Hi all, here is a new post on methods and statistical reporting in a manuscript. One of my mentors, Dr. Sundar Kumar, inspired me to write this post. For beginners like us, it isn’t very easy to understand and interpret our data analyses, which affect the way we report it in our manuscripts. Today, we will look at some of the common pitfalls while reporting methods and statistical analyses and ways to overcome them. There are multiple articles on this topic, and after going through some of them, I have tried to simplify it for you.
Main reasons for the rejection of a manuscript are methodological flaws, including statistical reporting. So here are some of the common errors in METHODS & RESULTS sections and ways to rectify them:
- Start with clearly reporting the aim and research questions in the introduction—state aim in the PICO (Population, Intervention, Comparators, and Outcome) format. We already know it but forget to implement it.
- Then specify the study design. Always make use of the available guidelines for study designs and try to include all items given in it. You can easily find a guideline that is relevant for your study design on the EQUATOR network (http://www.equator-network.org/). Additionally, you can use a guide for statistical analysis known as SAMPL (Statistical Analyses and Methods in the Published Literature.) I came across this one recently, and it is quite handy.
- Report if you have registered your trials or reviews previously in PROSPERO and Clinical trials registry. If there are any deviations from the original protocol and then report it in the manuscript.
- Write complete details for sample size calculation, including type I error, type II error, confidence interval, and minimum clinically relevant difference. In case of a pilot study where you have not computed a sample size, then explain it in the article.
- In case you have dropouts or missing data, then report the missing data, and justify if you have used any techniques to handle the missing data, especially in an RCT.
- Initially, I did not understand the importance of this, but, always provide a detailed description of the statistical analyses and a rationale for using a particular analysis. We usually report the test used but completely forget about the rationale for using it, such as assumptions or study hypotheses. Specify which variables were analyzed with each different statistical analysis. Statements such as ‘data were analyzed using chi-square test’ is considered vague.
- Also, provide details of statistical analysis such as specific analysis for a particular outcome and how the results are reported. E.g., distance walked is summarized using mean and SD. The results of logistic regression are summarized using the Odds ratio and confidence interval. These are a few details which we often omit.
- Again, I learned this hack recently to make sure that all the analyses mentioned in the methods section correspond with the set of results and vice versa. We unknowingly add a ‘new’ result that we forget to describe previously in the methods or objectives of the study.
- In the abstract, make sure that you have reported the same results as that given in the main text. Do not write additional/modified/different results that you have not mentioned in the main manuscript.
- While writing the methods section, write a separate detailed paragraph on outcomes. Provide details on measurement of the outcome, who measured it, when it was measured, and it was a primary or secondary outcome, units of measurement, tools for measurement, and its psychometric properties. A reviewer may ask you to remove these details during revision, but it is better to provide more information while submitting a manuscript for the first time.
- There are no studies without a limitation. Be very clear about your study limitations or biases or confounders. It pisses off the editors if you try to downplay your study limitations.
So, these were some of the common reporting problems that can be easily fixed in your manuscripts. A transparent and thorough reporting of methodology will always give you an edge for getting your paper accepted. I will write another post on choosing appropriate statistical analyses. You can use this post as a checklist before submitting your manuscripts. I hope it helps.
- Harhay MO, Donaldson GC. Guidance on Statistical Reporting to Help Improve Your Chances of a Favorable Statistical Review. Am J Respir Crit Care Med. 2020;201(9):1035-1038. doi:10.1164/rccm.202003-0477ED
- Swiss Med Wkly. 2015;145:w14076
- Lang T, Altman D. Statistical Analyses and Methods in the Published Literature: the SAMPL Guidelines