" Ioannidis's 2005 paper "
Why Most Published Research Findings Are False" In the paper, Ioannidis says that most published research does not meet good scientific standards of evidence. Ioannidis has also described the
replication crisis in diverse scientific fields including
genetics,
clinical trials,
neuroscience, and
nutrition. His work has aimed to identify solutions to problems in research, and on how to perform research more optimally. In a series of five papers about research published in
The Lancet and titled "Research: increasing value, reducing waste", and examining how to correct weaknesses in
research design, methods, and analysis by involving experienced
statisticians and
methodologists and avoiding stakeholders with
conflicts of interest. Ioannidis's research at Stanford focuses on
meta-analysis and
meta-research – the study of studies. Thomas Trikalinos and Ioannidis coined the term
Proteus phenomenon to describe tendency for early studies on a subject to find larger effect than later ones. He was an early and influential public critic of
Theranos, the now-fallen Silicon Valley blood test startup that at its height was valued at up to $9 billion. He criticized it for "stealth research" that it had not made available for other scientists to review.
Meta-research Ioannidis has defined meta-research to include "thematic areas of methods, reporting, reproducibility, evaluation, and incentives (how to do, report, verify, correct, and reward science)". He has performed large-scale assessments of the presence of reproducible and transparent research indicators such as
data sharing,
code sharing, protocol registration, declaration of
funding and conflicts of interest in
biomedical sciences,
social sciences, and
psychology. He has led or co-led efforts to define and improve reproducibility in science, e.g. computational reproducibility, and to reduce research waste in study design, conduct, and analysis. Ioannidis has co-authored the Manifesto for Reproducible Science, an eight-page document illuminating the need to fix the flaws in the current scientific process and mitigate the "reproducibility crisis" in science. In "Why Most Published Research Findings are False" (2005), Ioannidis focused on why most published research findings cannot be validated. and in a third paper (2016) he showed why
clinical research in particular is usually not useful and how this can be amended. In the first of the three PLOS papers he stated that "a research finding is less likely to be true when the studies conducted in a field are smaller; when effect sizes are smaller; when there is a greater number and lesser preselection of tested relationships; where there is greater flexibility in designs, definitions, outcomes, and analytical modes; when there is greater financial and other interest and prejudice; and when more teams are involved in a scientific field in chase of statistical significance". In the second paper, he discussed solutions: "adoption of large-scale collaborative research; replication culture; registration; sharing; reproducibility practices; better statistical methods; standardization of definitions and analyses; more appropriate (usually more stringent) statistical thresholds; and improvement in study design standards, peer review, reporting and dissemination of research, and training of the scientific workforce". In the third paper, he proposed eight features that are important for useful clinical research: problem base, context placement, information gain, pragmatism, patient-centeredness, value for money, feasibility, and transparency. Ioannidis was invited to present his findings as a
keynote speaker at the
"Evidence Live 2016" conference, hosted jointly by the
Centre for Evidence-Based Medicine (CEBM) at the
Nuffield Department of Primary Care Health Sciences, University of Oxford and the
BMJ. Meta-analysis Ioannidis has developed and popularized several methods for
meta-analysis and has made several conceptual advances in this field. These include methods for assessing
heterogeneity and its
uncertainty, methods for meta-analysis involving
multiple treatments, methods and processes for
umbrella reviews, and several approaches to identifying
bias and adjusting the results of meta-analyses for bias, such as
publication bias and
reporting bias resulting in
funnel-plot asymmetry. He has also alerted about the misuse and misinterpretation of bias tests. Along with
David Chavalarias, he catalogued 235 biases across the entire publication record of biomedical research. Ioannidis has been critical of flawed, misleading and redundant meta-analyses, estimating that few meta-analyses in medicine are both bias-free and clinically useful. He has performed
empirical evaluations of the
concordance of results between meta-analyses and large trials and between
randomized trials and non-randomized studies. In an essay written to honor his late mentor
David Sackett, he stated that: He has described four inter-related problems that create what he calls the Medical Misinformation Mess: He has supported these views by contributing to a meta-epidemiological study which found that only 1 in 20 interventions tested in Cochrane Reviews have benefits that are supported by high-quality evidence and a related study showing that the quality of this evidence does not seem to improve over time.
Statistical methods and inference Ioannidis has made methodological and conceptual contributions to the debates surrounding the use and misuse of
statistical methods and
inference. He has been an advocate of the approach to redefine
statistical significance by requesting more
stringent statistical significance thresholds; he has proposed and empirically validated stringent thresholds for
genome-wide significance in
genetics; and has been critical of the approach to entirely abandon statistical significance.
Reporting guidelines Ioannidis has contributed to several influential guidelines for reporting different types of research, such as
PRISMA for meta-analyses, TRIPOD for multivariable
prognostic and
diagnostic models, and others on clinical trials and
observational research. He is the lead author of the
CONSORT for harms, a guideline that provides guidance on how to properly report on harms in randomized trials and has contributed to PRISMA for harms, a guideline for reporting of harms in meta-analyses.
Genetic and molecular epidemiology Ioannidis was one of the first to advocate the use of meta-analysis in
genetic epidemiology to assess replication and the incorporation of meta-analysis in large-scale consortia of multiple investigators performing
genome-wide association studies. He led and contributed to many such efforts in diverse areas of genetic epidemiology and in other areas of
molecular epidemiology. By means of empirical reviews, he has highlighted that there are studies suggesting that almost every nutrient is associated with cancer risk, which is an implausible situation He has also suggested that more attention is needed for proper disclosures of both financial and non-financial
conflicts of interest in nutrition research. He also co-authored the DIETFITS randomized trial that showed no difference between a
low-fat and a
low-carb diet.
Association studies and big data In an effort to improve the credibility of research on
risk factors, Ioannidis has proposed that exposure-wide or environment-wide
association studies should be performed and he has outlined the similarities and differences between such studies and genome-wide association studies in genetics. By assessing all risk factors together instead of one at a time, this practice aims to reduce
selective reporting and publication bias. He has also advocated for the use of large national population databases with systematically collected data to minimize bias and improve yield of trustworthy discoveries. He has worked on the potential uses of such approaches in
big data and
artificial intelligence.
Psychiatry Ioannidis has performed critical assessments of the evidence behind
mental health interventions (
pharmacotherapy and
psychotherapy). He co-authored a
network meta-analysis on more than 500 randomized trials of
anti-depressants showing a modest benefit from these medications for
major depression. He has identified the potential for sponsorship bias in meta-analyses in mental health and has empirically assessed the totality of meta-analyses on mental health interventions, estimating that beneficial effects do exist, but they tend to be modest and thus a research agenda is needed to identify more effective interventions.
Neuroscience Along with colleagues, Ioannidis has performed empirical evaluations and meta-research assessments of large numbers of scientific studies in
neuroscience and have found that lack of power is a very common problem, leading to both
false-negatives (the inability to discover true signals) and
false-positives (finding spurious signals).
Economics In empirical assessments of all meta-analyses that have been conducted on
economics topics, Ioannidis and colleagues have found that most of the studies in these fields are small and
under-powered. Using bias detection and correction methods, they have concluded that nearly 80% of the reported effects in the empirical economics literature is exaggerated; typically by a factor of two, and with one-third inflated by a factor of four or more. == Editorial appointments ==