Hubbry Logo
Crossover studyCrossover studyMain
Open search
Crossover study
Community hub
Crossover study
logo
7 pages, 0 posts
0 subscribers
Be the first to start a discussion here.
Be the first to start a discussion here.
Crossover study
Crossover study
Crossover study
View on Wikipedia
from Wikipedia

In medicine, a crossover study or crossover trial is a longitudinal study in which subjects receive a sequence of different treatments (or exposures). While crossover studies can be observational studies, many important crossover studies are controlled experiments, which are discussed in this article. Crossover designs are common for experiments in many scientific disciplines, for example psychology, pharmaceutical science, and medicine.

Randomized, controlled crossover experiments are especially important in health care. In a randomized clinical trial, the subjects are randomly assigned to different arms of the study which receive different treatments. When the trial has a repeated measures design, the same measures are collected multiple times for each subject. A crossover trial has a repeated measures design in which each patient is assigned to a sequence of two or more treatments, of which one may be a standard treatment or a placebo.

Nearly all crossover are designed to have "balance", whereby all subjects receive the same number of treatments and participate for the same number of periods. In most crossover trials each subject receives all treatments, in a random order.

Statisticians suggest that designs should have four periods, which is more efficient than the two-period design, even if the study must be truncated to three periods.[1][2] However, the two-period design is often taught in non-statistical textbooks, partly because of its simplicity.

Analysis

[edit]

The data is analyzed using the statistical method that was specified in the clinical trial protocol, which must have been approved by the appropriate institutional review boards and regulatory agencies before the trial can begin. Most clinical trials are analyzed using repeated-measurements ANOVA (analysis of variance) or mixed models that include random effects.

In most longitudinal studies of human subjects, patients may withdraw from the trial or become "lost to follow-up". There are statistical methods for dealing with such missing-data and "censoring" problems. An important method analyzes the data according to the principle of the intention to treat.

Advantages

[edit]

A crossover study has two advantages over both a parallel study and a non-crossover longitudinal study. First, the influence of confounding covariates is reduced because each crossover patient serves as their own control.[3] In a randomized non-crossover study it is often the case that different treatment-groups are found to be unbalanced on some covariates. In a controlled, randomized crossover designs, such imbalances are implausible (unless covariates were to change systematically during the study).

Second, optimal crossover designs are statistically efficient, and so require fewer subjects than do non-crossover designs (even other repeated measures designs).

Optimal crossover designs are discussed in the graduate textbook by Jones and Kenward and in the review article by Stufken. Crossover designs are discussed along with more general repeated-measurements designs in the graduate textbook by Vonesh and Chinchilli.

Limitations and disadvantages

[edit]

These studies are often done to improve the symptoms of patients with chronic conditions. For curative treatments or rapidly changing conditions, cross-over trials may be infeasible or unethical.

Crossover studies often have two problems:

First is the issue of "order" effects, because it is possible that the order in which treatments are administered may affect the outcome. An example might be a drug with many adverse effects given first, making patients taking a second, less harmful medicine, more sensitive to any adverse effect.

Second is the issue of "carry-over" between treatments, which confounds the estimates of the treatment effects. In practice, "carry-over" effects can be avoided with a sufficiently long "wash-out" period between treatments. However, planning for sufficiently long wash-out periods requires expert knowledge of the dynamics of the treatment, which is often unknown.

See also

[edit]

Notes

[edit]

References

[edit]
Revisions and contributorsEdit on WikipediaRead on Wikipedia

Crossover study

View on Grokipedia
from Grokipedia
A crossover study, also known as a crossover trial or crossover design, is a longitudinal experimental design commonly used in where each participant receives two or more interventions sequentially over multiple periods, with the order of interventions randomized across participants to control for sequence effects. In this approach, all participants are exposed to every treatment condition, enabling each individual to serve as their own control and facilitating within-subject comparisons that account for individual variability. This design contrasts with parallel-group trials, where different groups receive different treatments simultaneously, and is particularly suited to Phase I and II studies evaluating , , or short-term efficacy. The primary advantages of crossover studies stem from their efficiency in reducing the required sample size, as the paired nature of the data enhances statistical power for detecting treatment differences with fewer participants compared to parallel designs. By minimizing between-subject variability, such as genetic or environmental factors, crossover trials provide more precise estimates of treatment effects, making them ideal for investigating reversible conditions like stable chronic diseases (e.g., or mild ) where participants can return to baseline between periods. Additionally, randomization of treatment sequences helps mitigate biases from period or order effects, ensuring balanced comparisons across groups. Despite these benefits, crossover studies have notable limitations that must be carefully managed in their design and analysis. A key disadvantage is the risk of carryover effects, where residual impacts from an earlier treatment influence responses in subsequent periods, potentially results unless addressed with sufficient washout intervals to allow full recovery to baseline. They are unsuitable for acute conditions, curative treatments, or scenarios involving high dropout risks due to the extended study duration, and dropout can introduce bias if not evenly distributed across sequences. Analytical challenges, such as handling period effects or unequal carryover, often require specialized statistical models like mixed-effects regression to ensure valid inferences.

