Analysis using SAS: AB/BA crossover trials

Mean Response values

Period 1 Period 2 D11.12.21.22 D11.12.22.21
Group 1 (AB) Y 11 Y 12 Y 11 Y 12 Y 11 Y 12
Group 2 (BA) Y 21 Y 22 Y 21 Y 22 Y 22 Y 21

Endpoints Fundamentals of Clinical Trials Summer I 22 / 49

Analysis using SAS: AB/BA crossover trials

(Y11 Y12 )(Y21 Y22 ) (Y11 +Y22 )(Y12 +Y21 )

Treatment differences based on 2 = 2

Endpoints Fundamentals of Clinical Trials Summer I 23 / 49

Analysis using SAS: AB/BA crossover trials

(Y11 Y12 )(Y22 Y21 ) (Y11 +Y21 )(Y12 +Y22 )

Period differences based on 2 = 2

Endpoints Fundamentals of Clinical Trials Summer I 24 / 49

Analysis using SAS: AB/BA crossover trials

The data are split into two groups according to treatment sequence
and analyzed as a two-independent-sample design
The analysis of the treatment effect is based on the half
D11.12.21.22 differences
The analysis for the period effect is based on the half D11.12.22.21
differences

Endpoints Fundamentals of Clinical Trials Summer I 25 / 49

Analysis using SAS: AB/BA crossover trials

Data: study comparing the effectiveness of two bronchodilators,

formoterol (for) and salbutamol (sal), in the treatment of
childhood asthma (Senn 2002, Chapter 3).
13 children are recruited for an AB/BA crossover design; 7 of the
children are assigned to the treatment sequence for/sal, receiving a
dose of formoterol upon an intial visit (period 1) and then a dose of
salbutamol upon a later visit (period 2); 6 children are assigned to
the sequence sal/for
Periods 1 and 2 are sufficiently spaced so that no carryover effects are
suspected
After a child inhales a dose of a bronchodilator, peak expiratory flow
(PEF) is measured. Higher PEF indicates greater effectiveness. The
data are assumed to be approximately normally distributed

Senn S. Crossover trials in clinical research, 2nd Edition.

SAS/STAT(R) 9.2 Users Guide, Second Edition
Endpoints Fundamentals of Clinical Trials Summer I 26 / 49
Analysis using SAS: AB/BA crossover trials

Endpoints Fundamentals of Clinical Trials Summer I 27 / 49

Analysis using SAS: AB/BA crossover trials

PEF1 = the responses for the first period

PEF2 = the responses for the second period
Drug1 = drug administered in the first period
Drug2 = drug administered in the second period

Endpoints Fundamentals of Clinical Trials Summer I 28 / 49

Analysis using SAS: AB/BA crossover trials

If the IGNOREPERIOD option is specified in the var statement, then the

treatment effect is analyzed as a paired analysis on the (treatment A,
treatment B) response value pairs, regardless of treatment sequence

Endpoints Fundamentals of Clinical Trials Summer I 29 / 49

Analysis using SAS: AB/BA crossover trials

Endpoints Fundamentals of Clinical Trials Summer I 30 / 49

Analysis using SAS: AB/BA crossover trials

Endpoints Fundamentals of Clinical Trials Summer I 31 / 49

Calculations

SAS output Treatment Diff(1-2):

30.7143 + 62.5
= 46.6071
2
SAS Output Period Diff(1-2):

30.7143 62.5
= 15.89285
2

Endpoints Fundamentals of Clinical Trials Summer I 32 / 49

Analysis using SAS: AB/BA crossover trials

We fail to reject the null hypothesis of equal variances (p=0.48)

Endpoints Fundamentals of Clinical Trials Summer I 33 / 49

Analysis using SAS: AB/BA crossover trials

Endpoints Fundamentals of Clinical Trials Summer I 34 / 49

Analysis using SAS: AB/BA crossover trials

Endpoints Fundamentals of Clinical Trials Summer I 35 / 49

Analysis using SAS: AB/BA crossover trials

Endpoints Fundamentals of Clinical Trials Summer I 36 / 49

Analysis using SAS: AB/BA crossover trials

Endpoints Fundamentals of Clinical Trials Summer I 37 / 49

Analysis using SAS: AB/BA crossover trials

The treatment differences are significant(p=0.0012) and the period

differences are not significant (p=0.1683)

Endpoints Fundamentals of Clinical Trials Summer I 38 / 49

Analysis using SAS: AB/BA crossover trials

If the assumption of normal distribution of the error terms is not met,

non-parametric analyses can be used in the analysis. The
NONPAR1WAY procedure in SAS can be used for Wilcoxon rank test
Analysis of binary data for AB/BA design:
The data need to be pre-processed into one record per patient
structure, with variables indicating the responses (0 or 1) at the end of
each treatment period.
Use McNemar test ( PROC FREQ procedure with the AGREE option
in the table statement)
More complex analyses: Random effects models (Model Fitting with
the MIXED Procedure in SAS)

Endpoints Fundamentals of Clinical Trials Summer I 39 / 49

Crossover trials: Advantages

Each patient receives both treatments, therefore variability is reduced

and the trial needs fewer subjects compared to a parallel design
The influence of confounding covariates is reduced because each
patient serves as her/his own control

Endpoints Fundamentals of Clinical Trials Summer I 40 / 49

Crossover trials: Disadvantages

Carryover effects (effect of the treatment from the previous time

period on the response at the current time period) may be
confounded with direct treatment effects; possible solution:
administer lengthy wash-out periods if appropriate; the length of the
washout period usually is determined using clinical expertise
There are risks to washing out medications in a patient, including
return or rebound of their medical/psychiatric problems
The disease of interest should be chronic and stable, and the
treatments should not result in total cures but only alleviate the
disease condition
Dropouts have strong effects since the trial duration is longer
The side effects may increase since patients are treated with more
drugs
Analysis are more complex than parallel group design, especially if
carry-over effects are considered
Endpoints Fundamentals of Clinical Trials Summer I 41 / 49