Free scoping read. A written assessment of your manuscript in two business days, at no cost.

Start one
Contents

Academy · Research Data, Analytics & AI · Biostatistics in Health Research

The Importance of Biostatistics in Health Data Research: Investment, Workforce and Reproducibility

A 2019 perspective in the Medical Journal of Australia warned that health research risks ‘drowning in data but starving for knowledge’ because biostatistics has never been treated as a core discipline. The warning has direct consequences for reproducibility.

Pubrica AcademyWritten by a biostatistics and health data research writer Reviewed bya biostatistics methodologist Published10 October 2026 Last checked10 October 2026 Reading time6 minutes Sources10 peer-reviewed

Statistics is often treated as a service function added to a study after the substantive research has been completed. A 2019 perspective in the Medical Journal of Australia, authored by Katherine J Lee, Margarita Moreno-Betancur, Jessica Kasza, Ian C Marschner, Adrian G Barnett and John B Carlin, argued the opposite: biostatistics is a fundamental discipline without which modern health data science cannot function, and chronic under-investment in it is already manifesting as unreliable findings.

Learning objectives

  • The authors’ characterisation of Australian health research as structurally under-invested in biostatistics, with no major university biostatistics department.
  • The contribution of superficial statistical practice, which the authors term ‘cargo-cult statistics’, to the reproducibility crisis.
  • An example from practice (hormone therapy and cardiovascular risk) in which improved statistical methods changed a clinical conclusion.

Common misconceptions

  • “Biostatistics is just a checklist step clinicians can learn adequately from a short course.” The authors argue that sound biostatistical work requires extensive training in mathematics, probability, bias and computing that most clinicians would find very difficult to acquire incidentally; they therefore call for dedicated training pathways and career structures rather than one-off courses.
  • “Statistical rigour mainly affects a study's precision, not its conclusions.” The authors' hormone-therapy example indicates otherwise: observational studies initially suggested a reduced risk of heart attack, a randomised trial contradicted them, and only a later re-analysis using improved methods for handling confounding reproduced the trial's result. The statistical approach itself therefore changed the clinical conclusion, not merely its margin of error.

Insufficient investment in biostatistics

Lee, Moreno-Betancur, Kasza, Marschner, Barnett and Carlin published "Biostatistics: a fundamental discipline at the core of modern health data science" in the Medical Journal of Australia, 211(10):444–446.e1, in 20191. Their central claim is that Australia has never systematically invested in biostatistics, either through universities or national funding schemes, and that, as a result, no major Australian university has a dedicated biostatistics department1. They contrast this with the United States, United Kingdom and continental Europe, where standalone biostatistics departments, national centres and dedicated funding streams have existed for decades1.

The authors identify the Victorian Centre for Biostatistics (ViCBiostat) as the only directly funded biostatistics group they could find in Australia and note that its funding ended in 2017, leaving short-term grants as the means of sustaining what should be a stable critical mass of methodological expertise1.

Workforce and training deficits in biostatistics

The practical consequence is evident in who conducts the statistical work on major studies. The authors note that researchers developing platform trials in Australia often rely on overseas biostatistical expertise because the local workforce and infrastructure are insufficient1. Where biostatistics expertise does exist, it tends to be scattered across small, loosely connected groups, often housed in mathematics or business faculties rather than embedded in medical research institutions1.

Training constitutes the second dimension of the problem. The authors argue that the depth of statistical and computing skill required is very difficult for most clinicians to acquire alongside clinical training, and call instead for structured pathways that take individuals from non-specialist roles through to research leadership in biostatistics1.

The perspective explicitly links superficial statistical understanding — which the authors describe using the term "cargo-cult statistics" — to unscientific practice and to the broader reproducibility crisis in health research1. They illustrate this with an example: observational studies of postmenopausal hormone therapy suggested that it reduced heart attack risk, but the Women's Health Initiative's large randomised controlled trial, published in JAMA in 2002, contradicted those findings8. A later re-analysis of the observational data, using improved methods to account for confounding, produced results consistent with the trial1. The statistical method, and not only the data, had produced the erroneous conclusion.

Propagation of cargo-cult statistics through p-hacking and flexible analysis

"Cargo-cult statistics" is not an abstract failure mode. Simmons, Nelson and Simonsohn's widely cited 2011 paper in Psychological Science, "False-Positive Psychology," demonstrated mathematically that undisclosed flexibility in how data are collected and analysed — choosing which outcome to report, when to stop collecting data, which covariates to include — can push the true false-positive rate for a nominal 5% significance threshold above 60%2. Head, Holman, Lanfear, Kahn and Jennions extended this concern empirically in their 2015 PLOS Biology paper "The Extent and Consequences of P-Hacking in Science," which examined the distribution of p-values reported across a large sample of published papers and found a pattern consistent with widespread, if not always deliberate, p-hacking — results clustering suspiciously just below the 0.05 threshold9. Neither paper concerns biostatistics specifically, but both describe the superficial, assumption-free statistical practice that Lee et al. warn produces unreliable health research when it is not subject to trained methodological oversight.

