Petros et al. (2003) studied the impact of different levels of alcohol on various flight performance measurements. This article provides a meta-analysis using the results of the original study.

The study used simulation equipment to measure the number of errors made by the pilot sample across 7 different performance measures. These were measured during two experimental flights. Both flights involved a takeoff with air traffic control (ATC) clearance, 16 x 3 minute legs with parameters given by ATC, intercepting a localiser, and initiating a descent. Three groups were measured, a placebo group who had drank lemonade 11 hours before, a 2ml/kg group who had reached this alcohol level 11 hours before, and a 3ml/kg group who had also reached this alcohol level 11 hours before. This article will summarise each parameter measured in the study using effect sizes to make sense of differences. The effect size used is Cohen's d.
* Note: The performance measure 'procedures correctly executed' was deliberately omitted from this meta-analysis as it was thought to only serve as an overall comparison (already provided here) and was not seen as unique to the other performance measures.

Results

Altitude

Altitude was the first performance measure which was studied. The number of errors was measured using 1 error added per foot off optimal altitude per second.

Illustration 1: Postintoxication effects of alcohol on altitude
	First flight pattern		Second flight pattern
	Mean difference	Effect size (d)	Mean difference	Effect size (d)
Placebo vs. 2ml/kg	-51	-0.016	825	0.190
2ml/kg vs. 3ml/kg	-556	-0.178	-3061	-0.704
3ml/kg vs. Placebo	607	0.194	2236	0.514
[Effect size (d) based on the average standard deviation (= 3130 for pattern 1, = 4349 for pattern 2)]

We can see in illustration 1 that there were only very small effect sizes for errors committed in altitude for the first flight pattern. The second flight pattern shows inconsistencies and so no conclusions can be drawn. In general it would seem that there is very little or no effect of postintoxication on a pilot's ability to maintain correct altitudes.

Bank

Bank was the second performance measure studied. The number of errors was measured using 1 error added per degree off optimal bank per second.

Illustration 2: Postintoxication effects of alcohol on bank
	First flight pattern		Second flight pattern
	Mean difference	Effect size (d)	Mean difference	Effect size (d)
Placebo vs. 2ml/kg	559	0.226	904	0.337
2ml/kg vs. 3ml/kg	-1677	-0.679	-3158	-1.177
3ml/kg vs. Placebo	607	0.453	2236	0.840
[Effect size (d) based on the average standard deviation (= 2469 for pattern 1, = 2682 for pattern 2)]

Illustration 2 shows that in both flight patterns there was a very small to small effect size between placebo and 2ml/kg groups, but that there was a medium to large effect size between the 2ml/kg and 3ml/kg groups. Although not as large, there was still a moderate difference between the placebo and the 3ml/kg group. In both flight patterns the 2ml/kg group outperformed the other two groups. To work out why would require further research, but the results still suggest that may be a postintoxication effect of alcohol on bank.

Heading

The third performance measure studied was ability to maintain a heading. The error levels were measured using 1 error added per degree off instructed heading per second.

Illustration 3: Postintoxication effects of alcohol on heading
	First flight pattern		Second flight pattern
	Mean difference	Effect size (d)	Mean difference	Effect size (d)
Placebo vs. 2ml/kg	-10	-0.002	940	0.248
2ml/kg vs. 3ml/kg	-941	-0.202	-868	-0.229
3ml/kg vs. Placebo	951	0.204	-72	-0.019
[Effect size (d) based on the average standard deviation (= 4662 for pattern 1, = 3787 for pattern 2)]

Illustration 3 shows that all the effect sizes were very small. This suggests that there was no observable effect of alcohol postintoxication on a pilot's ability to maintain a correct heading.

Airspeed

Ability to maintain an optimal airspeed was the fourth performance measure studied.

Illustration 4: Postintoxication effects of alcohol on airspeed
	First flight pattern		Second flight pattern
	Mean difference	Effect size (d)	Mean difference	Effect size (d)
Placebo vs. 2ml/kg	-1551	-0.445	-730	-0.252
2ml/kg vs. 3ml/kg	-449	-0.129	-311	-0.107
3ml/kg vs. Placebo	2000	6.574	1606	0.554
[Effect size (d) based on the average standard deviation (= 3484 for pattern 1, = 2900 for pattern 2)]

Illustration 4 shows that there was little observable difference between placebo and 2ml/kg, as well as 2ml/kg and 3ml/kg. However, the first flight pattern showed a exceptionally large effect size between the 3ml/kg and placebo groups. In the second flight pattern this was only a small to moderate effect size. However, both consistently show that there was an observable difference between the placebo group's and the 3ml/kg group's abilities to maintain optimal airspeed.

Rate of turn

Rate of turn was the fifth performance measure to be studied. Error was calculated using 1 error added per degree off optimum per second.

Illustration 5: Postintoxication effects of alcohol on rate of turn
	First flight pattern		Second flight pattern
	Mean difference	Effect size (d)	Mean difference	Effect size (d)
Placebo vs. 2ml/kg	49	0.191	179	0.628
2ml/kg vs. 3ml/kg	-151	-0.590	-360	-1.263
3ml/kg vs. Placebo	102	0.398	181	0.635
[Effect size (d) based on the average standard deviation (= 256 for pattern 1, = 285 for pattern 2)]

Illustration 5 shows that the first and second flight patterns have inconsistent results. No conclusions can be drawn.

