Study design and statistical protocols

Because a design flaw cannot be corrected at the analysis stage: you pay for it with data already collected. This applies to the preliminary phase of any observational or experimental study, while the question, the design, the outcomes and the sample can still be decided. The moment of greatest influence over a study's validity is the one before the first data point is collected: a research question poorly translated into a design, an ambiguous outcome or an undersized sample shape everything that follows, and no statistical technique can repair them after the fact. Discovering, data in hand, that the study lacked power, carried too much bias or used an outcome that did not measure what it should means either redoing the study or defending a fragile result. The protocol stage is where methodological rigour pays off most, because it acts on decisions that are frozen afterwards.

The design is chosen according to what the study needs to answer, estimate or evaluate, not from a default template: not every design answers every question. This applies at the initial phase, when the research question has to be translated into a structure that can answer it validly. The choice distinguishes between observational designs, such as cohort, case-control or cross-sectional studies, and experimental ones, such as randomised, pragmatic or implementation trials, depending on the objective and the real constraints of carrying the study out. Alongside the design, we define the target population, the inclusion and exclusion criteria and the unit of analysis, and identify foreseeable sources of bias and confounding, with the strategy to control them built in from the start. Choosing the design well is the decision that most shapes validity, because a design unsuited to the question is rarely offset later by a more sophisticated analysis.

The sample size is calculated from the type of outcome, the minimum relevant effect and the target power, with explicit assumptions about variability, expected event rates and loss to follow-up. This applies at the design stage, where the figure determines the feasibility of the whole study and is one of the first things a committee or an assessor examines. It is not an isolated number but a documented argument: each assumption is recorded so that the figure stands up to review. It comes with a sensitivity analysis showing how the sample size changes as the assumptions vary, and with specific scenarios for limited samples. When the standard assumptions do not fit the design, the sample size is validated by simulation. A sample size without its justification is the first decision a reviewer questions; with it, it stops being an arbitrary figure.

A pre-specified analysis plan is the document that sets, before the data are collected or seen, the analysis methods for each outcome, the analysis populations, the missing-data strategy and the handling of multiplicity. This applies to any study that wants to protect the credibility of its results, because defining the methods before seeing the data protects against opportunistic analysis. Choosing the method after the results are known opens the door to picking, consciously or not, the option that gives the desired result; pre-specifying reduces that risk and keeps the analysis traceable from the protocol onwards. The plan distinguishes the analysis populations and defines how multiplicity will be controlled when there are several comparisons. Having the plan written down from the protocol onwards is what helps distinguish a defensible confirmatory analysis from a later exploration of the data.

Yes: the statistical review of a protocol that is already written is one of the ways into the service, intended for the point before it goes to a committee, a funding call or a journal. This applies if you already have a written protocol and want to make sure its statistical component will hold up under evaluation. The review returns a report of recommended adjustments covering the assumptions behind the sample size calculation, the consistency between the objectives and the proposed analysis, and the points a reviewer or a committee might question. You do not need to have designed the study with ADI to commission this review: it can be contracted as a self-contained entry point, just as the full design from scratch is another. What you receive is a critical, constructive assessment of the statistical component before it is read by whoever decides, while there is still room to adjust it.

Study design and statistical protocols apply before data are collected, when the decision is how to set up the study: type of design, hypotheses, outcomes, sample size and analysis plan. Biostatistics for academic and pragmatic clinical trials covers the statistical cycle specific to a trial, including the analysis of the data once collected. They do not answer the same question: one helps you set up the study properly before it starts; the other supports and analyses the trial while it is under way or once it has closed. They can be two phases of the same project or two different entry points: if you have not started yet, you mainly need design; if your trial is already defined, under way or closed, you will usually need statistical support specific to the trial and, where relevant, the analysis of the data. This service covers any design, observational or experimental, whereas biostatistics for clinical trials focuses on the clinical trial as such.

No. The service covers the design of observational studies and of academic, pragmatic and implementation trials, with a statistical protocol aimed at research ethics committees, competitive funding calls and publication. It does not cover the design of a formal phase I–III regulatory trial intended for a submission to agencies such as the EMA or the FDA, nor the production of the associated regulatory dossier, which is a different competence with its own framework. The line does not blur: academic and pragmatic study design, yes; trial design for a formal regulatory dossier, no. What is included is the methodological work of design, sizing and pre-specification that helps make a study valid and defensible in academic and research settings. If you need the design of a regulatory trial, you will find a clear boundary here, not a vague promise.

The client provides the research question, knowledge of the context and the real constraints on carrying out the study; the service translates that question into a valid design, sizes the sample, pre-specifies the analysis plan and reviews the protocol. This applies to the phase before data collection, when there are no data yet and the work is methodological. At this stage we review the planned sources, the operational definition of each variable and the feasibility of capturing it, so that the design does not depend on information the study will never have. Afterwards, the client will carry out the study and collect and measure the data; the service leaves the methodological scaffolding, made up of the design, the sample size and the analysis plan, on which that data collection rests. What you receive is the methodological foundation of the study before it starts, not an analysis of data that do not yet exist.