交叉实验

A crossover experiment, also called a repeated measurements design, exposes all experimental units to every treatment across different time periods. Unlike simple randomized trials, subjects act as their own controls, eliminating inter-subject variability. This approach is particularly valuable in clinical trials where patients receive all treatments sequentially, with washout periods between exposures to clear previous treatments from the body.

A washout period eliminates residual effects of the previous treatment from the subject's body before introducing the next treatment. This prevents carryover effects that could confound results and compromise data validity. Without adequate washout time, the lingering effects of one drug could artificially influence responses to the subsequent treatment, making it impossible to isolate each treatment's true effect.

Crossover designs are ideal for late-phase clinical trials testing drugs that control symptoms rather than cure diseases completely. They work well for comparing bioavailability or drug ingestion in the human body against reference drugs. However, crossover designs are unsuitable for chronic, stable diseases or when drugs provide complete cures, as treatment effects would prevent meaningful comparison of subsequent treatments.

In crossover designs, each subject receives all treatments sequentially, allowing researchers to compare treatment effects within the same individual. Since subjects' characteristics remain constant throughout the study, differences in outcomes directly reflect treatment effects rather than individual variation. This internal comparison eliminates confounding variables associated with inter-subject differences, making crossover designs more statistically powerful than between-subject comparisons.

Crossover designs require smaller sample sizes than simple randomized trials because each subject provides data for all treatment conditions. Since subjects act as their own controls, researchers can detect treatment differences more efficiently with fewer participants. This efficiency makes crossover designs cost-effective and practical in fields like psychology, pharmaceuticals, agriculture, and medicine where sample recruitment may be limited.

In a typical example, asthmatic patients are randomly divided into two groups. Group one receives drug A for two weeks while group two receives drug B, with effects recorded. After a washout period, the groups switch treatments. This design allows researchers to compare both drugs' effectiveness within the same patients, ensuring fair comparison while controlling for individual patient characteristics that might influence drug response.

Crossover designs fail for chronic, stable diseases because if the first treatment cures or substantially improves the condition, the disease may not return during the washout period. This prevents meaningful evaluation of subsequent treatments. For example, if drug A cures asthma in the first period, drug B cannot demonstrate its effectiveness in the second period, making treatment comparison impossible and invalidating the study's conclusions.