A Retroactive Study on Factors Influencing the Efficacy of Treatment for Tuberculosis Patients with HIV: based on the data from 2010 to 2020 in Shanghai, China

Abstract

At present, the factors influencing Tuberculosis (TB) treatment effectiveness in HIV/TB co-infected patients need to be supported by more substantial real-world evidence. A retrospective study is conducted to fill the vacancy. 461 TB patients with HIV are defined as 742 samples according to each TB detection period. 7788 valid treatment records corresponding to 17 drug compositions for TB and 150 clinical indicators with more than 100 records are used to conduct data mining with consensus clustering, Fisher’s exact test, stratified analysis, and three modeling approaches, including logistic regression, support vector machine, and random forest. We find that A CD4+T cell count of 42 cells per μL may serve as a sensitive classification standard for the immune level to assist in evaluating or predicting the efficacy of TB (P=0.007); Rifabutin and levofloxacin alone or in combination may be more effective than other first- and second-line anti-TB agents in combination (P=0.037); Samples with low immune levels (CD4≦42) may be more resistant to first-line TB drugs (P=0.049); Age (P=0.015), bicarbonate radical (P=0.007), high-density lipoprotein cholesterol (P=0.026), pre-treatment CD8+T cell count (P=0.015, age<60, male), neutrophil percentage (P=0.033, age<60), rifabutin (P=0.010, age<60), and cycloserine (P=0.027, age<60) may influence the TB treatment effectiveness; More evidence is needed to support the relationship between pre-treatment clinical indicators or drug regimens and TB treatment effectiveness (The best AUC is 0.560∼0.763); The percentage of lymphocytes (P=0.028) can be used as an effective TB therapeutic target. These perspectives supplement knowledge in relevant clinical aspects.