Wider evidence on power, significance and the warning of the American Statistical Association (ASA)

Two further strands of evidence explain why Lee et al. regard this problem as urgent rather than theoretical. Button and colleagues' 2013 analysis in Nature Reviews Neuroscience, "Power failure: why small sample size undermines the reliability of neuroscience," found that the median statistical power of studies in that field was strikingly low, meaning most published "significant" findings were, by the numbers, more likely to be false positives or substantial overestimates of a real effect than genuine, precisely estimated discoveries7. And in 2016 the American Statistical Association itself, through Wasserstein and Lazar, issued a formal statement warning that the p-value is widely misinterpreted and misused, and that "a conclusion does not immediately become 'true' on one side of the divide and 'false' on the other" of a 0.05 threshold4. Because it originates from the ASA rather than an individual critic, the statement constitutes an institutional acknowledgement of the superficial statistical culture that Lee et al. describe: it is the discipline's own professional body stating the problem publicly rather than a marginal complaint.

The SAMPL guidelines for statistical reporting

One concrete response to under-trained statistical reporting is the SAMPL guidelines — Statistical Analyses and Methods in the Published Literature — developed by Lang and Altman and published through the European Association of Science Editors' Science Editors' Handbook in 20133. SAMPL sets out specific, verifiable reporting requirements: naming every statistical test used, reporting exact p-values rather than only “p < 0.05”, giving confidence intervals alongside point estimates, and describing how missing data were handled. It establishes a minimum standard rather than a substitute for methodological training: a paper can satisfy every SAMPL item and still have used an inappropriate test for its data, which is precisely the gap that, according to Lee et al., only a properly resourced biostatistics discipline can close.

Global workforce responses beyond Australia

Lee et al.'s 2019 warning was specific to Australia, but the workforce problem it describes is being addressed elsewhere through structured training programs of the kind they recommend. A 2025 paper in Clinical Trials describes CANSTAT, a pan-Canadian, multi-institutional training program built to address Canada's own shortage of biostatistics expertise for clinical trials, combining mentorship, trial-site placements, online workshops and tracked individual development plans; the program had enrolled 19 fellows at the time of reporting10. The existence of a dedicated national program elsewhere underlines what Lee et al. identify as missing in Australia specifically: a stable, funded structure for building biostatistical capacity rather than relying on short-term grants or overseas expertise study by study.

Actions for researchers and institutions

The authors' recommendations are organisational as much as technical: to build departmental structures with a genuine critical mass of academic biostatisticians, to support training and career pathways so that expertise does not depend on a small number of individuals, and for funding bodies to invest directly in methodological research such as graduate and postdoctoral programs in biostatistics1. The broader reproducibility literature supports those recommendations: Begley and Ioannidis, writing in Circulation Research in 2015 on reproducibility in basic and preclinical research, and Ioannidis's own earlier and widely cited 2005 PLoS Medicine paper "Why Most Published Research Findings Are False," both trace a meaningful share of irreproducible findings back to exactly the kind of under-resourced statistical practice Lee et al. describe65. Their closing warning is pertinent to the planning of any health data study: without sustained investment in biostatistics as a discipline, the research enterprise risks “drowning in data but starving for knowledge”1.

Key points

Key points

Treating biostatistics as an afterthought rather than a core discipline may alter a study’s conclusions as well as weaken its precision, as the hormone-therapy re-analysis in the 2019 MJA perspective by Lee et al. illustrates.

Practical application

Practical application

  1. A biostatistician should be involved at the design stage rather than only at analysis. Lee et al. argue that under-investment becomes apparent earliest at the design stage; statistical expertise introduced after data collection is too late to address confounding and bias adequately.
  2. The strategy for handling confounding should be stated explicitly in the methods. The hormone-therapy example shows that the method used to handle confounding, and not only the sample size, can reverse a conclusion; the approach should therefore be stated and justified rather than defaulting to an unadjusted analysis.
  3. Methodological expertise should be budgeted as a line item. Biostatistical input should be treated as a funded component of the study, consistent with the authors’ call for institutions and funders to invest directly in methodological research capacity.

Frequently asked questions

It is the term Lee et al. use for statistical practice that follows the surface form of a method, such as running a test or reporting a p-value, without an underlying understanding of what the method assumes or what its result means, which they argue contributes to unreliable findings.

