Exercise: Fisher's exact test?

Nuesch et al. (2002) studied responsiveness to treatment of hypertensive patients. Patients were classified as either responsive or non-responsive to treatment using blood pressure measurements. The authors compared the proportion of patients who were classed as responsive using clinic measurements of blood pressure with the proportion of patients who were classed as responsive using ambulatory blood pressure recording. The authors reported that:

‘Measurements of clinic blood pressure at the start of the study identified 43 patients who were responsive to antihypertensive treatment and 62 who were non-responsive. With the initial ambulatory blood pressure monitoring, we reclassified 12 of the non-responsive patients as responsive (Fisher's exact test of numbers of responsive and non-responsive patients as assessed by clinic blood pressure versus ambulatory blood pressure; P=0.127). Based on the results of ambulatory blood pressure, we classified 55 patients as responsive (control group) and 50 as non-responsive. Thus, about half of our patients taking two to four antihypertensive drugs fulfilled the criteria for treatment resistance.’

We can recreate the table as follows:

Clinic Ambulatory
Responder 43 55
Non-responder 62 50
Total 105 105


1. Why is Fisher’s exact test wrong here? What assumption is violated?

Check suggested answer 1.

2. What effect do you think this might have on the P value?

Check suggested answer 2.


Nuesch R, Schroeder K, Dieterle T, Martina B, Battegay E. (2001) Relation between insufficient response to antihypertensive treatment and poor compliance with treatment: a prospective case-control study. British Medical Journal 323, 142-146.

To Clinical Biostatistics index.

To Martin Bland's M.Sc. index.

To Martin Bland's home page.

This page maintained by Martin Bland.
Last updated: 10 August, 2006.

Back to top.