracy of the estimates. A valid sleepiness test for monitoring purposes in the field has to be accurate and precise. Therefore, this study aims to evaluate the accuracy and precision of repeated TA-estimates with an one-month protocol. If the accuracy and precision are high, this study validates the method.
2. Methods
The COP traces were filtered with a 0.3 s moving average. Balance scores were expressed as FT (equations in Forsman et al., 2007a, 2007b; Collins et al., 1995).
2.3. Analyses First experiment: We generated the subjects’ reference curves by plotting the balance scores against the known TA and known time-of-day (Forsman et al., 2007a) exemplifies reference curves). Second and third experiment: We estimated the TA at the time of testing in the second and third experiments employing the procedure in Forsman et al. (2007a): (1) superimpose the current test score onto the reference curve, (2) locate instant(s) where the test score equals the reference score(s), (3) accept the instant at which the time-of-day is closest to that of the current test score. We expressed the results as estimated- and actual TA at the time of the tests. We expressed the probability that an estimate is within 7 2.5 h of the actual TA as the positive predictive value (PPV), negative predictive value (NPV), sensitivity, specificity, accuracy and precision of the estimates (Altman and Bland, 1994).
2.1. Subjects, design, and procedure Twelve subjects (10 men and 2 women, age 21–38) participated in the study after giving their informed consent. Each subject participated in three experiments that were conducted in series. In the first experiment, the subjects were kept awake for 36 h (from 6:00 until 18:00 the next day) during which their balance was tested every 2 h starting from 8:00. This experiment provided the subjects’ reference curves. In the second experiment, the same subjects arrived for a balance test once a day during one week. This experiment generated seven TA-estimates per subject. Each subject was scheduled to arrive for testing with a TA that differed from his/ her previous ones by at least 2.5 h.1 This assured spread in the seven TAcoordinates; the range during the week was 17.5 h. The TA was not revealed to the experimenter or analyzer (blinded test). In the third experiment, the same subjects arrived for a balance test once a week during one month. This experiment generated four TA-estimates per subject. Each subject was scheduled to arrive for testing on the same weekday, on the same time-of-day, and with the same TA. This assured constant TA-coordinates and conditions during the month. In the second and third experiment the subjects did not spend the time between the balance tests in a controlled lab environment. This emulated workplace monitoring. With individual sleep-logs we checked that all subjects arrived rested to the experiments (criterion: 7–9 h of sleep during two nights preceding the test).
3. Results 3.1. First experiment The subjects’ reference curves showed that balance deteriorates during sustained wakefulness: between the 2nd and 36th hour the mean balance impairment was 2.66% per hour. Regression analysis of the reference curves showed a linear correlation between balance and TA that roughly accounted for 50% of the balance variance (F ¼21.6, Po0.01, R2 ¼0.52, n ¼12).
3.2. Second experiment The subject was tested with different TA-coordinates once a day for seven days. Regression analysis showed a linear correlation between the estimated- and the actual TA that accounted for 70% of the variance (Fig. 1). Table 1 shows the statistics of the estimates.
2.2. Balance testing and calculation of feedback time Balance was tested with a force platform (Toppila et al., 2006) on which the subject stood: unshod, feet together, arms crossed over the chest, looking at an eye-high fix-point provided for visual reference in front of the platform. The platform sampled body center-of-pressure (COP) excursions at 1 kHz for 30 s.
1 Based on the rate of change in balance scores, and the procedure in Forsman et al. (2007a).
3.3. Third experiment The subject was tested with constant TA-coordinates once a week for one month. Regression analysis showed a linear correlation between the estimated- and the actual TA that accounted for 70% of the variance (Fig. 1). Table 1 shows the statistics of the estimates.
3216
P. Forsman et al. / Journal of Biomechanics 43 (2010) 3214–3216
Table 1 Statistics of the TA-estimates in the present study and another study using the same method (Forsman et al., 2007a). Total increase in the FT-values in the present study (over 36 h) and a study comparing young and elderly subjects (22 vs. 75 years, Collins et al., 1995).
Experiment 1 Experiment 2a Experiment 3b Forsman et al. Collins et al.
