Crossover effects in experimental investigations with human subjects

Pharmacology Biochemistry & Behavior, Vol. :3, pp. 1141--1145. Copyright © 1975 by ANKHO International Inc. All rights of reproduction in any form reserved. Printed in the U.S.A.

BRIEF COMMUNICATION Crossover Effects in Experimental Investigations with Human Subjects ALI A. L A N D A U E R

Department o f Psychology, University o f Western Australia, Nedlands 6009, Australia (Received 19 June 1975) LANDAUER, A. A. Crossover effects in experimental investigations with human subjects. PHARMAC. BIOCHEM. BEHAV. 3(6) 1141 1145, 1975. - Simple crossover studies with drugs may introduce artifactual results. Volunteers are frequently less apprehensive when tested on the second occasion and experimental medication may result in different performance measures. Empirical evidence is given to demonstrate this point and a suggestion is made as to how it can be overcome. Crossover effects

Experimental design

Human drug research

THE use of crossover designs in pharmacological research seems to have gained a quite unwarranted status. For instance, in a recent review of the effects of antianxiety drugs on human performance McNair [2] analyzed the methods used in the numerous investigations which he reports and found that in 68 percent o f the studies a crossover design was used. He also reports that 64 percent of the investigations were performed under double-blind conditions, that the vast majority of experiments were made with normal males (88 percent), and that active medication consisted of a single dose (70 percent). Thus, the majority of pharmacological experiments with antianxiety drugs which test behavioral effects and which find their way into the scientific literature employ healthy male volunteers who receive a single dose of the active medication and a single dose of placebo (or of a comparison drug) under double-blind conditions. The simplest type of crossover design in drug research with human subjects is to give Drug A and Drug B consecutively and determine the order of administration by chance. In place of one of the drugs, placebo can be given. Therefore, in this type of design there are two groups of subjects: one group which starts on Drug A and receives Drug B afterwards; and one group which is given Drug B first, followed by Drug A. It is better to have an equal number of subjects in each group, though this is rarely achieved in practice owing to extra-experimental causes. However, there is nothing wrong with having an unequal number of subjects in the two groups, provided the appropriate statistics are applied to the data. It has been known for a long time that special care should be used in crossover studies in which a transfer of effects may take place. Poulton and Freeman [4] examined a large number of crossover studies in which transfer effects 1141

Method

Transfer effects

had been found. Though most of the experiments they review are taken from the field of human performance, the paradigm they advance holds equally well to other domains. According to these investigators, transfer can be symmetrical and asymmetrical. By starting conditions at random in a fully-balanced Latin Square design, symmetrical transfer effects are adequately controlled. A typical asymmetrical transfer effect occurs when performance is lowered under stress conditions during the first test, and only marginally improves when the second test is given under normal conditions; while subjects who are first tested under non-stress conditions perform at a higher level, which is only slightly debased in the subsequent test under stress conditions. If the experimenter suspects that there may be some asymmetrical transfer effects which occur in a crossover design with drugs, he can compare the effects of Drug A administered to Group 1 (Drug A first group), with the effects of Drug B administered to Group 2 (Drug B first group). For the time being he disregards the effect of the second drug which was administered to each group. Unfortunately this method is not very sensitive to small differences between treatment conditions, in particular if individual differences are large. Such a design becomes a two independent groups model with all the shortcomings the investigator was trying to avoid when he decided to use a crossover design. One factor which may account for asymmetrical transfer effects with normal volunteers in crossover experiments with drugs may be the following. Most subjects are somewhat apprehensive and anxious when they arrive for the first time at a laboratory or hospital to take part in a drug experiment. This is particularly marked in young men who have no medical or paramedical background and who

1142

LANDAUER TABLE 1 MEAN PERFORMANCE, POOLED OVER TESTING DAYS

Effect Due to Alcohol Overall Effect

Without Alcohol

With Alcohol

284

231

Overall Effect

Effect Due to Drug Oxprenolol

280

321

238

Placebo

235

246

224

TABLE 2 ANALYSIS OF VARIANCE, POOLED RESULTS

Source of Variance

SS

MS

df

Between Subjects

308104

19

Within Subjects

897500

60

Drug Effect

39650

39650

1

594912

31311

19

Alcohol Effect

55073

55073

I

Error (Alcohol Effect)

46569

2451

19

Drug X Alcohol Effect

18514

18514

I

142782

7515

19

Error (Drug Effect)

Error (Interaction)

