Maximizing sensitivity of the psychomotor vigilance test (PVT) to sleep loss

Abstract

Study objectives: The psychomotor vigilance test (PVT) is among the most widely used measures of behavioral alertness, but there is large variation among published studies in PVT performance outcomes and test durations. To promote standardization of the PVT and increase its sensitivity and specificity to sleep loss, we determined PVT metrics and task durations that optimally discriminated sleep deprived subjects from alert subjects.

Design: Repeated-measures experiments involving 10-min PVT assessments every 2 h across both acute total sleep deprivation (TSD) and 5 days of chronic partial sleep deprivation (PSD).

Setting: Controlled laboratory environment.

Participants: 74 healthy subjects (34 female), aged 22-45 years.

Interventions: TSD experiment involving 33 h awake (N = 31 subjects) and a PSD experiment involving 5 nights of 4 h time in bed (N = 43 subjects).

Measurements and results: In a paired t-test paradigm and for both TSD and PSD, effect sizes of 10 different PVT performance outcomes were calculated. Effect sizes were high for both TSD (1.59-1.94) and PSD (0.88-1.21) for PVT metrics related to lapses and to measures of psychomotor speed, i.e., mean 1/RT (response time) and mean slowest 10% 1/RT. In contrast, PVT mean and median RT outcomes scored low to moderate effect sizes influenced by extreme values. Analyses facilitating only portions of the full 10-min PVT indicated that for some outcomes, high effect sizes could be achieved with PVT durations considerably shorter than 10 min, although metrics involving lapses seemed to profit from longer test durations in TSD.

Conclusions: Due to their superior conceptual and statistical properties and high sensitivity to sleep deprivation, metrics involving response speed and lapses should be considered primary outcomes for the 10-min PVT. In contrast, PVT mean and median metrics, which are among the most widely used outcomes, should be avoided as primary measures of alertness. Our analyses also suggest that some shorter-duration PVT versions may be sensitive to sleep loss, depending on the outcome variable selected, although this will need to be confirmed in comparative analyses of separate duration versions of the PVT. Using both sensitive PVT metrics and optimal test durations maximizes the sensitivity of the PVT to sleep loss and therefore potentially decreases the sample size needed to detect the same neurobehavioral deficit. We propose criteria to better standardize the 10-min PVT and facilitate between-study comparisons and meta-analyses.

Keywords: PVT; alertness; attention; lapse; power; psychomotor vigilance; response speed; response time; sensitivity; sleep deprivation.

Figures

Figure 1

The analyses shown in A, B, and C are based on the TSD study and were restricted to the first 1, 2, 3, 4, 5, 6, 7, 8, 9, and 10 min of the 10-min PVT (abscissa) (A) The number of lapses and their standard deviation are shown for the sleep deprived and the non–sleep deprived state. (B) The within-subject differences between sleep deprived and non–sleep deprived states of the number of lapses and their standard deviations are shown. (C) Effect sizes calculated as the within-subject differences between sleep deprived and non–sleep deprived states divided by their standard deviation are shown for lapses including 95% nonparametric bootstrap confidence intervals.

Figure 2

Effect sizes (ES) and 95% nonparametric bootstrap confidence intervals are shown for 10 outcome metrics of the PVT for both partial (open circles) and total (black squares) sleep deprivation depending on the analyzed portion of the 10-min PVT. Effect sizes of Mean 1/RT, Slowest 10% 1/RT, and the Performance Score were multiplied by −1 to facilitate comparisons.

Figure 3

Effect sizes (ES) and 95% nonparametric bootstrap confidence intervals are shown for each of the 10 outcome metrics of the 10-min PVT and for each of the 17 tests performed during a 33-h period of total sleep deprivation (TSD). The average of the first 7 tests performed between 1 to 13 hours awake (corresponding to 09:00 and 21:00) served as the non–sleep deprived state. Effect sizes of Mean 1/RT, Slowest 10% 1/RT, and the Performance Score were multiplied by −1 to facilitate comparisons.

Figure 4

Effect sizes (ES) and 95% nonparametric bootstrap confidence intervals are shown for each of the 10 outcome metrics of the 10-min PVT and for each of the 7 conditions of the partial sleep deprivation (PSD) study (BL = baseline, R = restriction). Baseline day 2 (BL2) served as the non–sleep deprived state. Effect sizes of Mean 1/RT, Slowest 10% 1/RT, and the Performance Score were multiplied by −1 to facilitate comparisons.