PPV (%)
NPV (%)
Sensitivity (%)
Specificity (%)
Accuracy (%)
Precision (%)
Slopec(h/h)
DFTd(%)
N/A 57.4 65.2 69.0 N/A
N/A 91.4 97.6 93.0 N/A
N/A 55.6 82.9 56.0 N/A
N/A 91.3 93.9 96.0 N/A
N/A 85.5 92.6 N/A N/A
N/A 96.5 99.3 N/A N/A
N/A 0.88 0.87 1.02 N/A
90 N/A N/A N/A 56
N/A ¼not available or not addressed by study design. a
Seven days 12 subjects¼ 84 estimates. Four weeks 12 subjects ¼ 48 estimates. c The slope of the regression line in Fig. 1 quantifies the correspondence between the estimated- and actual TA. d FT increases as balance control decreases (Collins et al., 1995). b
4. Discussion Our results support earlier findings that balance control deteriorates during extended wakefulness (Nakano et al., 2001; Avni et al., 2006; Forsman et al., 2007a, 2007b). Our results also support the earlier findings that TA can be estimated with posturography (Forsman et al., 2007a), by showing that the accuracy of the estimates is high. Most importantly, the onemonth protocol in this study design extends that of Forsman et al. (2007a) and thereby allows evaluating the precision of the estimates (with several estimates per subject). Specifically, the one-month protocol shows that the precision of repeated posturographic TA estimates is high – both in conditions with variable TA (Experiment 2) and constant TA (Experiment 3). The high precision shows that the personal reference curve permits making estimates one month after its generation (calibration interval). Summing up, the method is valid and can be automated to facilitate implemented sleepiness monitoring. Calibration is needed less frequently than every month. Validating the results of the current study against those found in literature is hard because to our knowledge only one study has reported on posturographic TA estimates. The correspondence between the estimated- and actual TA in this study supports the validity of the method, but Forsman et al. (2007a) reported a higher correspondence (Table 1). They also reported higher PPV, NPV, sensitivity, and specificity of the estimates than we found (Table 1). An explanation for the discrepancies may be that the subjects in this study did not spend their time during Experiments 2 and 3 in a controlled laboratory environment. The overall decrease in balance that we observed in this study is in line with the decrease that Collins et al. (1995) found when comparing elderly and young subjects (Table 1). Three method-related issues and a study-limitation need commenting. First, personal reference curves are necessary since both the balance scores and the circadian phase and amplitude differ between subjects (Collins et al., 1995; Baehr et al., 2000). However, a 4% inter-individual variability in FT was reported in a group of 30 subjects (Forsman et al., 2007b). The small variability allowed the baseline scores to be separated (on group scale with statistical significance) from those recorded 2 and 5 h after the baseline test. This indicates that method generalization might be possible (making personal reference curves unnecessary). Second, the local time at the time of testing must be recorded to reduce the false positive rate of the estimate caused by the circadian variation (Forsman et al., 2007a). Third, this small-scale study underscores the potential of posturographic sleepiness monitoring. However, to prove test credibility requires large-scale
replication studies, as well as refinements to measurement- and analysis methods to further improve the accuracy of the estimates. Moreover, the current study validates the method for conditions where the actual TA is at most 17.5 h (Experiments 2 and 3). This limitation comes from the fact that we scheduled the testing in Experiments 2 and 3 according to the subjects’ ability to attend. As wakefulness extending beyond 17.5 h is common among exposed professions, studies with longer TA at testing would add method credibility.
