An estimate of the science-wise false discovery rate and application to the top medical literature
The DOI matches its Crossref record; 49 of the 49 references it cites were found in a registry or an index.
Abstract
The accuracy of published medical research is critical for scientists, physicians and patients who rely on these results. However, the fundamental belief in the medical literature was called into serious question by a paper suggesting that most published medical research is false. Here we adapt estimation methods from the genomics community to the problem of estimating the rate of false discoveries in the medical literature using reported $P$-values as the data. We then collect $P$-values from the abstracts of all 77 430 papers published in The Lancet, The Journal of the American Medical Association, The New England Journal of Medicine, The British Medical Journal, and The American Journal of Epidemiology between 2000 and 2010. Among these papers, we found 5322 reported $P$-values. We estimate that the overall rate of false discoveries among reported results is 14% (s.d. 1%), contrary to previous claims. We also found that there is no a significant increase in the estimated rate of reported false discovery results over time (0.5% more false positives (FP) per year, $P = 0.18$) or with respect to journal submissions (0.5% more FP per 100 submissions, $P = 0.12$). Statistical analysis must allow for false discoveries in order to make claims on the basis of noisy data. But our analysis suggests that the medical literature remains a reliable record of scientific progress.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“Analyses based on the distribution of P values reported in the medical literature, for example, suggested a false-discovery rate of only 14% (9).”
From Is science really facing a reproducibility crisis, and do we need it to? (2018) · cites this paper as [9]
Citations
Cites
Maximum Likelihood from Incomplete Data Via the EM Algorithm
Dempster, A. P., Laird, N. M., & Rubin, D. B. (1977) · Journal of the Royal Statistical Society Series B: Statistical Methodology · doi:10.1111/j.2517-6161.1977.tb01600.x · Cited by 50,444
Nonparametric Estimation from Incomplete Observations
Kaplan, E. L., & Meier, P. (1992) · Springer Series in Statistics · doi:10.1007/978-1-4612-4380-9_25 · Cited by 45,181 · cited 1×
“This is a similar problem encountered in the analysis of survival data, when patients are lost to follow-up (Kaplan and Meier, 1958).”
Nonparametric Estimation from Incomplete Observations
Kaplan, E. L., & Meier, P. (1958) · Journal of the American Statistical Association · doi:10.1080/01621459.1958.10501452 · Cited by 39,158
Why Most Published Research Findings Are False
Ioannidis, J. P. A. (2005) · PLoS Medicine · doi:10.1371/journal.pmed.0020124 · Cited by 10,505
Statistical significance for genomewide studies
Storey, J. D., & Tibshirani, R. (2003) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.1530509100 · Cited by 9,692 · cited 2×
“…similar statistical methods for estimating the rate of false discoveries in genomic (Kendziorski and others, 2003; Newton and others, 2001; Efron and Tibshirani, 2002; Storey and Tibshirani, 2003) or brainimaging studies (Genovese and others, 2002, where many hypotheses are tested simultaneously.”
Simmons, J. P., Nelson, L. D., & Simonsohn, U. (2011) · Psychological Science · doi:10.1177/0956797611417632 · Cited by 6,567
Casella, G., & Berger, R. (2024) · Chapman & Hall/CRC eBooks · doi:10.1201/9781003456285 · Cited by 5,711
Thresholding of statistical maps in functional neuroimaging using the false discovery rate
Genovese, C. R., Lazar, N. A., & Nichols, T. (2002) · NeuroImage · doi:10.1006/nimg.2001.1037 · Cited by 4,923
Why Most Published Research Findings Are False
Ioannidis, J. P. A. (2005) · CHANCE · doi:10.1080/09332480.2005.10722754 · Cited by 4,702
The interaction of TIGIT with PVR and PVRL2 inhibits human NK cell cytotoxicity
Stanietsky, N., Simic, H., Arapovic, J., Toporik, A., Levy, O., Novik, A., Levine, Z., Beiman, M., Dassa, L., Achdout, H., et al. (2009) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.0903474106 · Cited by 3,524
Publication bias in clinical research
Easterbrook, P. J., Gopalan, R., Berlin, J. A., & Matthews, D. R. (1991) · The Lancet · doi:10.1016/0140-6736(91)90201-y · Cited by 3,109
Generalized, Linear, and Mixed Models
McCulloch, C. E., & Searle, S. R. (2000) · Wiley Series in Probability and Statistics · doi:10.1002/0471722073 · Cited by 2,994 · cited 1×
“We were only able to obtain publication data for the years 2006– 2010 for The Lancet and for the years 2003-2010 for BMJ. 2.4 False discovery rates over time We used a linear model to estimate the rate of increase or decrease in false discovery rates over time (McCulloch and Searle, 2001).”