The 2019 MJA perspective focuses on Australia specifically, documenting the absence of a university biostatistics department and lost funding there, while noting by contrast that the US, UK and continental Europe have long-standing departments and funding structures — the underlying training and investment argument is not country-specific, but the workforce data cited is Australian.

References

  1. Lee KJ, Moreno-Betancur M, Kasza J, Marschner IC, Barnett AG, Carlin JB. Biostatistics: a fundamental discipline at the core of modern health data science. Medical Journal of Australia. 2019;211(10):444-446.e1. read 10 October 2026. View source
  2. Simmons JP, Nelson LD, Simonsohn U. False-Positive Psychology: Undisclosed Flexibility in Data Collection and Analysis Allows Presenting Anything as Significant. Psychological Science. 2011;22(11):1359-1366. read 10 October 2026. View source
  3. Lang TA, Altman DG. Basic statistical reporting for articles published in clinical medical journals: the SAMPL Guidelines. In: Science Editors' Handbook. European Association of Science Editors; 2013. read 10 October 2026. View source
  4. Wasserstein RL, Lazar NA. The ASA Statement on p-Values: Context, Process, and Purpose. The American Statistician. 2016;70(2):129-133. read 10 October 2026. View source
  5. Ioannidis JPA. Why Most Published Research Findings Are False. PLoS Medicine. 2005;2(8):e124. read 10 October 2026. View on PubMed Central
  6. Begley CG, Ioannidis JPA. Reproducibility in science: improving the standard for basic and preclinical research. Circulation Research. 2015;116(1):116-126. read 10 October 2026. View source
  7. Button KS, Ioannidis JPA, Mokrysz C, Nosek BA, Flint J, Robinson ESJ, Munafò MR. Power failure: why small sample size undermines the reliability of neuroscience. Nature Reviews Neuroscience. 2013;14(5):365-376. read 10 October 2026. View source
  8. Rossouw JE, Anderson GL, Prentice RL, et al. Risks and benefits of estrogen plus progestin in healthy postmenopausal women: principal results from the Women's Health Initiative randomized controlled trial. JAMA. 2002;288(3):321-333. read 10 October 2026. View on PubMed
  9. Head ML, Holman L, Lanfear R, Kahn AT, Jennions MD. The Extent and Consequences of P-Hacking in Science. PLOS Biology. 2015;13(3):e1002106. read 10 October 2026. View on PubMed Central
  10. Training the next generation of clinical trial biostatisticians. Clinical Trials. 2025;22(6):687-694. read 10 October 2026. View on PubMed

Every source above is a peer-reviewed article, a publisher’s record of one, or a standard issued by the body that owns it. Each carries the date it was last read. Where a standard moves, this article is revised rather than left up with a note.

How to cite this article

Pick your referencing style and copy it. Your university or target journal may vary the punctuation — check its guide if it has one.

Pubrica Academy. (2026, October 10). The Importance of Biostatistics in Health Data Research: Investment, Workforce and Reproducibility. Pubrica. https://pubrica.com/academy/biostatistics/biostatistics-is-a-critical-subject-in-current-health-data-research/

This article is written to be learned from. Cite it if you are referring to what it says; if you are using the evidence behind it, cite the sources listed above instead.

Biostatistics involvement checklist for a health data study

This checklist is intended for use when scoping a new health research project.

[ ] Biostatistician involved at design stage, before data collection
[ ] Confounding-handling method named and justified in the protocol
[ ] Statistical analysis plan written before unblinding/analysis, with no undisclosed flexibility in outcomes or stopping rules
[ ] Exact p-values, confidence intervals and missing-data handling reported per SAMPL
[ ] Methodological expertise budgeted as a funded line item
[ ] Re-analysis or sensitivity check planned for key observational findings

Pubrica AcademyWritten and maintained by Pubrica’s subject matter experts. Articles are revised when the underlying standards change, and each article states the date on which it was last reviewed.

How Pubrica can help

Pubrica assigns trained biostatisticians to work with the study team from the design stage through analysis, so that decisions on confounding, power and methods are made before data collection begins.

Find something else

Search the Pubrica Academy

Access is free and requires no account. Enter a topic, or open the full library to filter by category, sub-topic or resource type.

Integration of statistical design from the outset of a study

A scoping assessment is free and takes two business days. It provides a written assessment of what the work requires and which specialist would undertake it.

Get a free quote

Regulations

Check your project against this rule

  1. 1Your work
  2. 2Your details
  3. 3Send
What do you need?

Material you send is held under signed confidentiality, with access limited to the assigned team and encrypted transfer. It is not used to train anything and is not passed to third-party tools. Ask for the agreement before you send anything you would rather cover first.