Introduction

Definition and Purpose

A crossover study is a longitudinal research design in which each participant serves as their own control by sequentially receiving two or more different treatments or interventions over specified periods, typically with intervening washout phases to mitigate potential carryover effects from prior treatments. This approach allows for within-subject comparisons, where the order of treatment administration is randomized across participants to balance sequences and reduce systematic biases. The primary purpose of a crossover study is to enhance the precision of treatment effect estimates by minimizing inter-subject variability, as the same individuals experience all conditions, thereby increasing statistical power and requiring fewer participants compared to designs relying on between-subject comparisons. This efficiency is particularly valuable in clinical and experimental settings where individual differences, such as genetic or environmental factors, could otherwise confound results and inflate sample size requirements. Key principles underpinning crossover studies include the of treatment to prevent order effects, the of blinding to maintain objectivity where feasible, and the careful balancing of treatment orders across participants to avoid period or biases. Originating in mid-19th-century agricultural experiments to optimize resource use in field trials, the design gained prominence in after the 1950s as statistical methods advanced and its utility for drug comparisons became evident.

Comparison to Other Study Designs

Crossover studies differ from parallel-group designs in that the latter assign different groups of participants to receive distinct treatments simultaneously, which introduces between-subject variability and often requires larger sample sizes to achieve adequate statistical power. In contrast, crossover designs minimize this variability through within-subject comparisons, where each participant receives all treatments in sequence, allowing each individual to serve as their own control and thereby enhancing precision. This within-subject approach typically results in higher efficiency, with crossover trials requiring approximately 50% fewer participants than parallel-group trials to detect the same treatment effect size, assuming no significant carryover effects. Compared to general repeated-measures designs, which involve multiple observations on the same subjects over time and can include non-interventional factors like time or environmental changes, crossover studies represent a specialized that specifically sequences different treatments to directly compare their effects within individuals. While repeated-measures designs broadly capture intra-subject correlations without necessarily involving treatment alternation, crossover designs emphasize randomized treatment orders to control for period and effects in interventional contexts. Crossover designs are particularly suitable for evaluating treatments in chronic, stable conditions where the intervention's effects are short-acting and reversible, such as certain therapies or pharmacokinetic assessments, as this allows for complete washout between periods. They are less appropriate for acute illnesses, fluctuating symptoms, or scenarios involving irreversible treatment effects, like surgical interventions, where carryover or progression could confound results.

Study Design

Key Elements

In a crossover study, treatment sequences refer to the ordered assignment of interventions to participants, typically to balance potential biases. For instance, in a two-treatment, two-period design, participants are allocated to either sequence AB (receiving treatment A followed by treatment B) or BA (treatment B followed by A), with ensuring an equal number in each sequence to mitigate order effects. Washout periods are critical intervals between consecutive treatments, during which no intervention is administered, to minimize carryover effects from prior treatments. The duration is generally calculated as a multiple of the treatment's elimination , with a minimum of three to five recommended to allow residual effects to dissipate and return outcomes to baseline; for example, a with a 24-hour might require a 5–7 day washout. Period effects encompass time-dependent variations in responses across study periods, such as progression, seasonal influences, or participant , which must be accounted for in the to avoid treatment comparisons. These effects are addressed by structuring periods uniformly within sequences and using within-subject analyses that inherently control for them. and balancing involve randomly assigning participants to treatment sequences in equal proportions to counteract sequence or order biases, ensuring that each sequence has the same number of subjects for unbiased estimation of treatment differences. This approach reduces and supports the validity of subsequent statistical inferences. Sample size considerations in crossover studies leverage the within-subject to achieve greater than parallel designs, focusing on the lower within-subject variance for power calculations. A common formula for the total sample size nn in a two-period, two-treatment crossover is: n=(Z1−α/2+Z1−β)2⋅2σe2δ2n = \frac{(Z_{1-\alpha/2} + Z_{1-\beta})^2 \cdot 2\sigma_e^2}{\delta^2} where Z1−α/2Z_{1-\alpha/2} is the critical value for type I error, Z1−βZ_{1-\beta} for power, σe2\sigma_e^2 is the within-subject variance, and δ\delta is the minimum detectable treatment difference; this is adjusted for crossover , often requiring roughly half the sample size of a parallel-group study assuming moderate between periods.

