Evaluation of measures of impairment in real and simulated driving: Results from a randomized, placebo-controlled study

<p><b>Objective</b>: Standard deviation of lateral position (SDLP) is often the primary outcome in experimental studies on impaired driving. However, other measures may be easier and more practical to obtain and reflect a broader range of driving-related behaviors. We wanted to assess the validity and sensitivity of a range of measures in a driving simulator as well as during real driving and compare these to SDLP.</p> <p><b>Methods</b>: Twenty healthy male volunteers undertook 6 driving trials each, 3 in a regular car on a closed track resembling rural road conditions and 3 in a simulator with an identical driving scenario. Ethanol was used as impairing substance due to its well-characterized effects on driving. The subjects were tested sober and at blood alcohol concentrations (BAC) of approximately 0.5 and 0.9 g/L. We explored dose–response relationships between BAC and a range of driving-related measures, as well as their BAC-dependent effect sizes.</p> <p><b>Results</b>: In simulator driving, ethanol intake increased steering wheel reversal frequency, steering wheel movement measures, average speed, standard deviation of speed, and pedal use frequency. At the test track, only steering wheel movement and standard deviation of speed were significantly correlated to BAC. Likewise, reaction to unexpected incidents and observance of red traffic lights were adversely affected by ethanol in the simulator but not at the test track. Whereas SDLP showed a relatively large effect size that was similar in simulated and real driving, all other measures demonstrated smaller effect sizes, with less pronounced BAC effects on the test track than in the simulator.</p> <p><b>Conclusions</b>: The results suggest that the driving-related measures explored in this study are less sensitive to alcohol-mediated driving impairment than SDLP, especially during real (test track) driving. The discrepancy in effect sizes between simulated and real driving may imply low external validity of these measures in simulator studies.