InfluentialConfidence intervals rather than P values: estimation rather than hypothesis testing.
Gardner, M. J., & Altman, D. G. (1986) · BMJ · doi:10.1136/bmj.292.6522.746 · Cited by 2,010 · cited 1×
“Specifically, it is still important to report estimates and confidence intervals in addition to or instead of P-values when possible so that both statistical and scientific significance can be judged by readers (Altman, 2005; Altman and Gardner, 1986).”
Replication validity of genetic association studies
Ioannidis, J. P. A., Ntzani, E. E., Trikalinos, T. A., & Contopoulos-Ioannidis, D. G. (2001) · Nature Genetics · doi:10.1038/ng749 · Cited by 1,832
Survival Analysis: A Self-Learning Text
Carpenter, M. (1997) · Technometrics · doi:10.1080/00401706.1997.10485091 · Cited by 1,811
Reproducible research in computational science
Peng, R. D. (2011) · Science · doi:10.1126/science.1213847 · Cited by 1,449 · cited 1×
“By distributing our methods, code, and software our approach adheres to the highest standards of reproducible computational research (Peng, 2012).”
Microarray data analysis: from disarray to consolidation and consensus
Allison, D. B., Cui, X., Page, G. P., & Sabripour, M. (2006) · Nature Reviews Genetics · doi:10.1038/nrg1749 · Cited by 1,339
Does high public debt consistently stifle economic growth? A critique of Reinhart and Rogoff
Herndon, T., Ash, M., & Pollin, R. (2014) · Cambridge Journal of Economics · doi:10.1093/cje/bet075 · Cited by 1,197
Toward Evidence-Based Medical Statistics. 1: The P Value Fallacy
Goodman, S. N. (1999) · Annals of Internal Medicine · doi:10.7326/0003-4819-130-12-199906150-00008 · Cited by 1,185 · cited 1×
“There are serious problems with interpreting individual P-values as evidence for the truth of the null hypothesis (Goodman, 1999).”
Biostatistics: A Methodology for the Health Sciences.
D'Agostino, R. B., Fisher, L. D., & van Belle, G. (1995) · Journal of the American Statistical Association · doi:10.2307/2291095 · Cited by 957
Testing a Point Null Hypothesis: The Irreconcilability of P Values and Evidence
Berger, J. O., & Sellke, T. (1987) · Journal of the American Statistical Association · doi:10.1080/01621459.1987.10478397 · Cited by 775 · cited 1×
“Gelman and O’Rourke, 2013) and in the statistical literature (Berger and Sellke, 1987), hypothesis testing has known limitations as a method for hypothesis screening.”
