Analysis of nutrition intervention studies

Not always. In nutrition, a large share of intervention studies use pre-post or quasi-experimental designs, because randomising a diet or a food programme is not always feasible, acceptable or compatible with the setting in which it is implemented. This applies when you want to evaluate the effect of supplementation, a change of diet or a programme on health indicators, and field conditions do not allow a randomised control group. Analysis of nutrition intervention studies covers the full range: randomised trials where they are possible, and pre-post or quasi-experimental designs where they are not, choosing the one that fits the question and what the setting allows. A pre-post or quasi-experimental design can be useful, but it does not offer the same capacity for attribution as a well-designed randomised trial. That choice shapes what the study will be able to conclude, so it is made when the study is designed, not when it closes.

The best time is before data collection starts, ideally at the study design stage. The service can come in at any point, but sizing the sample and fixing the analysis plan before collection is what best protects the validity of the result; coming in with the data already collected limits what can be corrected afterwards. To get started, you need a definition of the intervention, the planned groups, the indicators to be followed and the intended design. The client runs the intervention and collects the data; the service sizes the study, defines the analysis plan, models the data and estimates the effect that the available design supports. If you arrive with data collection already done, the analysis is still possible, but bear in mind that some design decisions can no longer be undone, and that limits what the study can conclude.

Dietary adherence is modelled as a gradient of exposure, not as a yes or no, because in nutrition nobody follows a diet perfectly, and treating adherence as binary ignores precisely what determines the effect. This applies to any study in which the intervention is a diet, supplementation or a food programme that sits alongside the participant's baseline diet. Unlike more tightly protocolised exposures, a dietary intervention is usually followed to varying degrees and overlaps with what the person was already eating. That actual exposure is therefore modelled as a continuum and integrated with the baseline diet and, where the data allow, the dose-response relationship between degree of exposure and effect is analysed. The estimated effect then does not assume full compliance, which rarely occurs uniformly. That is what separates a realistic nutrition analysis from one that copies the template of a protocolised exposure.

Pre-post change can be analysed with mixed-effects models for repeated measures when the follow-up structure justifies it, following each participant over time instead of comparing isolated means at each time point. This applies to intervention studies that measure the same people before and after, sometimes at several time points. These models use all the follow-up information and account for within-participant correlation, which gives more reliable estimates than comparing isolated group averages. The specific approach depends on the number of time points, the distribution of the outcome, the design and the sample size. The analysis includes intention-to-treat and per-protocol analyses where the design allows, depending on the question. Handling dropouts and missing data over follow-up is part of the analysis, not a detail: how they are treated shapes how the result is read, and it is decided explicitly.

Yes. The different outcome types in a nutrition study are analysed consistently with one another, not as separate studies stitched together, so that the effect is read as a whole. This applies to studies that collect measures of different kinds: anthropometric (weight, height), biochemical (biomarkers), dietary (consumption) and scale-based (well-being, food security, quality of life). Each type of data has its own statistical treatment, and forcing them all into the same mould distorts the result. The service acts as the framework for the study and brings those analyses together into an integrated reading of the effect. When a component needs specialist treatment, such as scales through psychometrics, the microbiota through bioinformatics or the diet-to-outcome pathway through mediation analysis, it can be referred to the corresponding service or combined with it, with this service remaining the framework for the intervention study. The result is an interpretation of the effect that does not leave each measure in its own compartment.

Analysis of nutrition intervention studies measures or estimates the effect of an intervention, such as a diet, supplementation or a programme, on health indicators; population dietary intake estimation quantifies the dietary intake and nutrient availability of a population at a given point in time. One answers "what changed when this intervention was introduced?"; the other, "which nutrients are lacking, and in which subgroups?", as a snapshot of the starting point. They do not answer the same question: this service works on a study with an intervention and follow-up; intake estimation works on consumption data to describe nutritional status. The two can be chained, since an intake estimate can serve as the baseline against which an intervention is later evaluated. But the choice depends on whether there is already an intervention to evaluate, in which case what this service provides is the estimated or attributable effect of an action, depending on the available design, or whether the starting situation is still being characterised.

They share statistical tools, such as mixed-effects models, pre-post designs and sample size calculation, but they serve different fields and answer different questions. Analysis of nutrition intervention studies focuses on what is specific to a dietary study: adherence as a gradient of exposure, integration with the baseline diet and the combination of anthropometric, biochemical, dietary and scale-based outcomes. Biostatistics for academic and pragmatic clinical trials works within the general clinical-pharmaceutical or biomedical framework of a trial, with its own vocabulary and requirements. The difference lies not in the method but in the subject matter and in what each field demands of the analysis: a dietary intervention does not behave like a protocolised exposure, and its analysis cannot assume uniform compliance. If the intervention is nutritional or dietary and the challenge lies in modelling an exposure that is followed to varying degrees, it fits this service. If the main challenge is the general statistical cycle of a clinical trial not centred on dietary exposure, it fits biostatistics for academic and pragmatic clinical trials.

You receive the analysis of the intervention's effect, interpreted and documented: the estimated change in the indicators, the comparison between groups where the design includes one, the analysis of adherence as exposure and the joint treatment of the different outcomes, with results and figures ready for scientific publication. This applies to anyone who needs a result ready for a manuscript, a report or a decision on the programme or intervention. The deliverable also records the analysis decisions taken, such as the groups, the handling of dropouts and missing data, and the model assumptions, so that the result is traceable and reproducible. The research team authors and leads the publication and decides on the intervention; the service provides the design, the analysis, its interpretation and the methodological justification behind it.