Introduction
The purpose of this study was to investigate the usefulness and relationships between different measures of compliance among tuberculoses patients in real world context: the study focussed on patients with a diagnosis of TB registered with the Sindh District Tuberculosis Programme, Pakistan. Most of the registered patients on this programme were utilising WHO recommended DOTS (directly observed treatment � short course).

Method and sample
165 patients were sequentially recruited from a cohort of about 1000 registered TB patients in the Sindh district. In the absence of a "gold standard" compliance measure, this prospective study examined the patient sample using the following indicators of compliance; pill counts, urine tests, face interview (including use of the Morisky¹ scale) and clinic attendance. A questionnaire incorporating these measures was implemented by TB paramedic staff to patients attending clinic and followed up at the next appointed clinic. Unannounced house spot checks were also used on the sample.

Results
Urine tests measured compliance to be 88.8%, 93.2% and 93.0% in first and second clinics and spot check respectively (n=165). Pill counts measured lower compliance of 65.2% and 49.2% for clinics and spot check respectively. Compliance values from 53.4% to 70.9% using pill counts were observed for DOTS patients compared with 35% (range 33.4 � 37.5%) for non-DOT patients. Compliance measured by urine tests was around 95% for combination formulations compared with 86% for separate formulations of the same drugs. The results suggest that differences exist between the measures tested; patient interview was sound to be of limited use in this setting, probably because of cultural influences, similarly little value can be placed on clinic attendance data. Urine testing highlighted fewer non-compliers than did pill counts and appeared to be a more specific measure. Patients who were absolutely non-compliant according to urine tests were also non-compliant by pill counts but did not exhibit overly poor attendance.

Discussion
There are a number of factors which may have biased the results, however unavoidable. In retrospect, some indicators were more practical than others; in conjunction with tablet overages, pill count was probably the most viable to implement and urine testing was easiest to perform, although there were greater constraints on the latter. Clinic attendance rates would need to have been followed for a much longer time duration to reflect compliance to therapy. Follow-up studies are now being conducted to investigate relationships between clinical outcome and compliance measures in order to determine which is the most applicable or true reflection of compliance in this setting. In summary, differences in compliance as measured by urine tests and pill counts were found between prescribed formulations. Patients on DOTS also appeared more compliant.

Reference

1. Morisky, DE. et al. Concurrent and predictive validity of a self-reported measure of medication adherence. Med Care, 1986; 24: 67-74.