Newton, M. A., Kendziorski, C. M., Richmond, C. S., Blattner, F. R., & Tsui, K. W. (2001) · Journal of Computational Biology · doi:10.1089/106652701300099074 · Cited by 739
Empirical Bayes methods and false discovery rates for microarrays
Efron, B., & Tibshirani, R. (2002) · Genetic Epidemiology · doi:10.1002/gepi.1124 · Cited by 684 · cited 5×
“So, the P-value distribution can be modeled as a mixture of these two distributions (Newton and others, 2001; Efron and Tibshirani, 2002): f (p|a, b, π0) = π0 Uniform(0, 0.05) + (1 − π0)t Beta(a, b; 0.05) (2.1) A similar approach is taken in genomics or brain-imaging studies when many hypotheses…”
InfluentialTesting a Point Null Hypothesis: The Irreconcilability of P Values and Evidence
Berger, J. O., & Sellke, T. (1987) · Journal of the American Statistical Association · doi:10.2307/2289131 · Cited by 548
Macdonald, P. D. M., & Pitcher, T. J. (1979) · Journal of the Fisheries Research Board of Canada · doi:10.1139/f79-137 · Cited by 546 · cited 1×
“…can calculate these probabilities by integrating the distributions over the intervals rounding to 0, 0.01, 0.02, 0.03, 0.04, and 0.05: [0, 0.005), [0.005, 0.015), [0.015, 0.025), [0.025, 0.035), [0.035, 0.045), and [0.045, 0.05], respectively (MacDonald and Pitcher, 1979; Wengrzik and Timm, 2011).”
InfluentialPounds, S., & Morris, S. W. (2003) · Bioinformatics · doi:10.1093/bioinformatics/btg148 · Cited by 425 · cited 2×
“…distribution, but statistical research has shown that when considered as one group, the distribution of true alternative P-values can be modeled as a Beta distribution, Beta(a, b) or mixtures of Beta distributions (Allison and others, 2002, 2006; Pounds and Morris, 2003; Leek and Storey, 2011).”
A general framework for multiple testing dependence
Leek, J. T., & Storey, J. D. (2008) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.0808709105 · Cited by 374 · cited 1×
“However, there are a variety of reasons that this may not be true, including common, unmeasured confounders (Leek and Storey, 2008) or P-value hacking, where individual steps in the statistical processing pipeline are changed to get more significant results (Simmons and others, 2011).”
Self-citationKendziorski, C. M., Newton, M. A., Lan, H., & Gould, M. N. (2003) · Statistics in Medicine · doi:10.1002/sim.1548 · Cited by 370
A mixture model approach for the analysis of microarray gene expression data
Allison, D. B., Gadbury, G. L., Heo, M., Fernández, J. R., Lee, C.-K., Prolla, T. A., & Weindruch, R. (2002) · Computational Statistics & Data Analysis · doi:10.1016/s0167-9473(01)00046-9 · Cited by 360
Publication bias examined in meta-analyses from psychology and medicine: A meta-meta-analysis
van Aert, R. C. M., Wicherts, J. M., & van Assen, M. A. L. M. (2019) · PLOS ONE · doi:10.1371/journal.pone.0215052 · Cited by 334
Baggerly, K. A., & Coombes, K. R. (2009) · The Annals of Applied Statistics · doi:10.1214/09-aoas291 · Cited by 312 · cited 1×
“Reproducible research is a burgeoning field (Peng, 2009, 2011) that has grown in importance with the discovery of statistical coding errors in recent high-profile medical (Baggerly and Coombes, 2009) and economic (Herndon and others, 2013) studies.”
Are all Economic Hypotheses False?
De Long, J. B., & Lang, K. (1992) · Journal of Political Economy · doi:10.1086/261860 · Cited by 295 · cited 1×
“Similar efforts have suggested that almost no null hypotheses are true in the economic literature and most tests that fail to reject the null do so because of a lack of power (De Long and Lang, 1989).”
Survival Analysis: A Self-Learning Text (2nd ed.)
Cavanaugh, J. E. (2006) · Quarterly Publications of the American Statistical Association · Cited by 266
The vexed question of authorship: views of researchers in a British medical faculty
Bhopal, R., Rankin, J., McColl, E., Thomas, L., Kaner, E., Stacy, R., Pearson, P., Vernon, B., & Rodgers, H. (1997) · BMJ · doi:10.1136/bmj.314.7086.1009 · Cited by 242
Reproducible research and Biostatistics
Peng, R. D. (2009) · Biostatistics · doi:10.1093/biostatistics/kxp014 · Cited by 237
The scandal of poor epidemiological research: ***
Elm, E. V., & Egger, M. (2004) · BMJ · doi:10.1136/bmj.329.7471.868 · Cited by 134 · cited 1×
“On the other hand, evidence-based medicine focuses on determining whether specific medical hypotheses are true or false by aggregating data from multiple studies and performing meta-analyses (von Elm and Egger, 2004; Altman, 2002).”
