The DOI matches its Crossref record; 6 of the 6 references it cites were found in a registry or an index.
Response to Hartwig and Davies
Open accessAbstract
Thank you for the opportunity to respond to the thoughtful and timely letter by Hartwig and Davies.1 They do indeed raise a very important practical issue with the implementation of MR-Egger regression in the single-sample setting, or with the use of so called `internal’ weights, namely weak instrument bias. They are right to point out the unsatisfactory nature of our analysis of the height data in our original publication,2 and that our use of weak instruments had the likely effect of biasing the MR-Egger estimate towards that of the observational association. Their re-analysis of these data with external weights appears to provide a much more satisfactory answer and, when such weights are available, it is both a simple and an attractive way to circumvent the problem. Although previous simulation studies have highlighted this fact, further research is needed to completely understand the issue of weak instrument bias for MR-Egger in the single-sample context. What is clear however, is that the standard notion of instrument strength, as quantified by the F statistic, cannot naively be applied to estimate the magnitude of this bias; new (or at least newly borrowed) theory is required. Before covering initial progress in this vein, we now briefly discuss related (and more mature) work in the two-sample context. A strength of MR-Egger regression, along with the weighted median3 and inverse-variance weighted (IVW)4 methods, is that they do not require gene, exposure and outcome data on a single sample of subjects at the individual level. MR-Egger regression can be implemented with only summary data estimates of the single nucleotide polymorphism (SNP)-exposure and SNP-outcome associations, making it most natural to use within the two-sample summary data context. This is becoming an increasingly popular way of performing Mendelian randomization investigations and, in recent work to appear in this journal,5 we extensively explore the issue of weak instrument bias for MR-Egger in this simpler setting. We show that weak instruments (with small F statistics) have the effect of inducing regression dilution bias into the MR-Egger estimate, thus shrinking its value towards the null (although such bias does not inflate Type 1 error rates when testing for a causal effect). This same phenomenon has also been shown for the IVW estimate.6,7 However, whereas the magnitude of the attenuation in the IVW estimate can be gauged from the instrument’s F statistics (e.g. a mean F of 20 would indicate an approximate dilution of 1/20 = 5%), this will almost certainly underestimate the dilution for MR-Egger. A new statistic - which is a simple modification of Higgins’ I2 - can be used to quantify the dilution (we call it IGX2). It lies between 0 and 100%, and an IGX2of 95% would indicate an expected dilution of 5%. We also describe how the established method of simulation extrapolation (SIMEX)8 can be used to calculate a bias-adjusted MR-Egger estimate to mitigate the effect of this dilution. The first author of this letter is currently conducting a theoretical investigation of the weak instrument bias of MR-Egger regression in the single-sample setting, with particular focus on comparing its performance to that of the IVW estimate. Initial results highlight the following: (i) when all genetic instruments are valid, both MR-Egger and IVW estimates are consistent, but the MR-Egger estimate will always be more strongly affected by weak instrument bias (towards the observational association) than the IVW estimate; and (ii) when some or all of the genetic instruments are invalid due to directional pleiotropy, but the sample covariance of the pleiotropy and instrument strength terms is zero (e.g. the InSIDE assumption in Bowden et al.2 holds), then MR-Egger is a consistent estimate of the causal effect whereas the IVW estimate is asymptotically biased. However, the finite-sample bias of the IVW estimate may still be less than that of MR-Egger if the genetic variants are sufficiently weak. Research is under way to investigate whether a suitable statistic, perhaps analogous to IGX2, can be derived to quantify the likely magnitude of dilution of MR-Egger regression in the single-sample setting. An obvious follow-on question will be whether the SIMEX approach outlined in this journal5 can also be transferred to this setting to yield a bias-adjusted estimate. This is not trivial because such an implementation is likely to require at least some a priori knowledge of the strength of correlation between the standard errors of the SNP-exposure and SNP-outcome association estimates due to shared confounding (in the two-sample context we assume a priori this correlation is zero). Theoretical work on this topic already exists9 and will no doubt serve as a useful starting point. Even in cases which appear superficially to be two-sample settings, sample overlap may mean that the bias due to weak instruments is more similar to the single-sample setting.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