Grand Total

1205604

F-Ratio

1.266

22.469*

2.464

79

*p<0.001 have n o t p a r t i c i p a t e d in previous investigations. These subjects r e p r e s e n t t h e t y p e m o s t s o u g h t after in drug experiments. S u b s e q u e n t t o t h e i r first session the v o l u n t e e r s b e c o m e aware t h a t psychological and physiological tests are n o t u n p l e a s a n t a n d t h a t even b l o o d - l e t t i n g b y e x p e r i e n c e d s t a f f is only mildly d i s c o m f o r t i n g . C o n s e q u e n t l y , w h e n subjects r e t u r n for t h e i r s e c o n d trial, t h e y are no longer as a n x i o u s as t h e y were on the first testing day. T h e r e f o r e , if an a n t i a n x i e t y drug such as a tranquillizer is being evaluated, the drug will t e n d to affect p e r f o r m a n c e of subjects on the first testing day only. Some c o n f i r m a t i o n for this p o i n t of view c o m e s f r o m a recent crossover s t u d y in w h i c h the Profile of M o o d States q u e s t i o n n a i r e [3] was a d m i n i s t e r e d to 20 y o u n g men. No significant difference was f o u n d b e t w e e n t h e active drug (a

b e t a - a d r e n e r g i c blocking agent) and placebo, t l o w e v e r , w h e n t h e data were a n a l y z e d according to testing days, it was f o u n d t h a t there was a significant d i f f e r e n c e on two d i m e n s i o n s w h i c h this q u e s t i o n n a i r e measures. On the first testing day subjects were significantly m o r e a n x i o u s (t = 3.053, df = 19, p < 0 . 0 1 ) and more depressed (t = 2.647, d f = 19, p < 0 . 0 2 ) t h a n on the s e c o n d testing day. The results d e n o t e t h a t d i f f e r e n t stages of a n x i e t y and depression existed w h e n t h e active drug was a d m i n i s t e r e d to subjects. The effect appears to be s y m m e t r i c a l and b y starting an equal n u m b e r of subjects in each c o n d i t i o n , p r o b a b l y is controlled adequately. As far as p e r f o r m a n c e measures are c o n c e r n e d , similar effects have been o b t a i n e d . In a r e c e n t s t u d y w h i c h investigated t h e b e h a v i o u r a l effects of o x p r e n o l o l and alcohol on skills similar to t h o s e used in car driving [ I ] a

C R O S S O V E R E F F E C T S IN E X P E R I M E N T S W I T H H U M A N S

1143

TABLE 3 MEAN PERFORMANCE ON EACH TESTING SESSION, DATA OF GROUP 2 IN ITALICS. THE MEAN OF THE TWO ENTRIES APPEARS IN TABLE 1 . G R O U P 1 : DRUG--PLACEBO. GROUP 2: PLACEBO-DRUG.

Effect Due to Alcohol Overall Effect

Without Alcohol

With Alcohol

249

213

319

250

178

178

177

382

464

300

284

319

248

187

174

200

Overall Effect

Effect Due to Drug Oxprenolol

Placebo

simple crossover design was used. On each o f the t w o testing days ( o n c e w i t h m e d i c a t i o n and o n c e w i t h p l a c e b o ) subjects were tested twice: first w i t h o u t a n d l a t e r w i t h 1 m l / k g o f e t h a n o l . This c o n f o u n d i n g o r d e r effect has always b e e n used in o u r drug-alcohol research. In o n e t r e a t m e n t of the data p e r f o r m a n c e over testing days was p o o l e d . This could be d o n e since half the s u b j e c t s s t a r t e d o n p l a c e b o and half o n the active drug. T h e results o f o n e test, the driving score o n t h e M a r t i n S i m u l a t o r , are s h o w n in T a b l e 1. A s u m m a r y of the analysis of variance is given in Table 2. As can be seen the o n l y d i f f e r e n c e w h i c h was significant was due to a l c o h o l c o n s u m p t i o n . A n o t h e r way o f e x a m i n g the results of this e x p e r i m e n t is to consider t h a t two g r o u p s o f subjects were used: G r o u p 1 , which was given o x p r e n o l o l o n the first e x p e r i m e n t a l day and p l a c e b o on t h e s e c o n d day, and G r o u p 2, w h i c h received p l a c e b o o n the first day a n d o x p r e n o l o l o n the second. Table 3 s h o w s the m e a n for each day u n d e r t h e same c o n d i t i o n s as Table 1 : e a c h of the t w o m e a n s in T a b l e 3 has a single m e a n in T a b l e 1 • A n e w analysis of variance was m a d e of these data. T h e m o d e l used was the T h r e e - f a c t o r E x p e r i m e n t with R e p e a t ed Measures (Case 1) m o d e l w h i c h has been fully described by Winer [ 5 ] . It has one b e t w e e n subjects variable ( G r o u p 1 c o m p a r e d w i t h G r o u p 2) a n d two w i t h i n subjects variables, w h i c h were i d e n t i c a l to t h o s e used in t h e o t h e r analysis. In o t h e r words, the effects o f drug and a l c o h o l were again e x a m i n e d a f t e r any effect due to t r e a t m e n t o r d e r h a d been isolated. T h e s u m m a r y o f the analysis o f variance is p r e s e n t e d in Table 4. The drug a n d the a l c o h o l c o n d i t i o n s each had a significant main effect. F u r t h e r m o r e , the i n t e r a c t i o n s of drug w i t h g r o u p , drug w i t h alcohol, a n d t h e triple i n t e r a c t i o n of drug, alcohol a n d g r o u p were all significant. This s e c o n d - o r d e r i n t e r a c t i o n is p l o t t e d in Fig. 1. It can be seen t h a t on t h e first testing session p e r f o r m a n c e b o t h with and w i t h o u t o x p r e n o l o l was very similar. Moreover, on

~0.

