SOME INDICES OF DISEASE ACTIVITY IN RHEUMATOID ARTHRITIS

Mohamad  F.M. ZIADE

Rheumatoid Arthritis is a heterogeneous disease in which assessment is difficult. Measurement of disease inflammation is important when determining rheumatoid activity and the response to the drugs used in its treatment. Unfortunately no one measure, either clinical or laboratory, accurately reflects inflammatory activity. To address this problem 'Stoke index' and 'Mallaya and Mace index' were designed by clinicians to give a global measure of disease activity in Rheumatoid Arthritis. Both indices utilize a combination of chemical and physical inflammatory measures to derive an activity score. These indices have not been validated in terms of their reversibility and sensitivity before this study.

Results of seven demographic measures, seven laboratory variables and six clinical markers of disease activity on cohort of 371 rheumatoid patients have been used to verify the clinical ability of the indices to measure disease activity. A comparison, in this aspect, is carried out between the two indices. Variable selection routines using Principal Component Analysis were obtained to define a subset of the most influential variables in determining the disease activity. It demonstrated the superiority of Soke index in providing a useful index of global disease activity for use in the assessment of Rheumatoid Arthritis. A modification of the overall Stoke index is suggested.

Different statistical procedures are implemented to construct new ways of allocating treatments. Discriminate Analysis is used to construct linear functions based on our data set, which may be used for allocating new patients. A new index based on z‑scores was introduced. Different subsets of variables were used to obtain different indices. The usage of two indices, one for  the chemical aspects of the disease and another for the physical aspects, rather than the traditional way of using one index is introduced and discussed. A set of 49 patients was chosen as a test set to evaluate the new indices.

Finally, predicting the improvers and non‑improvers after a period of treatment were obtained by using Linear Logistic Regression, firstly by using the chemical and physical variables and secondly by using the demographic variables. A simplified model for expecting the possible response of each patient using the demographic variables was obtained for the first time. The implementation of the new allocation methods in the practical life is discussed and some further suggestions and recommendations are presented.