Development of Non-Genetic Risk Stratification Model of Severe Allopurinol Hypersensitivity (NoG-ALLOH Score): A Multicenter Study in Thailand

Abstract

Background The side effect of allopurinol was severe allopurinol hypersensitivity (SAH). There are two major causes including genetic risk factor, HLA-B*58:01 and non-genetic risk factors including female, old age, renal impairment, inadequate starting dosage of allopurinol and received diuretic. Although HLA-B*58:01 testing is recommended in Thai, the several problems including cost, limitation of the testing and the delay of the result. Therefore, we aimed to use these non-genetic risk factors for developing the model to predict SAH. Method The study was a retrospective observational incidence density sampling. SAH cases were collected from tertiary care medical center which some cases were referred from primary care medical center; where most non-SAH cases were collected from primary care hospitals and tertiary care medical centers. The data of non-genetic factors particularly sex, age, renal function, co-diuresis, starting dosage of allopurinol and serum uric acid (SUA) of non-SAH cases selected the first description of allopurinol. The binary prevalence-weighted logistic regression statistical method was used to develop the prediction models. Three models were developed following general practice. Result Totally, there were 209 cases of SAH and 23,068 cases of non-SAH. Factors that were associated with the development of SAH within 90 days were female, old age (¿ 65 years old), renal impairment, inadequate starting dosage of allopurinol, co-medication(s) with diuretic and high SUA before prescription of allopurinol. Model 1a and model 1b were applied for patients who did not have and have SUA when starting allopurinol, respectively. Model 2 was applied for patients who had all non-genetic risk factors and started allopurinol within 60 days but have not SAH. The area under the receiver operating characteristic curve for model 1a, model 1b and model 2 were 0.72, 0.81 and 0.82, respectively (Figure 1). The performance for each predictions SAH were good. Conclusion Model 1a and model 1b predict SAH for the patients who had their first prescribed allopurinol, model 2 predicts SAH for the patients who had been taken allopurinol within 60 days but no SAH. The scoring system of each model helps clinician to prescribe allopurinol in real clinical practice before the patients develop SAH. The score of 0-1%, 1-2% and 2-100% indicates the low, moderate and high risk, respectively. The low-risk group can start allopurinol. The moderate-risk group considers to start allopurinol with closed monitoring of SAH. The high-risk group suggests to change to other urate lowering agents for preventing SAH (Table 1).