Types of Crossover Designs

Crossover designs vary in complexity to accommodate different numbers of treatments, periods, and subjects, allowing researchers to balance factors such as carryover effects and period biases while maintaining efficiency. The simplest and most commonly used variant is the two-period, two-treatment design, often denoted as AB/BA, where subjects are randomized into two sequences: one group receives treatment A in the first period followed by treatment B in the second (AB sequence), and the other group receives B followed by A (BA sequence). This design ensures each subject serves as their own control, reducing inter-subject variability, and is particularly suitable for direct comparisons between two treatments, such as assessing relative of drug formulations. A washout period between treatments helps mitigate carryover effects in this structure. For studies involving more than two treatments, higher-order crossover designs extend the principle to multiple periods and sequences, such as three-period designs with sequences like ABC, ACB, BAC, BCA, CAB, and CBA. These designs allow each subject to receive all treatments across periods, enabling comprehensive pairwise comparisons while controlling for sequence and period effects through balanced . They are useful when evaluating multiple interventions, though they require larger sample sizes to achieve balance and may increase the risk of dropout or carryover if periods are extended. Latin square designs provide a structured approach for trials with multiple treatments (k > 2) over an equal number of periods, where each treatment appears exactly once in each row (subject) and each column (period) across the . This arrangement ensures balanced exposure, minimizing biases from subject-period interactions, and is often implemented as a k × k square for k treatments. Variants like the Williams design, a type of , further balance first-order carryover effects by ensuring no treatment immediately follows itself in any sequence. are well-suited for dose-response studies in , where escalating doses must be compared within subjects to model concentration-time profiles efficiently. When the number of treatments exceeds the feasible number of periods or subjects, balanced incomplete block designs (BIBD) are employed, treating subjects as blocks and sequences as incomplete sets where not every treatment is received by every subject, but each pair of treatments appears together an equal number of times. In crossover contexts, these designs construct balanced sequences to minimize biases from missing treatment combinations, using parameters like block size (v treatments), replication (r times per treatment), and (λ pairwise comparisons). For instance, a BIBD with v=4 treatments and block size k=3 might use sequences that cover all pairs equally across subjects, making it ideal for resource-limited studies with unequal treatment allocations. This approach maintains statistical power despite incompleteness, though it requires careful construction to avoid .

Statistical Analysis

Addressing Confounding Effects

In crossover studies, confounding effects such as carryover, period, and sequence biases can distort estimates of treatment differences if not properly addressed during analysis. These effects arise from the repeated measures nature of the design, where treatments are administered sequentially to the same subjects, potentially leading to residual influences or systematic variations across periods. Detection and mitigation strategies are essential to ensure valid inferences, typically involving statistical tests, model adjustments, and diagnostic tools tailored to the specific confounder. Carryover effects occur when the influence of a treatment from a previous period persists into the subsequent period, biasing the observed response in the later phase. For instance, in a two-treatment, two-period , this residual effect can confound the direct treatment by inflating or deflating responses in the second period. Detection often relies on pre-treatment baseline measurements to assess residual influences or statistical tests such as a t- comparing outcomes between sequences in the second period to identify differential carryover. If significant carryover is detected, analysts may restrict inference to the first period only, discarding second-period data to avoid . The seminal Grizzle formalizes this approach by first testing for carryover using an analysis of data from both periods, such as a t- on the treatment-by-period interaction; if non-significant, it proceeds to estimate treatment effects using the full . Period effects represent systematic differences in responses across treatment periods, independent of the treatments themselves, often due to temporal factors like disease progression or environmental changes. These can be adjusted for by incorporating period as a fixed effect in the , allowing estimation of treatment effects while accounting for period-specific shifts. For example, in a , the period term isolates these variations, enabling unbiased treatment comparisons across sequences. Analysis of variance (ANOVA) is commonly used to test the significance of period effects, with non-significant results supporting the assumption of uniformity. Sequence effects arise from the order in which treatments are administered, potentially introducing if the design does not balance sequences equally. Mitigation primarily occurs through balanced at the design stage, ensuring equal numbers of subjects in each (e.g., AB and BA), which orthogonalizes sequence effects from treatment comparisons in balanced designs. While sequence effects are assumed absent under proper , they cannot always be statistically tested directly; instead, their impact is evaluated indirectly through model residuals or sequence-stratified analyses. Other confounders, such as subject-by-treatment interactions, reflect heterogeneity in treatment responses across , which can mimic or exacerbate carryover biases. These interactions are using ANOVA to assess variance components for subject-specific treatment effects, identifying if individual differences significantly modify outcomes. If detected, subgroup analyses or mixed-effects models with random interaction terms may be employed to model this heterogeneity without discarding data. Diagnostic approaches complement statistical s by providing visual insights into potential confounders. Graphical methods, such as period-treatment interaction plots (e.g., boxplots of responses by period and treatment), help identify patterns like diverging trends indicative of carryover or period shifts. The Grizzle integrates such diagnostics within its two-stage procedure, combining t-tests with residual plots for comprehensive evaluation. These tools, often implemented in software like SAS PROC GLM or MIXED, facilitate early detection before full analysis.