Conflict of interest statement The authors declare no conflict of interests. References Altman, D., Bland, M., 1994. General practice statistics notes: diagnostic tests 2: predictive values. British Medical Journal 309, 102. Avni, N., Avni, I., Barenboim, E., Azaria, B., Zadok, D., Kohen-Raz, R., Morad, Y., 2006. Brief posturographic test as an indicator of fatigue. Psychiatry and Clinical Neurosciences 60, 340–346. Baehr, E., Revelle, W., Eastman, C., 2000. Individual differences in the phase and amplitude of the human circadian temperature rhythm: with an emphasis on morningness–eveningness. Journal of Sleep Research 9, 117–127. Collins, J., De Luca, C., Burrows, A., Lipsitz, L., 1995. Age-related changes in openloop and closed-loop postural control mechanisms. Experimental Brain Research 104, 480–492. Dawson, D., Reid, K., 1997. Fatigue, alcohol and performance impairment. Nature 388, 235. ¨ ainen, ¨ ¨ Forsman, P., Wallin, A., Tietav A., Hæggstrom, E., 2007a. Posturographic sleepiness monitoring. Journal of Sleep Research 16, 259–261. ¨ ¨ I., 2007b. Daytime Forsman, P., Hæggstrom, E., Wallin, A., Toppila, E., Pyykko, changes in postural stability and repeatability of posturographic measurements. Journal of Occupational and Environmental Medicine 49, 591–596. Guilleminault, C., Brooks, S., 2001. Excessive daytime sleepiness: a challenge for the practicing neurologist. Brain 124 (Pt 8), 1489–1491. Leger, D., 1994. The cost of sleep-related accidents: a report for the national commission on sleep disorders research. Sleep 17, 84–93. Nakano, T., Araki, K., Michimori, A., Inbe, H., Hagiwara, H., Koyama, E., 2001. Nineteen-hour variation of postural sway, alertness and rectal temperature during sleep deprivation. Psychiatry and Clinical Neurosciences 55, 277–278. Philip, P., 2005. Sleepiness of occupational drivers. Industrial Health 43, 30–33. Philip, P., Vervialle, F., Le Breton, P., Taillard, J., Horne, J., 2001. Fatigue, alcohol, and serious road crashes in France: factorial study of national data. British Medical Journal 322, 829–830. ¨ I., Starck, J., Tossavainen, T., Uitti, J., Oksa, P., 2006. Toppila, E., Forsman, P., Pyykko, Effect of styrene on postural stability among reinforced plastic boat plant workers in Finland. Journal of Occupational and Environmental Medicine 48, 175–180. Van Dongen, H., Baynard, M., Maislin, G., Dinges, D., 2004. Systematic interindividual differences in neurobehavioral impairment from sleep loss: evidence of trait-like differential vulnerability. Sleep 27, 423–433. Webb, W., 1995. The cost of sleep-related accidents: a reanalysis. Sleep 18, 276–280.
Contents lists available at ScienceDirect
Journal of Biomechanics journal homepage: www.elsevier.com/locate/jbiomech www.JBiomech.com
Short communication
Validation of a posturographic approach to monitor sleepiness ¨ c,d Pia Forsman a,b,n, Anders Wallin c, Edward Hæggstrom a
Sleep and Performance Research Center, Washington State University Spokane, USA Finnish Institute of Occupational Health, Helsinki, Finland Department of Physics, University of Helsinki, Helsinki, Finland d Helsinki Institute of Physics, Helsinki, Finland b c
a r t i c l e in f o
a b s t r a c t
Article history: Accepted 11 August 2010
Sleepiness is a major risk factor in traffic- and occupational accidents. While sleepiness is a persistent concern, there is no convenient test to monitor impending levels of sleepiness. We show that force platform posturographic balance testing addresses this need because it estimates time awake (TA) accurately and precisely. Testing the TA is appropriate because TA drives the sleep homeostatic process, a component in sleepiness. With 12 subjects we evaluated the accuracy and precision of repeated estimates of TA. Our extended study design that allows evaluating the accuracy and precision of posturographic TA-estimates is new. First, we tested the subjects’ balance every 2 h during 36 h of sustained wakefulness. This comprised the subjects’ reference curves (balance as a function of known and increasing TA). Then, we tested the subjects’ balance once a day over one week. We also tested the subjects’ balance once a week over one month. Finally, to estimate the subjects’ TA, we equated the balance scores with the scores in their reference curves. The accuracy of the estimates was 86%, and the precision was 97%. The high accuracy and precision of the estimates obtained with this one-month protocol validates the method of posturographic monitoring of sleepiness. So far, force platform posturographic balance testing has generally been used for clinical purposes, to quantify balance control and musculoskeletal performance. Our main result is that we now validated that balance testing provides accurate and precise estimates of TA, and hence, also provides an approach towards an automated monitor of sleepiness. & 2010 Elsevier Ltd. All rights reserved.
Keywords: Extended wakefulness Time awake Work safety Circadian rhythm Balance Quantitative sleepiness test
1. Introduction Sleepiness is related to accidents because it increases the risk for micro sleeps and lowers performance (Philip, 2005). For example, sleepiness ensuing from extending time awake (TA) beyond 16 h impairs performance as much as the legally proscribed blood alcohol concentration of 0.05% (Dawson and Reid, 1997). Whereas alcohol causes 20% of all road accidents, sleepiness causes 10% (Philip et al., 2001). Notably sleepiness causes 50% of all occupational accidents (Leger, 1994; Webb, 1995). Sleepiness is a normal physiological state that exhibits two trait-like concurrent processes: the TA-dependent sleep homeostatic process, and the time of day-dependent circadian rhythm (Baehr et al., 2000; Van Dongen et al., 2004). The sleepiness-issue has attracted interest from policymakers, but the lack of a breathalyzer-like test precludes convenient monitoring of
n Corresponding author at: Sleep and Performance Research Center, Washington State University Spokane, 702 South Campus Facility, P.O. Box 1495, Spokane, WA 99210-1495, USA. Tel.: + 1 509 358 7756; fax: +1 509 358 7810. E-mail address: [email protected] (P. Forsman).
