Crossover Trial for a
Cycling Performance Study

RESEARCH QUESTION

What the Study Set Out to Ask

This crossover trial example follows an exercise physiology laboratory that wants to know whether a nitrate-rich drink, compared with a placebo drink, changes performance in a 4-km cycling time trial among trained cyclists. Each cyclist takes both drinks, so each serves as their own comparison and a small group can give a precise answer.

The question is the within-cyclist difference in mean power between drink and placebo, and whether it exceeds the smallest change that matters in competition. A crossover design raises problems that a parallel trial does not: the order, the period or a lingering effect of the first drink can distort the comparison. The plan states how each is handled before the first visit, and the difference is read against a threshold set in advance.

STUDY DESIGN AND DATA

Two Periods for About 24 Cyclists

A randomized, double-blind crossover study of about 24 trained cyclists, each assigned to drink then placebo or to placebo then drink, with a familiarization trial before the first experimental visit and a washout of at least one week between periods, justified from the time course of the supplement. The dataset holds the randomization sequence, period, drink, mean power in each time trial, and baseline fitness and body mass. A cyclist may miss the second period, and the analysis keeps every trial that was completed.

METHOD AND ASSUMPTIONS

Why Sequence and Period Are in the Model

The two measurements from one cyclist are correlated, so they are analyzed in a mixed model with drink, period and sequence as fixed effects and a random intercept for cyclist. When every cyclist completes both periods, this gives the same estimate as the classic crossover analysis, which compares the within-cyclist differences between the two sequences so that the period effect cancels out; a plain paired comparison would ignore period. The model also uses what a cyclist with one completed period contributes, under the assumption that a missed visit is unrelated to the outcome given the model. The familiarization data are set aside from the estimate because learning is largest on first exposure.

Carryover is handled by the design, through the washout and the balanced sequences, and not by a preliminary test for carryover, which in a two-period design is the between-cyclist test of sequence: it has low power, and choosing the analysis from its result biases the estimate that follows. The plan records the evidence for the washout length and shows the means by sequence and period as descriptive information. If carryover were suspected, the first period alone would give an estimate free of it, with less precision. Normality tests do not choose the method in a sample this small: residual plots are examined, and a sensitivity analysis checks whether any single cyclist drives the result.

OUTPUT AND INTERPRETATION

How the Results Are Read

The output has the mean within-cyclist difference in power with its 95% confidence interval, a plot of each cyclist's two values joined by a line, and the means by sequence and period. The difference is placed against the smallest worthwhile change, which the plan sets from the typical error of the time trial and the performance literature, and the interval is read against that threshold: an interval inside it, across it or beyond it supports different statements.

Sensitivity analyses refit the model on the cyclists who completed both periods, on the first period alone and on log-transformed power. The interpretation describes the effect of this drink in these cyclists under these laboratory conditions, with its uncertainty, and does not extend to other doses, events or populations. Whether the effect is large enough to matter in training or competition is a decision for the laboratory and the athletes' coaches.

REPORTING

What the Researcher Received

A written report structured to the CONSORT extension for crossover trials, with participant flow by sequence and period and a baseline table by sequence, and statistical wording for the methods, matched to the output, for the team to check and incorporate. The research question, physiological interpretation and conclusions remain with the research team.