Bowden, J., Davey Smith, G., & Burgess, S. (2015) · International Journal of Epidemiology · doi:10.1093/ije/dyv080 · Cited by 9,816 · cited 2×
“E-mail: Dave.Evans@bristol.ac.uk In their letter, Hartwig and Davies1 raise an important issue that was not discussed in the original Mendelian randomization Egger regression (MR Egger) paper by Bowden et al.2 Hartwig and Davies point out that, similar to other varieties of MR, MR-Egger is also susceptible to weak ins…”
Self-citationBowden, J., Davey Smith, G., Haycock, P. C., & Burgess, S. (2016) · Genetic Epidemiology · doi:10.1002/gepi.21965 · Cited by 8,922
Self-citationMendelian Randomization Analysis With Multiple Genetic Variants Using Summarized Data
Burgess, S., Butterworth, A., & Thompson, S. G. (2013) · Genetic Epidemiology · doi:10.1002/gepi.21758 · Cited by 6,050
Self-citationGenetic studies of body mass index yield new insights for obesity biology
Study, T. L. C., Locke, A. E., The ADIPOGen Consortium, The AGEN-BMI Working Group, The CARDIOGRAMplusC4D Consortium, The CKDGen Consortium, The, G. L. G. C., The, I. C. B. P., The MAGIC Investigators, The MuTHER Consortium, et al. (2015) · Nature · doi:10.1038/nature14177 · Cited by 4,534
Bias due to participant overlap in two‐sample Mendelian randomization
Burgess, S., Davies, N. M., & Thompson, S. G. (2016) · Genetic Epidemiology · doi:10.1002/gepi.21998 · Cited by 1,624
Self-citationPierce, B. L., & Burgess, S. (2013) · American Journal of Epidemiology · doi:10.1093/aje/kwt084 · Cited by 1,389
Self-citation
Research works printed but not linked: 6
- 10.1093/ije/dyw240 · DOI
- Bowden J Del Greco FM Minelli C . Assessing the validity of MR-Egger regression for two-sample Mendelian randomization: the role of the I 2 statistic. Int J Epidemiol 2016; doi: 10.1093/ije/dyw220. · DOI · Paper page
- Angrist, 1995, Split-sample instrumental variables estimates of the return to schooling, J Bus Econ Stat, 13, 225, 10.1080/07350015.1995.10524597 · DOI · Paper page
- 10.2307/2290994 · DOI
- 10.1007/s10182-008-0086-x · doi:10.1007/s10182-008-0086-x
- 10.1038/ng.2797 · DOI
Cited by
Verbanck, M., Chen, C.-Y., Neale, B., & Do, R. (2018) · Nature Genetics · doi:10.1038/s41588-018-0099-7 · Cited by 8,591
Mapping the human genetic architecture of COVID-19
Initiative, C.-1. H. G., InitiativeLeadership, C.-1. H. G., Niemi, M. E. K., Karjalainen, J., Liao, R. G., Neale, B. M., Daly, M., Ganna, A., Writing group, Writing group leaders, et al. (2021) · Nature · doi:10.1038/s41586-021-03767-x · Cited by 983
Jones, S. E., Lane, J. M., Wood, A. R., van Hees, V. T., Tyrrell, J., Beaumont, R. N., Jeffries, A. R., Dashti, H. S., Hillsdon, M., Ruth, K. S., et al. (2019) · Nature Communications · doi:10.1038/s41467-018-08259-7 · Cited by 691
Self-citationMendelian randomization in cardiometabolic disease: challenges in evaluating causality
Holmes, M. V., Ala-Korpela, M., & Smith, G. D. (2017) · Nature Reviews Cardiology · doi:10.1038/nrcardio.2017.78 · Cited by 679
Self-citationthe 23andMe Research Team, Pasman, J. A., The Substance Use Disorders Working Group of the Psychiatric Genomics Consortium, International Cannabis Consortium, Verweij, K. J. H., Gerring, Z., Stringer, S., Sanchez-Roige, S., Treur, J. L., Abdellaoui, A., et al. (2018) · Nature Neuroscience · doi:10.1038/s41593-018-0206-1 · Cited by 498
Metadata
- Authors
- Bowden, J., Burgess, S., & Davey Smith, G.
- Venue
- International Journal of Epidemiology
- Publisher
- Oxford University Press (OUP)
- Published
- 2016-09-20
- Volume
- 45
- Issue
- 5
- Pages
- 1679-1680
- ISSN
- 0300-5771 (ISSN-L)
- 1464-3685
- OpenAlex
- W2521215217
- Cited by
- 174
- References
- 6
- FWCI
- 6.85
- Type
- journal article
- Language
- en
- Open access
- hybrid
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
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: Response to Hartwig and Davies. https://app.cochat.ai/papers/10.1093/ije/dyw252
- The paper
- Bowden, J., Burgess, S., & Davey Smith, G. (2016). Response to Hartwig and Davies. International Journal of Epidemiology. https://doi.org/10.1093/ije/dyw252
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.