0021-9290/$ - see front matter & 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.jbiomech.2010.08.013
exposed professions. The current sleepiness tests are expensive, time-consuming, invasive, biased by motivation, or subjective (Guilleminault and Brooks, 2001). Therefore, there is a need for a test that addresses these issues. Force platform posturography is traditionally used in clinical examinations to quantify balance control and musculoskeletal performance (Toppila et al., 2006). However, monitoring balance during sustained wakefulness has shown that balance exhibits the homeostatic- and circadian processes associated with sleepiness (Nakano et al., 2001; Avni et al., 2006). In fact, the homeostatic TA-dependency explains about 60% of the variance in the balance scores (Forsman et al., 2007a). This finding means that balance is a marker of sleepiness and it advocates the development of force platform posturographic balance testing to estimate TA. Moreover, in contrast to existing sleepiness tests, balance testing is fast, simple, objective, and reliable. With 20 subjects Forsman et al. (2007a) developed a procedure to estimate TA. First, they scored the subjects’ balance every 2 h during 36 h of sustained wakefulness (this generated the subjects’ reference curves). On another occasion they scored the subjects’ balance once. Then, to estimate the subjects’ TA at this occasion, they equated the balance scores with those in the reference
P. Forsman et al. / Journal of Biomechanics 43 (2010) 3214–3216
20
3215
10
16
14 12 10 8
6 4
y = 0.88 x + 1.35 r 2 = 0.73, F =156, p < 0.0001 n = 84
2
Estimated hours awake
Estimated hours awake
18
8 6
4 2
y = 0.88 x + 0.92 r 2 = 0.69, F =72.1, p < 0.0001 n = 48
0
0
0
2
4
6
8 10 12 14 16 18 20
0
Hours awake
2
4
6
8
10
Hours awake
Fig. 1. Scatter plot and linear regression of estimated against actual TA. Left pane shows estimates in the Second experiment. Right pane shows estimates in the Third experiment. Significant correlations (p o0.0001) between the estimated- and actual TA were found in both experiments.
curves. Balance was scored as the feedback time (FT) of balance control. The FT reflects how the balance control allows the body to drift from its equilibrium. The FT increases with decreasing balance control (Collins et al., 1995). The results in Forsman et al. (2007a) are promising but rely on one estimate per subject. This prevents evaluating the precision of the method. Moreover, the 2007 paper does not report the accuracy of the estimates. A valid sleepiness test for monitoring purposes in the field has to be accurate and precise. Therefore, this study aims to evaluate the accuracy and precision of repeated TA-estimates with an one-month protocol. If the accuracy and precision are high, this study validates the method.
2. Methods
The COP traces were filtered with a 0.3 s moving average. Balance scores were expressed as FT (equations in Forsman et al., 2007a, 2007b; Collins et al., 1995).
2.3. Analyses First experiment: We generated the subjects’ reference curves by plotting the balance scores against the known TA and known time-of-day (Forsman et al., 2007a) exemplifies reference curves). Second and third experiment: We estimated the TA at the time of testing in the second and third experiments employing the procedure in Forsman et al. (2007a): (1) superimpose the current test score onto the reference curve, (2) locate instant(s) where the test score equals the reference score(s), (3) accept the instant at which the time-of-day is closest to that of the current test score. We expressed the results as estimated- and actual TA at the time of the tests. We expressed the probability that an estimate is within 7 2.5 h of the actual TA as the positive predictive value (PPV), negative predictive value (NPV), sensitivity, specificity, accuracy and precision of the estimates (Altman and Bland, 1994).
