CD4 count and tuberculosis risk in HIV-positive adults not on ART: a systematic review and meta-analysis

Abstract

Background: CD4 cell count in adults with human immunodeficiency virus (HIV) infection (PLHIV) not receiving antiretroviral therapy (ART) influences tuberculosis (TB) risk. Despite widespread use in models informing resource allocation, this relationship has not been systematically reviewed.

Methods: We systematically searched MEDLINE, Aidsinfo, Cochrane review database and Google Scholar for reports in English describing TB incidence stratified by updated CD4 cell count in cohorts of HIV-positive adults (age ≥15 years) not on ART (PROSPERO protocol no: CRD42016048888). Among inclusion criteria were: reporting precision for TB incidence, repeated CD4 measurements, and TB incidence reported for those not on ART or monotherapy. Study quality was assessed via the Newcastle-Ottawa tool for cohort studies. A Bayesian hierarchical model was fitted to estimate the pooled factor increase in TB incidence with respect to CD4 cell count decrement.

Results: A total of 1,555 distinct records were identified from which 164 full text articles were obtained. Common reasons for exclusion of full texts were: no valid TB incidence, no repeat CD4 measurements, and not reporting TB incidence by ART status. The seven studies included reported on 1,206 TB cases among 41,271 individuals, with a typical median follow-up of four years. Studies were generally rated as moderate or high quality. Our meta-analysis estimated a 1.43 (95% credible interval: 1.16-1.88)-fold increase in TB incidence per 100 cells per mm³ decrease in CD4 cell count.

Discussion: Our analysis confirms previous estimates of exponential increase in TB incidence with declining CD4 cell count in adults, emphasizing the importance of early ART initiation to reduce TB risk in PLHIV.

Keywords: Antiretroviral therapy; CD4 cell count; Modeling; Opportunistic infections; TB.

Figure 1. PRISMA flow diagram of review process.

Figure 2. Incidence of tuberculosis in adults living with HIV (age ≥15 years) not on antiretroviral therapy, by study and CD4-positive lymphocyte count.

Study data reported as incidence are plotted as triangles with confidence intervals; study data reported as hazard ratios are reported as dots, with confidence intervals except for the reference category. Note: the rightmost data point for Wolday, 2003 had zero TB cases which is not defined on a logarithmic scale—only the top of the Poisson exact confidence interval is shown, with the bottom truncated.

Figure 3. Forest plot of the rate of increase logarithmic tuberculosis incidence with CD4-positive lymphocyte count in adults living with HIV (age ≥15 years) not on antiretroviral therapy.

Summary and individual study measures are based on the posteriors from the hierarchical meta-analysis.

Figure 4. Increase in relative risk of tuberculosis incidence in adults living with HIV (age ≥15 years) not on antiretroviral therapy by CD4-positive lymphocyte count.

Thick dashed lines represent 95% credible intervals around the point estimate (thick solid line); horizontal lines represent means over depicted CD4 categories; dotted lines represent 95% credible intervals for these category means; the horizontal dashed line represents an incidence rate ratio of 1 (no change). It is assumed that individuals have a CD4 count of 1,000 cells/mm³ at the point of HIV infection.