Departmental biostatisticians, commercial review services, automated checkers, and RigorMD — an honest comparison of scope, turnaround, and cost, so you can match the tool to the question.
Before a clinical manuscript goes out, most authors want a second read on its methods and statistics. There is no single right way to get one — the options differ in what they catch, how long they take, what they cost, and where in the project they help most. A second read is really two checks at once — whether the methodology and design hold up, and whether the numbers and their interpretation are sound — and the options below are not equally good at both. The honest framing is that they are complementary, not interchangeable. Below is what each category does well, where it falls short, and how RigorMD fits among them.
One distinction runs through all of it: design-stage input versus pre-submission review. The highest-value statistical help comes before you collect data, when the design can still change. Most of the options here, including ours, operate later — on a finished draft. That is useful, but it is not a substitute for getting the design right at the start.
The gold standard, when you can get it. A qualified statistician involved from the design stage will shape your endpoints, sample size, and analysis plan before a single patient is enrolled — the point of maximum leverage. They understand your field, can defend the choices to reviewers, and are accountable for their judgment in a way no automated tool is.
The constraint is access. Biostatistical expertise is scarce and gated: academic statistical units are capacity-strained, and the NIH now expects a statistician on trial grants ↗, which concentrates limited time on funded work. A trainee or an unfunded project may wait weeks, or find no one available at all. When you can secure design-stage involvement from a statistician, take it. This is not the category to replace — it is the one to protect for the questions that genuinely need a human expert.
Human review, on a turnaround. A number of services offer paid statistical or methodological review of a finished manuscript by a contracted expert, typically returned over a span of days. This buys you human judgment without an institutional queue, which is genuinely useful when no local statistician is available.
The trade-offs are cost, turnaround, and variability. Published list prices vary, but a manuscript-level statistical review typically runs in the range of $150–$400, and more for fast turnaround or a comprehensive methodological critique. Quality depends on the individual reviewer assigned, and turnaround is measured in days, not hours. For a high-stakes manuscript with no other access to expertise, the cost can be worth it; for routine pre-submission checking, it is a heavier instrument than many projects need.
Narrow, fast, and free — within their scope. A small number of open tools check specific, mechanical properties of a manuscript. The best known is statcheck, which recomputes the consistency of statistical results reported in APA style — pairing each p-value with its test statistic and degrees of freedom and flagging arithmetic that does not hold. It is fast, deterministic, and well validated for what it does.
The limit is scope. statcheck was built for psychology-style reporting and checks essentially one thing: whether reported statistics are internally consistent. It does not assess whether your design supports your claim, whether you were powered, whether your conclusion matches your results, or whether you met your reporting guideline. Within its lane it is excellent; it simply covers a narrow lane.
Automated, broad, and verifiable — for a finished draft.RigorMD sits between the narrow automated checker and the human reviewer. It classifies your study design and claim type, appraises the manuscript with two independent engines, and runs a deterministic layer that recomputes statistics where the reported numbers allow — then reconciles all of it into a severity-scored report across six domains: design–claim fit, results–conclusion alignment, statistical appropriateness, reporting-guideline adherence, numerical consistency, and clinical interpretability. Every serious finding is grounded in a direct quote from your manuscript or a recomputed number you can check. Where the two engines disagree, the disagreement is surfaced in the report rather than averaged away, so you can see where the assessment is confident and where it is not. Turnaround is hours, not days. The pre-submission review is $25. A separate design-planning option — right test, variables, hypothesis, power — is available for studies still in the planning stage (coming soon); that is a different product for a different moment, not a second tier of the same review.
What it does not do, stated plainly.It flags; it does not certify. It is not a substitute for peer review, not a guarantee of acceptance, and not a replacement for a statistician's input on study design. It reads a finished draft — so it cannot recover a design decision that was settled before the data were collected, and without your raw dataset it cannot answer every question. It distinguishes what it checked and passed from what is not checkable from the submitted files, and says which is which.
| Option | Who or what reviews | Best for | Turnaround | Typical cost | Certifies? |
|---|---|---|---|---|---|
| Departmental biostatistician | Human expert, your institution | Design-stage input; defensible analysis plans | Days to weeks, subject to availability | Institutional / grant-funded | No — advises |
| Commercial expert-review service | Contracted human reviewer | A high-stakes manuscript with no local access | Days | Typically $150–$400 (published list prices vary) | No |
| Automated checker (e.g., statcheck) | Deterministic script | Reporting-consistency arithmetic | Seconds | Free | No |
| RigorMD | Two LLM engines + deterministic recompute | A broad pre-submission read on a finished draft | Hours | $25 | No — flags, never certifies |
Cost and turnaround figures for the human and commercial categories are general ranges, stated only as published-list-price ranges; specific providers vary.
Match the tool to the question. If you are still designing the study, find a statistician — no review of a finished draft, automated or human, recovers a design decision made before the data existed. If you have a finished manuscript and access to a trusted biostatistician with time, that human read remains the most authoritative. If you do not, the practical question is breadth versus depth: an automated reporting checker verifies arithmetic in seconds for free; a commercial service buys human judgment over days; RigorMD returns a broad, severity-scored, quote-grounded read in hours for $25. Many authors use more than one — an automated pass to catch the mechanical errors, a human for the questions that need accountability. For example, a funded trial with a named statistician needs no automated breadth check on its arithmetic; an unfunded retrospective cohort drafted without statistical support is exactly where a fast, broad, low-cost read earns its place.
Whichever you choose, the goal is the same: catch the methodological and statistical problems before a reviewer does. For the specific patterns that draw a rejection, see why manuscripts get rejected for statistics. To see what a RigorMD finding looks like, read a full sample report → or review pricing.