Analytical Methods and Models

Mixed-effects models form the cornerstone of statistical analysis in crossover studies, accommodating the correlated nature of repeated measurements within subjects through random effects for subjects and fixed effects for treatments, periods, and interactions. These models enable robust estimation of treatment effects while controlling for potential period biases and individual variability. A general for the response in a crossover design is given by the linear : Yijk=μ+τi+πj+σk+(τπ)ij+εijk,Y_{ijk} = \mu + \tau_i + \pi_j + \sigma_k + (\tau \pi)_{ij} + \varepsilon_{ijk}, where YijkY_{ijk} represents the observed response for the kk-th subject in the jj-th period under the ii-th treatment, μ\mu is the grand mean, τi\tau_i is the fixed treatment effect, πj\pi_j is the fixed period effect, σk∼N(0,σs2)\sigma_k \sim N(0, \sigma_s^2) is the random subject effect, (τπ)ij(\tau \pi)_{ij} is the fixed treatment-by-period interaction, and εijk∼N(0,σ2)\varepsilon_{ijk} \sim N(0, \sigma^2) is the residual error. This structure assumes independence across subjects and periods conditional on fixed effects, with estimation typically via restricted maximum likelihood (REML) to handle unbalanced data or missing observations. For simpler two-period, two-treatment (2x2) crossover designs, analysis of variance (ANOVA) provides an accessible method to decompose the total variability into components for treatments, periods, subjects (nested within sequences), and residuals, facilitating tests of treatment differences via the for treatments divided by the residual . This approach assumes normality and but offers straightforward F-tests for fixed effects, with subject variability partitioned to enhance precision over parallel designs. In practice, the ANOVA framework underlies many software implementations and is particularly useful when interactions are minimal or absent. When parametric assumptions such as normality fail or data are ordinal, non-parametric methods like the are applied to the within-subject differences between treatments, providing a distribution-free assessment of the treatment effect while accounting for paired structure. This test ranks the absolute differences and signs them according to direction, offering robustness to outliers and non-normal distributions common in small crossover samples. It is especially valuable in early-phase trials or with skewed outcomes, though it requires symmetric difference distributions for validity. Implementation of these analyses is facilitated by statistical software such as R's lme4 package for fitting mixed-effects models via the lmer function, which supports complex random structures and REML estimation for crossover data. Similarly, SAS PROC MIXED offers versatile tools for specifying fixed and random effects in crossover settings, including options for handling unequal periods or dropouts through structures like unstructured or compound symmetry. These tools automate variance component estimation and hypothesis testing, with lme4 emphasizing open-source flexibility and PROC MIXED providing robust integration with datasets. Power and sample size planning in crossover studies must incorporate intra-subject correlation ρ\rho, which reflects the similarity of measurements within individuals and drives efficiency gains over parallel designs. The variance of the treatment effect estimator is reduced by a factor of (1 - ρ)/2 relative to a parallel design, so the required sample size is (1 - ρ)/2 times that of a parallel design for equivalent power. For instance, with ρ = 0.5, this quarters the required size compared to independent groups. Calculations typically use simulation or analytic formulas based on mixed models, adjusting for dropout rates and ensuring adequate power (e.g., 80-90%) to detect clinically meaningful effects while briefly accounting for potential carryover via sensitivity analyses.

