Participants
Thirteen healthy senior individuals (mean(SD) age: 61.5(4.4) years, 9 male) volunteered to participate in this study. None of the participants was, or used to be a habitual drinker. Inclusion criteria were absence of any known serious neurological, orthopaedic or cognitive dysfunction, and age between 50-70 years. Exclusion criteria were a bodyweight exceeding 100 kg or the use of (prescribed) medication(s) that could interfere with alcohol. As the experiment took place in the late afternoon, participants were instructed to just have an early light lunch (e.g. a sandwich), and not to drink caffeinated drinks in the 4 hours before arriving at the laboratory. Subjects were informed about the experimental procedure before they gave their written informed consent in accordance with the ethical standards of the Declaration of Helsinki. The protocol was approved by the ethical committee of the region Arnhem-Nijmegen.
Equipment and procedure
The participants were instructed to avoid obstacles while walking on a treadmill (ENRAF Nonius, Type EN-tred Reha) at a fixed velocity of 3 km/hr (Figure 1A), wearing their own comfortable shoes (no high heels). A wooden obstacle (measuring 40x30x1.5cm) with an embedded piece of iron was held by an electromagnet just above the treadmill surface. Its release could be triggered by the computer. The obstacle was always presented to the left foot. On both feet, three reflective markers (diameter 14 mm) were attached at heel, hallux and lateral malleolus. A single marker was placed on top of the obstacle. Marker positions were recorded by an 8-camera 3-D motion analysis system (Vicon®, Oxford Metrics, London, UK) at a sample rate of 100 Hz. The marker positions were processed in real time in order to determine the moment of obstacle release related to gait phase. The real time processing also enabled the experimenter to check online the foot position with respect to the obstacle, while the participants were instructed to walk at a fixed distance to the obstacle that was approximately 10 cm from the most anterior position reached by the toes in the swing phase. If they deviated more than 3 cm from this position, participants received verbal feedback to correct the distance to the obstacle. The obstacle was not released until a regular walking pattern was observed and until at least five unperturbed strides for the trial had been completed. Stride regularity was defined as a maximum difference of 50 ms between two consecutive strides. The obstacle was dropped at one of three different phases of the step cycle (late stance (LSt,45-60% of the step cycle), early swing (ESw, 60-70%) or mid swing (MSw,70-85%)) to create different levels of difficulty to avoid the obstacle as time pressure increased (Figure 1B). Available response time (ART), the measure of time pressure, was defined as the time between obstacle release and the estimated moment of foot contact with the obstacle if no adjustment of the stride had been made [12]. The obstacle release phases corresponded with ARTs greater than 450 ms (LSt), 300-450 ms (ESw), and 150-300 ms (MSw). Ten obstacles in each of the three phases of the gait cycle were presented in random order during a series of 30 trials.
The participants were instructed to look at the obstacle, and step over it after its release. Stepping to the side was discouraged, and any contact of the left foot with the obstacle was defined as a failure. Since the m. biceps femoris (BF) is known to be the prime mover involved in the avoidance reaction [6], surface electromyography (EMG) data were collected from this muscle to assess response times. Self-adhesive Ag-AgCl electrodes (Tyco Arbo ECG) were placed approximately 2 cm apart and longitudinally on the belly of the muscle, according to European guidelines [13]. The EMG signals were sampled at 2400 Hz (ZeroWire®, Aurion S.r.l., Italy) and recorded synchronously with the marker data.
Three series of 30 obstacle avoidance trials were performed, each 30 minutes after ingestion of a drink (Figure 1C). Subjects were informed that these drinks contained alcohol, and had to finish them within 10 minutes. The first (A0) was a placebo consisting of water mixed with orange juice (ratio 1:3) with a drop of vodka floated on top to give the scent of alcohol. The following two drinks (A1 and A2) each contained 40% vodka mixed with orange juice (ratio 1:3). We aimed to reach a BAC that was around the common legal limits for driving (0.05% for most European countries or 0.08% for most US states, Canada and UK) 30 minutes after A2, having used the Widmark formula [14] to adjust the alcohol dosage for the individual's gender and weight. A Dräger Alcotest® 7410 Plus com breathalyzer was used to determine the BAC before, during, and after the experimental task (Figure 1C). For safety reasons, all participants were taken home by a taxi after the experiment was finished.
Data analysis
Successful obstacle avoidance for each trial was scored. This was easily determined by two observers by eye, and by feedback from the participant. If there was any doubt about the successfulness, the marker data were checked (this happened in less than 1% of the cases). As the primary outcome measure, failure rates (as defined by the number of failed trials divided by the total number of trials) were calculated for each alcohol condition and each step cycle phase.
To assess the EMG responses, the EMG activity of the m. biceps femoris (BF) was full-wave rectified and low-pass filtered at 25 Hz (zero lag, 4th order Butterworth filter). Background EMG was calculated for each series separately as the average BF activity over 25 control strides (i.e. the stride preceding that in which obstacle release occurred). For each participant and alcohol condition, BF response times were determined as the time between obstacle release and the moment at which BF activity exceeded the average control stride by at least 2 SDs for more than 30 ms (for example, see Figure 2). This was done with the help of a custom made computer algorithm (Matlab® software, version 7.4.0, The Mathworks Inc., US). Each trial for which a response time was calculated was visually checked for correct determination of the response onset. In about 2% of the trials the onsets were corrected. The onsets of the avoidance responses for each subject were averaged for each phase of obstacle release within each alcohol condition. The responses amplitude was calculated as the average amplitude during the 100 ms following the onset of the BF response [5, 15]. The amplitudes were normalized with respect to the maximum average background activity during the whole step cycle in the A0 condition. A similar procedure was performed to calculate and normalize the average control stride activity in the 100 ms following the BF response onset.
Statistical analysis
To check whether within participants, the series were equally difficult in the three alcohol conditions, we compared the average ARTs in a repeated measures MANOVA (within-subjects factors: alcohol condition (A0, A1, A2); phase of obstacle release (LSt, ESw, MSw), α = 0.05) with post-hoc pairwise comparisons.
To identify the effect of BAC on avoidance failure rates we used a binary logistic regression with alcohol condition as categorical factor (A0 as reference category, α = 0.05) in Egret® for Windows (version 2.0.31). A statistical model was fitted to the data of the MSw phase to predict the probability of a failure with increasing BAC for the most time critical situations.
For the analysis of EMG data, we used repeated measures MANOVAs with post-hoc pairwise comparisons to test for differences between the three alcohol conditions (within subjects-factor: A0, A1, A2; α = 0.05) for average BF response times, normalized response amplitudes, and normalized background activity. The relationship between BF response time and BAC was assessed by means of bivariate correlation (Pearson Correlation Coefficient). One sample Students' t-test with bins of 0.005% BAC was used to determine the BAC from which the BF response time was significantly delayed. These analyses were carried out in SPSS® (version 12.0.1) with α set at 0.05. Means are presented with their standard errors (SE).
Pilot data indicated that the difference in obstacle avoidance response time after 2-3 standard alcoholic drinks was 20 ms (SD: 18 ms). To be able to identify a difference of 20 ms in the mean response time between A0 and A2, a sample size of 11 subjects would be needed in this repeated measures design (β = 0.9, α = 0.05).