Recall of clearnances

Both flight patterns involved 8 radio calls, with 2 being recalled by the participants. One of these involved a low memory load, the other a high memory load. These were worked out as a percentage

Illustration 6: Postintoxication effects of alcohol on low memory load recall clearances
	First flight pattern		Second flight pattern
	Mean difference	Effect size (d)	Mean difference	Effect size (d)
Placebo vs. 2ml/kg	-1.65	-0.179	-0.77	-0.094
2ml/kg vs. 3ml/kg	3.50	0.380	4.98	0.612
3ml/kg vs. Placebo	-1.85	-0.201	-4.21	-0.518
[Effect size (d) based on the average standard deviation (= 9.21 for pattern 1, = 8.13 for pattern 2)]

Illustration 6 shows the inconsistencies of the data, suggesting that there was no observable effect of postintoxication on low memory load recalls.

Illustration 7: Postintoxication effects of alcohol on high memory load recall clearances
	First flight pattern		Second flight pattern
	Mean difference	Effect size (d)	Mean difference	Effect size (d)
Placebo vs. 2ml/kg	5.87	0.568	3.04	0.362
2ml/kg vs. 3ml/kg	3.73	0.361	1.15	=0.122
3ml/kg vs. Placebo	-9.60	-0.929	-4.19	-0.499
[Effect size (d) based on the average standard deviation (= 10.33 for pattern 1, = 8.39 for pattern 2)]

Illustration 7 shows that there was a small observable effect size between placebo and 2ml/kg groups. The effect size was much smaller between 2ml/kg and 3ml/kg groups, but in both flight patterns there was a moderate to large effect size between 3ml/kg and placebo groups. This suggests that there was a reasonably observable postintoxication effect of alcohol on high-memory load recalls.

Conclusion

This meta-analysis has provided a comparison of the effect of alcohol postinotxication on different flgiht performance measures by using data from a study by Petros et al. (2003). This showed that there may be postintoxication effects on a pilot's ability to maintain optimal bank and airspeed and also on a pilot's ability to correctly recall high memory load clearances.

Methods

Research approach

Experimental research into alcohol postintoxication effects on flight performance.

Sample

36 participants, all male, with at least 200 hours flight experience and 50 hours IFR. Aged between 21 and 35, all social drinkers and in good health according to FAA medical tests. Participants were paid $100 each for taking part in the study.

Design

Quantitative research, using both between and within subject experimental design.

Variables

The independent variable in this experiment was the alcohol intoxication level achieved by participants 11 hours before the experimental flights. The dependent variables were the pilots' abilities to maintain optimal altitude, bank, heading, airspeed, rate of turn and abilities to recall clearances of different memory loads.

Procedure

Participants were assessed on their mental rotation ability (Vadenberg's test), visual ability (Block design and digit symbol tests), working memory (forward and backward digit span) and vocabularies (WAIS-R subtest) prior to the experiment. Explanations of these tests can be found in the original article.

Participants, then, underwent a practice flight in a simulator for an hour. This was to ensure that there was no difference in flying abilities. The practice included 2 flight patterns which will be used in the experiment.

Participants were then grouped into either "Placebo", "2ml/kg", or "3ml/kg" which correspond to the treatments that they would receive. Participants were then given their respective drinks which were:
Placebo: lemonade with 1ml of vodka wiped around the rim of the cup
2ml/kg: lemonade and vodka mixture (dependent on the weight of the participant to give a final blood alcohol level of 2ml/kg)
3ml/kg: lemonade and vodka mixture (dependent on the weight of the participant to give a final blood alcohol level of 3ml/kg)

An alcohol breath test was then carried out every 15 minutes for 90 minutes after consuming the beverages. Participants were then sent to bed. The following morning, participants were all woken up at similar time and provided breakfast. Participants, then, took the balance test (Sharpened Romberg test) followed by the flight test.
The flight tests for each participant comprises of two flight patterns. Each flight began with the ATC providing take off clearance. Participants were told to climb out at a certain speed, heading and climb to a certain altitude. Three minutes after the take off, participants were given the first ATC instructions. This comprised of a new heading, altitude and radio contact frequency. Participants were required to read back the ATC command and execute accordingly. Three minutes after the first instruction, a new set of instructions were given and must be followed accordingly. This was repeated 16 times. Of the 8 radio calls made during the flight, a high load and a low load memory calls were to be recalled. Participants were, finally, given their last heading, altitude and were told to contact approach once they had reached the localiser. They would then descend and the simulation would end. This was repeated for the second flight which had different instructions but in the same format.

Data analysis

Descriptive statistics: Mean differences and effect sizes.

Generalisation potential

The study has a low generalisability due to not having a large enough sample size to properly reflect the characteristics of each the treatment groups. The sample was also limited to pilots from North Dakota so the results may be geographically limited. The inferential statistics used in the original study may not have been appropriate because it appears that there was not homogeneity of variance. The tests used were a mixed ANOVA and a one-way ANOVA which assume homogeneity of variance. This is why this meta-analysis has focused on describing the sample used without inferring results onto the population. Generalising results should be done with caution.

References

1: Petros T., Bridewell J., Jensen W., Ric Ferraro F., Bates J., Moulton P., Turnwell S., Rawley D., Howe T. & D. Gorder. (2003). Postintoxication effects of alcohol on flight performance after moderate and high blood alcohol levels. The International Journal of Aviation Psychology, 2003, volume 13, number 3, pages 287-300.

Want to know more?

Explanation of effect size: To understand this meta-analysis then a sound knowledge of effect size is required. This presentation is very suitable for someone with a basic knowledge of statistics.