2.1. Subjects, design, and procedure Twelve subjects (10 men and 2 women, age 21–38) participated in the study after giving their informed consent. Each subject participated in three experiments that were conducted in series. In the first experiment, the subjects were kept awake for 36 h (from 6:00 until 18:00 the next day) during which their balance was tested every 2 h starting from 8:00. This experiment provided the subjects’ reference curves. In the second experiment, the same subjects arrived for a balance test once a day during one week. This experiment generated seven TA-estimates per subject. Each subject was scheduled to arrive for testing with a TA that differed from his/ her previous ones by at least 2.5 h.1 This assured spread in the seven TAcoordinates; the range during the week was 17.5 h. The TA was not revealed to the experimenter or analyzer (blinded test). In the third experiment, the same subjects arrived for a balance test once a week during one month. This experiment generated four TA-estimates per subject. Each subject was scheduled to arrive for testing on the same weekday, on the same time-of-day, and with the same TA. This assured constant TA-coordinates and conditions during the month. In the second and third experiment the subjects did not spend the time between the balance tests in a controlled lab environment. This emulated workplace monitoring. With individual sleep-logs we checked that all subjects arrived rested to the experiments (criterion: 7–9 h of sleep during two nights preceding the test).
3. Results 3.1. First experiment The subjects’ reference curves showed that balance deteriorates during sustained wakefulness: between the 2nd and 36th hour the mean balance impairment was 2.66% per hour. Regression analysis of the reference curves showed a linear correlation between balance and TA that roughly accounted for 50% of the balance variance (F ¼21.6, Po0.01, R2 ¼0.52, n ¼12).
3.2. Second experiment The subject was tested with different TA-coordinates once a day for seven days. Regression analysis showed a linear correlation between the estimated- and the actual TA that accounted for 70% of the variance (Fig. 1). Table 1 shows the statistics of the estimates.
2.2. Balance testing and calculation of feedback time Balance was tested with a force platform (Toppila et al., 2006) on which the subject stood: unshod, feet together, arms crossed over the chest, looking at an eye-high fix-point provided for visual reference in front of the platform. The platform sampled body center-of-pressure (COP) excursions at 1 kHz for 30 s.
1 Based on the rate of change in balance scores, and the procedure in Forsman et al. (2007a).
3.3. Third experiment The subject was tested with constant TA-coordinates once a week for one month. Regression analysis showed a linear correlation between the estimated- and the actual TA that accounted for 70% of the variance (Fig. 1). Table 1 shows the statistics of the estimates.
3216
P. Forsman et al. / Journal of Biomechanics 43 (2010) 3214–3216
Table 1 Statistics of the TA-estimates in the present study and another study using the same method (Forsman et al., 2007a). Total increase in the FT-values in the present study (over 36 h) and a study comparing young and elderly subjects (22 vs. 75 years, Collins et al., 1995).
Experiment 1 Experiment 2a Experiment 3b Forsman et al. Collins et al.
PPV (%)
NPV (%)
Sensitivity (%)
Specificity (%)
Accuracy (%)
Precision (%)
Slopec(h/h)
DFTd(%)
N/A 57.4 65.2 69.0 N/A
N/A 91.4 97.6 93.0 N/A
N/A 55.6 82.9 56.0 N/A
N/A 91.3 93.9 96.0 N/A
N/A 85.5 92.6 N/A N/A
N/A 96.5 99.3 N/A N/A
N/A 0.88 0.87 1.02 N/A
90 N/A N/A N/A 56
N/A ¼not available or not addressed by study design. a
Seven days 12 subjects¼ 84 estimates. Four weeks 12 subjects ¼ 48 estimates. c The slope of the regression line in Fig. 1 quantifies the correspondence between the estimated- and actual TA. d FT increases as balance control decreases (Collins et al., 1995). b
4. Discussion Our results support earlier findings that balance control deteriorates during extended wakefulness (Nakano et al., 2001; Avni et al., 2006; Forsman et al., 2007a, 2007b). Our results also support the earlier findings that TA can be estimated with posturography (Forsman et al., 2007a), by showing that the accuracy of the estimates is high. Most importantly, the onemonth protocol in this study design extends that of Forsman et al. (2007a) and thereby allows evaluating the precision of the estimates (with several estimates per subject). Specifically, the one-month protocol shows that the precision of repeated posturographic TA estimates is high – both in conditions with variable TA (Experiment 2) and constant TA (Experiment 3). The high precision shows that the personal reference curve permits making estimates one month after its generation (calibration interval). Summing up, the method is valid and can be automated to facilitate implemented sleepiness monitoring. Calibration