Advantages and Limitations

Advantages

Crossover studies offer substantial advantages in efficiency and precision over parallel-group designs by leveraging within-subject comparisons, where each participant serves as their own control, thereby eliminating between-subject variability and reducing overall noise in the data. This approach minimizes inter-individual differences in baseline characteristics, , and environmental factors that can confound results in parallel designs, leading to increased statistical sensitivity and power. For instance, by focusing on intra-subject differences, crossover designs can achieve a substantial reduction in the required sample size for detecting treatment effects, often up to 50% or more depending on the within-subject , making them particularly valuable for studies where resources are limited. The reduced need for participants translates to significant cost savings and ethical benefits, especially in trials involving rare diseases, expensive interventions, or vulnerable populations where recruiting large cohorts is challenging. Fewer subjects are required to achieve adequate statistical power, as the design's allows for robust comparisons with smaller groups, potentially halving the sample size compared to parallel studies while maintaining equivalent precision. Ethically, crossover designs ensure that all participants receive every treatment condition, avoiding the need for placebo-only arms and providing equitable access to potentially beneficial interventions, which enhances participant and . Additionally, crossover studies excel at detecting individual-level responses by capturing intra-subject variability, offering insights into personalized treatment effects that parallel designs often overlook due to aggregated between-subject . This granularity is crucial for understanding heterogeneity in responses, such as varying across individuals, and supports advancements in precision medicine. In the context of testing, crossover designs have been the standard recommended by the FDA since 1992 for evaluating generic drugs, enabling efficient demonstration of comparable pharmacokinetic profiles with minimal subjects through direct within-subject comparisons of test and reference formulations.

Limitations

One major limitation of crossover studies is the risk of carryover effects, where the influence of a prior treatment persists into subsequent periods, potentially treatment effect estimates if the washout period is inadequate. This can lead to overestimation or underestimation of effects, particularly in designs without sufficient separation between treatments. For instance, in pharmacological studies, residual drug activity from the first period may confound outcomes in the second, violating the assumption of no carryover and requiring complex adjustments to mitigate . Period effects and subject dropout further complicate crossover studies by introducing time-dependent variability and incomplete data. Period effects arise when external factors, such as disease progression or seasonal influences, differ across study periods, inflating variance and potentially masking true treatment differences. Dropout, often higher due to prolonged study duration or adverse events after the first period, can unbalance the design, leading to loss of and reduced statistical power; this is especially problematic in two-period designs where withdrawals disrupt the within-subject comparison. Crossover designs are unsuitable for certain medical conditions, including curative treatments, acute diseases, or therapies with high , such as cancer interventions. In curative scenarios, successful treatment in the first period may eliminate the condition, preventing valid assessment in the second period. For acute or rapidly evolving diseases, the assumption of stable underlying conditions is violated, rendering the design infeasible. In , crossover can be ethically problematic and misleading, as it may dilute survival endpoints or expose patients to sequential toxicities without clear benefits. The statistical analysis of crossover studies is inherently complex, demanding advanced models to account for period, , and potential carryover effects, with violations of key assumptions like treatment-period interactions leading to invalid results. Unlike parallel designs, crossover analyses require specialized techniques, such as mixed-effects models, to handle within-subject correlations, increasing the of errors in and interpretation, particularly in smaller samples. Addressing effects, as discussed in related statistical frameworks, adds further layers of methodological rigor but does not eliminate these analytical challenges. Regulatory agencies, including the FDA and EMA, impose restrictions on crossover designs for specific applications, such as long-term outcome studies. In oncology trials evaluating overall survival, agencies recommend limiting crossover to prevent dilution of treatment effects, as outlined in recent FDA draft guidance. However, in public health emergencies like the (2020-2021), crossover designs have been adapted for some vaccine trials to allow placebo participants access to active treatment while preserving trial integrity.

