Q 1. I am not clear about the difference between odds ratios and relative risks and indeed, I am not sure how to decide which is the most appropriate measure to use when comparing levels of risk. How can I find out more?

A.  You may wish to start with a primer for the uninitiated and then take a look at a very comprehensive but lucid account of the issues involved in choosing the correct measure, as well as examples on how to perform the relevant calculations. As a further reference on appropriate methodology, the BMJ letter When can odds ratios mislead? is also most instructive.

Q 2. What is meant by the Number Need to Treat (NNT) and how is it related to risk?

A. To get a good intuitive grasp of what is meant by the NNT and why it is useful, take a look at the BMJ article The number needed to treat: a clinically useful measure of treatment effect. You may also wish to have a look at a general overview of risk measurements and the NNT, where relevant formulae are provided.

Q. 3. I am keen to obtain odds ratios and relative risks using SPSS.  I would also like to obtain 95% confidence intervals for these measures. Can you point me to an appropriate tutorial?

A. Certainly; please refer to Q. 2 and the corresponding solution within the FAQ section Hypothesis Tests for Categorical Data, where within the contest of testing for an association between two categorical variables – e.g. gender and choice of drug (new or existing) – you will find detailed worked examples, where it is assumed that you do not have background knowledge in this area. You will also find exemplars on how to present your findings in a reader-friendly fashion within the context of a write-up, such as an dissertation report or a publication at a later stage of your career.

Q 4.  I have been using Minitab to carry out the chi-square test of association and am interested in obtaining odds ratios for 2 x 2 cases.  The relevant instructions do not appear to be forthcoming from the dialogue box for the above test. Is there an alternative way in to obtaining the odds ratios I require.

A. Yes; you should use the instructions in the worked example provided under Calculating the odds ratio and confidence interval for the odds ratio for a 2 x 2 contingency table .

. Q 5. I am currently reading a paper reporting a study designed to compare likelihoods of receipt of IV rembinant tisue plasmoinogen activator (rtPA) treatment for stroke in eligible men and women. The study reports use of a z-test to compare the corresponding unadjusted odds ratios across subgroups of patients as defined by a range of factors, including time period within which study took place. How can I find out more about how the above test should be conducted?

A. The required mathematical calculations are provided at Test for a difference in two odds ratios, where you can also find advice on how to derive the corresponding p-value in Excel. The content at the above link is supported with the supplementary material below to support understanding.

Note, however, that within the context of the above type of meta-analysis, there is not a strong case for taking this approach instead of simply confirming if the 95% CIs for the odds ratios don’t overlap, except that if there are multiple comparisons to carry out, the introduction of correction for spurious statistical signficance based on sampling variation can be addressed by a correction such as the Bonferonni correction. For basic step-by-step instructions on how to perform this correction, refer to Q. 3 on the StatsforMedics page HYPOTHESIS TESTS FOR CATEGORICAL DATA.

Guidance notes for use of the resource Test for a difference in two odds ratios referred to above
In the resource  the terms ‘n’₁, ‘n’₂, ‘n’₃ and ‘n’₄ refer to the four frequencies that appear in a 2 x 2 table (ignoring the totals) for calcuating odds ratios. These frequencies are often labelled as ‘a’, ‘b’, ‘c’ and ‘d’. The remaining notation in the resource is clearly defined.

Odds Ratios, Relative Risks and the Number Needed to Treat by Margaret MacDougall is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.