Introduction

Use of isoniazid monotherapy (H) or rifampicin/isoniazid combination (RH) is an established strategy in the prevention of tuberculosis (TB) in individuals with latent infection1. We have previously found that patient choice of regimen is an important consideration in management2 but the concept of choice may not be applicable in all circumstances, e.g. child populations. Homogeneity analysis stems from the idea that if different variables are measuring the same concept, the fundamental characteristics would ideally be the same3. Multiple homogeneity analysis employing HOMALS program (SPSS®) allows graphical representation of categorical data using more than two variables, permitting a multivariate approach to data exploration.

Method

Data for TB patients receiving chemoprophylaxis between 1998 and 2002 in Homerton University Hospital were analysed retrospectively (n=560, excluding those who developed TB disease). Before 1st April 2000, drug regimen was decided by clinician but thereafter patients were allowed choice between daily 6-month H regimen and 3-month RH. Completion outcomes were defined by those who attended all outpatient clinics with evidence of tablet taking (success), incomplete attendance or negative urine tests explained by patient with sufficient medication supply (potential success), or neither of the above (failure). Relationships between variables (regimen, choice of regimen and outcome) were explored in a) the whole sample and b) patients over 16 years.

Results

Figure 1 demonstrates that locality of variable categories in relation to each other reflects association between categories, as previously described2. Differences are observed when the child group is excluded, in that greater clustering occurs for the three variable categories: (1) no choice; (2) H regimen; (3) failure outcome. These results are consolidated quantitatively (see Table 1) noting that statistical significance is reached when the child population is excluded.

^b Excluding patients where change of regimen occurred.

Discussion

These results demonstrate that HOMALS is an effective and appropriate multivariate technique for informing further analysis of categorical data. Choice of regimen in this context appears to be an adult concept inferring that concordance is more easily addressed in this subset and raises the issue of how best to target the child population.

References
1. Control and prevention of tuberculosis in the United Kingdom. Thorax 2000;55:887-901.
2. Rennie TW, Bates IP, Engova D, Bothamley G H. Exploring choice of regimen in TB prevention. Proceedings UKCPA Nov 2003, 41-42.
3. Gifi A. Homogeneity Analysis. In Gifi A, ed. Nonlinear Multivariate Analysis, pp 81-149. Chichester: John Wiley & Sons, 1991.