Applications

In Clinical Research

Crossover studies are widely employed in to evaluate , , and in controlled settings, particularly for conditions where patient stability allows within-subject comparisons. These designs are especially valuable in early-phase trials for their ability to reduce variability and sample size requirements compared to parallel-group studies. In trials, crossover designs facilitate direct comparisons between generic and brand-name drugs by administering both formulations to the same participants under fasting or fed conditions. Regulatory standards, such as the U.S. and Drug Administration's 2021 guidance, require that the 90% intervals for key pharmacokinetic parameters—area under the curve (AUC) and maximum concentration (C_max)—fall within 80% to 125% of the reference product to establish therapeutic equivalence. This approach minimizes inter-subject variability, enabling efficient approval of generics for a broad range of medications. Pharmacokinetic evaluations often utilize crossover studies in healthy volunteers to assess absorption rates, , and elimination profiles. For instance, single-dose, two-period crossovers compare oral versus intravenous administration, revealing differences in absorption efficiency, as demonstrated in trials with compounds like where was estimated at approximately 3%. These studies provide critical data for dosing recommendations and formulation optimization without exposing patients to unnecessary risks. Dose-finding studies in stable patient populations, such as those with , leverage crossover designs to sequence escalating doses and evaluate dose-response relationships. In trials involving dihydropyridine calcium antagonists like and , participants received sequential treatments to compare antihypertensive effects, highlighting the design's utility in identifying optimal dosing with fewer participants. Notable historical applications include insulin crossover trials for from the onward, such as double-blind comparisons of porcine versus bovine insulin in established patients, which informed species-specific and efficacy. In the 2020s, crossover designs continue in prophylaxis, exemplified by randomized trials assessing prophylactic caffeine's role in reducing hypercapnia-induced severity, demonstrating sustained benefits over . The International Council for Harmonisation's E10 guideline (2000) endorses crossover designs within placebo-controlled or active-control frameworks for early-phase clinical trials, emphasizing their role in ethical and efficient demonstrations when carryover effects are manageable.

In Non-Clinical Fields

Crossover designs have been employed in agricultural research since the mid-19th century to compare crop yields under varying treatments, allowing each experimental unit, such as a plot of land, to receive multiple interventions sequentially while controlling for environmental variability. One of the earliest documented applications occurred in at the Rothamsted Experimental Station, where John Bennet Lawes and Joseph Henry Gilbert used a crossover approach to assess the effects of different types on production over successive periods. Ronald A. Fisher advanced these methods in his 1935 book , incorporating principles like and blocking into trials to enhance precision in yield comparisons across rotations. This design's efficiency in reducing between-unit variation made it foundational for modern agronomic studies evaluating amendments or crop varieties. In and , crossover studies facilitate the sequencing of interfaces or stimuli to minimize individual differences in factors , particularly in controlled environments like simulators. For instance, a 2x2 crossover design has been used to evaluate the impact of adding a to simulator displays on cybersickness and , with participants experiencing conditions in randomized order across visits to isolate effects. Similarly, repeated-measures crossover approaches in factors experiments compare attention maintenance training programs to assess behavioral adaptations. These applications highlight the 's value in psychological studies of user interaction, enabling within-subject comparisons that reveal subtle effects on and response times without large sample sizes. Environmental science utilizes crossover designs in animal models to examine sequential exposures, providing insights into cumulative or alternating effects on physiological responses. A notable example involves a 2x2 crossover study on mice exposed to real-world ambient during developmental stages, combined with treatments to perturb , demonstrating how pollution sequences influence metabolic outcomes and microbial composition. This approach controls for animal-specific baselines, allowing researchers to attribute changes in biomarkers, such as , to specific exposure orders in chambers. By randomizing treatment sequences, these studies mitigate carryover biases and enhance the reliability of findings on in ecological contexts. Economic experiments leverage crossover designs for consumer preference testing by presenting products in varied orders to the same participants, capturing order effects and individual preferences efficiently. In auctions evaluating demand, a two-phase crossover setup exposed groups to price treatments sequentially, revealing how familiarity influences willingness-to-pay without between-group . Another application tested date label formats ("Best by" vs. "Use by") on food waste decisions using a crossover , where participants bid on identical items under alternating conditions, showing label impacts on perceived quality and consumption behavior. This method's within-subject structure reduces sample requirements while quantifying preference shifts in and . In the 2020s, crossover designs have extended to validation, enabling controlled simulations to compare through sequential application to the same datasets or users, thus isolating performance differences. For example, a crossover experiment contrasted AI-assisted tactical tools against traditional methods in , with participants alternating approaches to evaluate accuracy and learning gains. Such designs support robust benchmarking in simulations by crossing over models across subjects or tasks to assess and efficiency metrics like accuracy and convergence speed.