Why Most Published Research Findings Are False: Problems in the Analysis
Goodman, S., & Greenland, S. (2007) · PLoS Medicine · doi:10.1371/journal.pmed.0040168 · Cited by 109 · cited 2×
“Submission data are available only for the years 2006–2010 for The Lancet and the years 2003–2010 for The BMJ. misquantified (Goodman and Greenland, 2007).”
Why we need confidence intervals
Altman, D. G. (2005) · World Journal of Surgery · doi:10.1007/s00268-005-7911-0 · Cited by 88 · cited 2×
“It is also well established that reporting measures of scientific uncertainty such as confidence intervals are critical for appropriate interpretation of published results (Altman, 2005).”
Ioannidis, J. P. A. (2014) · Biostatistics · doi:10.1093/biostatistics/kxt036 · Cited by 60 · cited 10×
“By using web-based version control, we were able to correct a minor typo identified by a discussant and push the changes to the repository quickly (Ioannidis, 2013).”
InfluentialBenjamini, Y., & Hechtlinger, Y. (2014) · Biostatistics · doi:10.1093/biostatistics/kxt032 · Cited by 54 · cited 10×
“primary versus secondary) (Ioannidis, 2013; Gelman and O’Rourke, 2013; Goodman, 2013; Benjamini and Hechtlinger, 2013).”
InfluentialGelman, A., & O'Rourke, K. (2014) · Biostatistics · doi:10.1093/biostatistics/kxt034 · Cited by 48
Poor-Quality Medical Research: What Can Journals Do?
Altman, D. G. (2002) · JAMA · doi:10.1001/jama.287.21.2765 · Cited by 34 · cited 1×
“On the other hand, evidence-based medicine focuses on determining whether specific medical hypotheses are true or false by aggregating data from multiple studies and performing meta-analyses (von Elm and Egger, 2004; Altman, 2002).”
Goodman, S. N. (2013) · Biostatistics · doi:10.1093/biostatistics/kxt035 · Cited by 27 · cited 9×
“We would therefore like to thank the discussants at Biostatistics (Cox, 2013; Gelman and O’Rourke, 2013; Schuemie and others, 2013; Benjamini and Hechtlinger, 2013; Goodman, 2013; Ioannidis, 2013) andworldwidewho have also brought rational, empirical reasoning to bear on this important topic, and provided outstanding…”
InfluentialLarge-Scale Assessment of the Effect of Popularity on the Reliability of Research
Pfeiffer, T., & Hoffmann, R. (2009) · PLoS ONE · doi:10.1371/journal.pone.0005996 · Cited by 26 · cited 2×
“interest in this question has led to significant discussion in both the scientific literature (Ioannidis, 2005; Goodman and Greenland, 2007; Pfeiffer and Hoffmann, 2009) and the popular press (Freedman, 2010; Staff, 2007; Hotz, 2007).”
The Joint Null Criterion for Multiple Hypothesis Tests
Leek, J. T., & Storey, J. D. (2011) · Statistical Applications in Genetics and Molecular Biology · doi:10.2202/1544-6115.1673 · Cited by 25 · cited 2×
“…distribution, but statistical research has shown that when considered as one group, the distribution of true alternative P-values can be modeled as a Beta distribution, Beta(a, b) or mixtures of Beta distributions (Allison and others, 2002, 2006; Pounds and Morris, 2003; Leek and Storey, 2011).”