is needed less frequently than every month. Validating the results of the current study against those found in literature is hard because to our knowledge only one study has reported on posturographic TA estimates. The correspondence between the estimated- and actual TA in this study supports the validity of the method, but Forsman et al. (2007a) reported a higher correspondence (Table 1). They also reported higher PPV, NPV, sensitivity, and specificity of the estimates than we found (Table 1). An explanation for the discrepancies may be that the subjects in this study did not spend their time during Experiments 2 and 3 in a controlled laboratory environment. The overall decrease in balance that we observed in this study is in line with the decrease that Collins et al. (1995) found when comparing elderly and young subjects (Table 1). Three method-related issues and a study-limitation need commenting. First, personal reference curves are necessary since both the balance scores and the circadian phase and amplitude differ between subjects (Collins et al., 1995; Baehr et al., 2000). However, a 4% inter-individual variability in FT was reported in a group of 30 subjects (Forsman et al., 2007b). The small variability allowed the baseline scores to be separated (on group scale with statistical significance) from those recorded 2 and 5 h after the baseline test. This indicates that method generalization might be possible (making personal reference curves unnecessary). Second, the local time at the time of testing must be recorded to reduce the false positive rate of the estimate caused by the circadian variation (Forsman et al., 2007a). Third, this small-scale study underscores the potential of posturographic sleepiness monitoring. However, to prove test credibility requires large-scale
replication studies, as well as refinements to measurement- and analysis methods to further improve the accuracy of the estimates. Moreover, the current study validates the method for conditions where the actual TA is at most 17.5 h (Experiments 2 and 3). This limitation comes from the fact that we scheduled the testing in Experiments 2 and 3 according to the subjects’ ability to attend. As wakefulness extending beyond 17.5 h is common among exposed professions, studies with longer TA at testing would add method credibility.
Conflict of interest statement The authors declare no conflict of interests. References Altman, D., Bland, M., 1994. General practice statistics notes: diagnostic tests 2: predictive values. British Medical Journal 309, 102. Avni, N., Avni, I., Barenboim, E., Azaria, B., Zadok, D., Kohen-Raz, R., Morad, Y., 2006. Brief posturographic test as an indicator of fatigue. Psychiatry and Clinical Neurosciences 60, 340–346. Baehr, E., Revelle, W., Eastman, C., 2000. Individual differences in the phase and amplitude of the human circadian temperature rhythm: with an emphasis on morningness–eveningness. Journal of Sleep Research 9, 117–127. Collins, J., De Luca, C., Burrows, A., Lipsitz, L., 1995. Age-related changes in openloop and closed-loop postural control mechanisms. Experimental Brain Research 104, 480–492. Dawson, D., Reid, K., 1997. Fatigue, alcohol and performance impairment. Nature 388, 235. ¨ ainen, ¨ ¨ Forsman, P., Wallin, A., Tietav A., Hæggstrom, E., 2007a. Posturographic sleepiness monitoring. Journal of Sleep Research 16, 259–261. ¨ ¨ I., 2007b. Daytime Forsman, P., Hæggstrom, E., Wallin, A., Toppila, E., Pyykko, changes in postural stability and repeatability of posturographic measurements. Journal of Occupational and Environmental Medicine 49, 591–596. Guilleminault, C., Brooks, S., 2001. Excessive daytime sleepiness: a challenge for the practicing neurologist. Brain 124 (Pt 8), 1489–1491. Leger, D., 1994. The cost of sleep-related accidents: a report for the national commission on sleep disorders research. Sleep 17, 84–93. Nakano, T., Araki, K., Michimori, A., Inbe, H., Hagiwara, H., Koyama, E., 2001. Nineteen-hour variation of postural sway, alertness and rectal temperature during sleep deprivation. Psychiatry and Clinical Neurosciences 55, 277–278. Philip, P., 2005. Sleepiness of occupational drivers. Industrial Health 43, 30–33. Philip, P., Vervialle, F., Le Breton, P., Taillard, J., Horne, J., 2001. Fatigue, alcohol, and serious road crashes in France: factorial study of national data. British Medical Journal 322, 829–830. ¨ I., Starck, J., Tossavainen, T., Uitti, J., Oksa, P., 2006. Toppila, E., Forsman, P., Pyykko, Effect of styrene on postural stability among reinforced plastic boat plant workers in Finland. Journal of Occupational and Environmental Medicine 48, 175–180. Van Dongen, H., Baynard, M., Maislin, G., Dinges, D., 2004. Systematic interindividual differences in neurobehavioral impairment from sleep loss: evidence of trait-like differential vulnerability. Sleep 27, 423–433. Webb, W., 1995. The cost of sleep-related accidents: a reanalysis. Sleep 18, 276–280.