Background: Mortality in hospitalised patients with HIV-associated tuberculosis remains high. Cytomegalovirus (CMV) organ disease is one of the co-infections found in autopsies of such patients. We investigated the association of CMV viraemia with mortality in this setting.
Methods: HIV-infected inpatients in Khayelitsha Hospital with CD4< 350cells/mL and new diagnosis of tuberculosis were enrolled from January 2014-June 2015. Plasma CMV qPCR was performed and categorized as detectable (CMV+) or undetectable (CMV-). Endpoint was 12-week mortality.
Results: We included 256 patients with median age 36 years (IQR 31-44 years), 49% male, 35% on ART, median CD4=64 cells/mL(IQR 24-117) and 79(30.9%) CMV+. By 12 weeks, 26/77(38.0%) of CMV+ and 31/174(17.8%) of CMV- patients died (p=0.008); 5 were lost to follow-up. In CMV+ patients with < 1000 copies/ml mortality was 12/36(33.3%) compared to 14/41(34.1%) in those with higher viral load (p=1.0).
Mortality was higher in older patients (≥36 years): 32.8% vs. 14.1%(p< 0.001). Older patients were more likely to be CMV+ (38.0% vs. 23.6%, p=0.015) and a larger proportion of older patients had CD4 count< 50 cells/mL (48.5% vs. 37.9%, p=0.106). In Kaplan-Meier analysis, CMV+ was associated with mortality in older, not younger patients.
In multivariate Cox proportional-hazards regression, age (aHR=1.70, 95%CI=1.34-2.15 per 10years increase) was associated with mortality; CMV status was not.

Figure: Survival by CMV status: whole cohort and stratified by age
[Figure: Survival by CMV status: whole cohort and stratified by age ]

 UnivariateMultivariate: n=232 with complete observations
 HRlower 95%CIupper 95%CIp.valueaHRlower 95%CIupper 95%CIp.value
Age, per 10 years1,751,412,19<0.0011,701,342,15<0.001
Male sex0,840,501,410,5050,690,411,190,185
CD4, per 50 cellsx10^6/L0,780,640,960,0170,800,641,000,052
HIV viral load, Log10 copies/mL0,900,801,030,1170,940,821,070,316
Albumin, per 5 g/L0,730,580,920,0060,840,661,080,177
CMV viraemia2,091,243,530,0041,670,952,930,077
Likelihood ratio test= 37.2 on 7 d.f., p<0.0001
[Table: Cox proportional hazards regression analysis of factors associated with 12-week mortality]

Conclusions: CMV viraemia was associated with higher mortality, but not after adjusting for potential confounders. Older patients had higher mortality and were more likely to have CMV viraemia. CMV viraemia is likely a marker of more severe immunodeficiency rather than a direct contributor to mortality.