Self-citationSchuemie, M. J., Ryan, P. B., Suchard, M. A., Shahn, Z., & Madigan, D. (2013) · Biostatistics · doi:10.1093/biostatistics/kxt037 · Cited by 18
Fletcher, R. H. (1989) · JAMA · doi:10.1001/jama.1989.03430130118044 · Cited by 13 · cited 1×
“As Cox (2013) points out in his discussion, there have been relatively few published true replications before whole genome association studies, likely due to the cost/benefit tradeoff of performing clinical studies (Fletcher, 1989).”
Discussion: Comment on a paper by Jager and Leek
Cox, D. R. (2013) · Biostatistics · doi:10.1093/biostatistics/kxt033 · Cited by 9 · cited 1×
“FARRINGTON, C. P. (1995). Relative incidence estimation from case series for vaccine safety evaluation.”
Comparing Several Methods to Fit Finite Mixture Models to Grouped Data by the EM Algorithm
Wengrzik, J., & Timm, J. (2011) · Cited by 5
Research works printed but not linked: 4
- 1989, Journal of Political Economy, 100, 1257
- 2011, STATISTICAL APPLICATIONS IN GENETICS AND MOLECULAR BIOLOGY, 10, 28
- 2012, PNAS, 334, 1226
- 2011, PROCEEDINGS OF THE WORLD CONGRESS ON ENGINEERING, 1, 1
Cited by
The Extent and Consequences of P-Hacking in Science
Head, M. L., Holman, L., Lanfear, R., Kahn, A. T., & Jennions, M. D. (2015) · PLOS Biology · doi:10.1371/journal.pbio.1002106 · Cited by 1,388
“We chose p 0.05 as the cutoff for our upper bin (following [3]), rather than p = 0.05 (see [46]) because we suspect that many authors do not regard p = 0.05 as significant.”
BackgroundMethodInfluentialDonoho, D. (2017) · Journal of Computational and Graphical Statistics · doi:10.1080/10618600.2017.1384734 · Cited by 695
Why, When and How to Adjust Your P Values?
Jafari, M., & Ansari‐Pour, N. (2019) · PubMed · doi:10.22074/cellj.2019.5992 · Cited by 468
Detecting and avoiding likely false‐positive findings – a practical guide
Forstmeier, W., Wagenmakers, E., & Parker, T. H. (2017) · Biological Reviews · doi:10.1111/brv.12315 · Cited by 436
Is science really facing a reproducibility crisis, and do we need it to?
Fanelli, D. (2018) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.1708272114 · Cited by 416
Metadata
- Authors
- Jager, L. R., & Leek, J. T.
- Affiliations
- United States Naval Academy (US)
- Johns Hopkins University (US)
- Corresponding author
- Leek, J. T.
- Venue
- Biostatistics
- Publisher
- Oxford University Press (OUP)
- Published
- 2013-09-25
- Volume
- 15
- Issue
- 1
- Pages
- 1-12
- ISSN
- 1465-4644 (ISSN-L)
- 1468-4357
- OpenAlex
- W2098523432
- Cited by
- 179
- References
- 48
- Citations by year
- 2026: 2 · 2025: 8 · 2024: 9 · 2023: 15 · 2022: 10 · 2021: 12 · 2020: 12 · 2019: 11 · 2018: 23 · 2017: 17 · 2016: 34 · 2015: 17 · 2014: 8 · 2013: 1
- FWCI
- 22.04
- Type
- journal article
- Language
- en
- Open access
- closed
- Locations
- Biostatistics (publishedVersion)
- PubMed (publishedVersion)
- Funders
- Johns Hopkins University
- Topics
- Social Sciences › Decision Sciences › Statistics, Probability and Uncertainty › Meta-analysis and systematic reviews
- Social Sciences › Library and Information Sciences › scientometrics and bibliometrics research
- Social Sciences › Library and Information Sciences › Academic Publishing and Open Access
- Keywords
- false discovery rate, false positives, p-values, journal abstracts
- MeSH terms
- Algorithms, Computer Simulation, Data Interpretation, Statistical, False Positive Reactions, United Kingdom, Humans, Publications, Software, United States, Biomedical Research
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.