References

  1. https://www.cancer.gov/publications/dictionaries/cancer-terms/def/crossover-study
  2. https://pmc.ncbi.nlm.nih.gov/articles/PMC8342834/
  3. https://online.stat.psu.edu/stat509/book/export/html/749
  4. https://www.fda.gov/media/175389/download
  5. https://pmc.ncbi.nlm.nih.gov/articles/PMC1113275/
  6. https://online.stat.psu.edu/stat509/lesson/15/15.2
  7. https://online.stat.psu.edu/stat509/lesson/15
  8. https://digitalcommons.wayne.edu/cgi/viewcontent.cgi?article=4547&context=oa_dissertations
  9. https://math.montana.edu/grad_students/writing-projects/Pre-2000/98choochao.pdf
  10. https://pmc.ncbi.nlm.nih.gov/articles/PMC6434767/
  11. https://pmc.ncbi.nlm.nih.gov/articles/PMC3345345/
  12. https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0133023
  13. https://online.stat.psu.edu/stat502/lesson/12
  14. https://pubmed.ncbi.nlm.nih.gov/6657039/
  15. https://pmc.ncbi.nlm.nih.gov/articles/PMC11534298/
  16. https://www.fda.gov/media/87219/download
  17. https://online.stat.psu.edu/stat503/lesson/4/4.6
  18. https://www.bmj.com/content/366/bmj.l4378
  19. https://taylorandfrancis.com/knowledge/Medicine_and_healthcare/Medical_statistics_%2526_computing/Latin_squares/
  20. https://pmc.ncbi.nlm.nih.gov/articles/PMC2772210/
  21. https://www.jstor.org/stable/2529652
  22. https://anale-informatica.tibiscus.ro/download/lucrari/17-1-05-Adebara.pdf
  23. https://www.taylorfrancis.com/books/mono/10.1201/b17537/design-analysis-cross-trials-byron-jones-michael-kenward
  24. https://pmc.ncbi.nlm.nih.gov/articles/PMC10985791/
  25. https://www.statsdirect.com/help/analysis_of_variance/crossover.htm
  26. https://pubmed.ncbi.nlm.nih.gov/10544305/
  27. https://www.routledge.com/Design-and-Analysis-of-Cross-Over-Trials/Jones-Kenward/p/book/9781439861424
  28. https://pubmed.ncbi.nlm.nih.gov/7889227/
  29. https://pmc.ncbi.nlm.nih.gov/articles/PMC8278783/
  30. https://www.lexjansen.com/phuse/2006/st/ST07.pdf
  31. https://www.sciencedirect.com/science/article/pii/S0002916523166883
  32. https://www.nap.edu/read/10078/chapter/4
  33. https://pmc.ncbi.nlm.nih.gov/articles/PMC4359650/
  34. https://www.fda.gov/media/163638/download
  35. https://www.fda.gov/downloads/drugs/guidances/ucm070244.pdf
  36. https://pmc.ncbi.nlm.nih.gov/articles/PMC3981898/
  37. https://www.ncss.com/wp-content/themes/ncss/pdf/Procedures/NCSS/Analysis_of_2x2_Cross-Over_Designs_using_T-Tests_for_Superiority_by_a_Margin.pdf
  38. https://www.fda.gov/media/188274/download
  39. https://pmc.ncbi.nlm.nih.gov/articles/PMC8882397/
  40. https://database.ich.org/sites/default/files/E10_Guideline.pdf
  41. https://pmc.ncbi.nlm.nih.gov/articles/PMC4759723/
  42. https://link.springer.com/article/10.1007/BF02125734
  43. https://www.sciencedirect.com/science/article/pii/S0140673682905487
  44. https://pubmed.ncbi.nlm.nih.gov/40019582/
  45. http://tankona.free.fr/fisher1935.pdf
  46. https://www.tandfonline.com/doi/full/10.1080/00140139.2018.1518543
  47. https://www.researchgate.net/publication/283330123_Crossover-Repeated_Measures_Designs_Clarifying_Common_Misconceptions_for_a_Valuable_Human_Factors_Statistical_Technique
  48. https://pmc.ncbi.nlm.nih.gov/articles/PMC12257794/
  49. https://www.intechopen.com/chapters/79570
  50. https://www.jstor.org/stable/43556172
  51. https://onlinelibrary.wiley.com/doi/full/10.1002/aepp.13507
  52. https://www.sciencedirect.com/science/article/pii/S0001691825007449
Add your contribution
Related Hubs
User Avatar
No comments yet.