GROUP I~

DRUG'ePLACEBO •

GROUP 2:

PLACEBO.PDRUG DRUG -

450,

PLACEBO

400. W O ¢O 3,~O O) ~ 300> ¢-~ ~: ZSOZ la.I 200-

150-

// WITHOUT ALCOHOl.

WITH ALCOHOL

FIG. I. Mean driving score. Each point is based on l 0 observations. The higher the score, the better the performance.

1144

LANDAUER TABLE 4 ANALYSIS OF VARIANCE, SEPARATE GROUPS

Source of Variance

MS

df

F-Ratio

57943

57943

l

4.169

250161

13898

18

SS

Between Subjects Between Groups Error (Between Groups)

Total Between Subjects

308104

19

Within Subjects

Drug Effect

39650

39650

1

5.045*

Drug X Group Effect

453457

453457

l

57.702+

Error (Drug Effect)

141455

7859

18

55073

55073

1

24.115~-

5462

5462

1

2.391

Error (Alcohol Fffect)

41107

2284

18

Drug × Alcohol Effect

185 l 4

18514

1

5.711 *

Drug × Alcohol × Group Effect

84435

84435

l

26.048+

Error (Drug × Alcohol Effect)

58347

3242

18

Alcohol Effect Alcohol X Group Effect

Total Within Subjects

Grand Total *p<0.05

897500

60

1205604

79

tp<0.001

this o c c a s i o n the w e l l - d o c u m e n t e d d e t r i m e n t a l effects of alcohol i n t a k e were a p p a r e n t l y a t t e n u a t e d b y the effect of practice in b o t h t h e drug and t h e placebo group. O n the second testing day, w h e n subjects were m o r e familiar w i t h e q u i p m e n t and p r o c e d u r e , t h e r e was a general i m p r o v e m e n t in p e r f o r m a n c e . However, t h e s u b j e c t s w h o received o x p r e n o l o l on this occasion achieved a b e t t e r score t h a n those w h o received placebo. With s u b s e q u e n t alcohol c o n s u m p t i o n the p e r f o r m a n c e decline was q u i t e m a r k e d in b o t h groups of subjects. This p r o b a b l y a c c o u n t s for t h e significant drug-group, drug-alcohol and drug-alcohol-group interactions. It m a y be p r u d e n t to abstain f r o m crossover designs in p h a r m a c o l o g i c a l research w i t h h u m a n subjects. This applies

in particular to investigations with drugs w h i c h have an a n t i a n x i e t y effect and where b e h a v i o r a l measures are used. In this case it can r e a s o n a b l y be p r e d i c t e d t h a t a smaller a n t i a n x i e t y effect s h o u l d o c c u r w i t h n o r m a l subjects d u r i n g the s e c o n d testing session, since m o s t test a n x i e t y s h o u l d have dissipated. One way to o v e r c o m e some possible p r o b l e m s caused by such a s y m m e t r i c a l t r a n s f e r effects w o u l d be to test subjects on 3 occasions. T h e first session w o u l d be o n placebo only (single-blind) a n d the s u b s e q u e n t sessions w o u l d be t h e usual crossover d o u b l e - b l i n d c o n d i t i o n s . This p r o c e d u r e , w h i c h is being used by some investigators, w o u l d t e n d to r e d u c e any a r t e f a c t u a l effects w h i c h are due to the test a n x i e t y of n o r m a l , naive volunteers.

C R O S S O V E R E F F E C T S IN E X P E R I M E N T S WITH H U M A N S

1145

REFERENCES l.

2.

Landauer, A. A., D. A. Pocock and F, W. Prott. The effect of oxprenolol and alcohol on skills similar to those used in car driving. Paper presented at the 2nd International Congress of C.I.A.N.S., Prague, 1975. McNair, D. M. Antianxiety drugs and human performance. Archs gen. Psychiat. Chicago, 29: 6 1 1 - 6 1 7 , 1971.

3. 4. 5.

McNair, D. M., M. Lorr and L. F. Droppleman. Profile of Mood States Manual. San Diego: Educational and Industrial Testing Service, 1971. Poulton, B. C. and P. R. Freeman. Unwanted asymmetrical transfer effects with balanced experimental designs. Psychol. Bull. 66: 1-8, 1966. Winer, B. J. Statistical Principles in Experimental Design. New York: McGraw-Hill, 1962.