| Publication Abstract Display | | Type: Published Abstract | | Title: Development and validation of a symptom score for acute HIV infection in San Diego. | | Authors: Lin T, Anderson C, Tenenbaum T, Gianella S, Little S, Hoenigl M | | Year: | | Publication: CROI 2018 | | Volume: Issue: Pages: | | Abstract:Background
Diagnosis and treatment of acute HIV infection (AHI) is critical to decreasing transmission and preserving immune function. While behavioral risk-based scores have been developed to identify those most likely to have AHI, a symptom-based score has not yet been developed. The objective of this study was to develop and validate a symptom-based score with fair accuracy [receiver operating characteristic (ROC) area under the curve (AUC) >0.7] for AHI.
Methods
Adults who tested positive for AHI from 2007 to 2017 and who tested negative for HIV in 2017 with the Early Test community-based HIV testing program San Diego were assessed for 11 symptoms present during the 14 days prior to testing. The study sample was retrospectively randomized 2:1 into a derivation and validation dataset. In the derivation dataset, symptoms significant for AHI in univariate logistic regression models were entered into a multivariate model. Significant symptoms in the multivariate model were assigned a score equivalent to its odds ratio rounded to the nearest integer. In the validation dataset, a summed symptom score was assessed with ROC. An optimum cut-off score was determined using Youden’s index.
Results
A total of 1003 participants (738 men who have sex with men (MSM), 151 non-MSM men, 111 ciswomen, 2 transwomen, 1 declined to disclose gender) were included in the analysis. There were 114 AHI cases, including 109 MSM, 1 ciswoman and 1 transwoman. Median age of AHI cases was similar to that of HIV-negative cases [32 (IQR 25-42) vs 33 (27-43), p = 0.11]. AHI cases reported a greater number of symptoms than HIV-negative cases [4 (IQR 2-6) vs. 0 (0-1), p<0.001]. In the derivation dataset, all 11 symptoms were included in the multivariate model. Myalgia [OR 7.9 (95%CI 3.3-18.7)], fever [10.9 (4.6-26.1)], and weight loss [4.1(1.1-15.1)] remained significant. In the validation dataset, the weighted symptom score yielded a ROC curve with AUC of 0.85 (95%CI 0.77 to 0.93). An optimal cut-off score of >=9 was 71% sensitive, 97% specific, with PPV of 76%, NPV of 96%, accuracy of 93%, and diagnostic odds ratio of 65.1 (95%CI 26.4 to 160.8).
Conclusions
Use of a symptom-based score to identify individuals with AHI is novel, fairly accurate, and may inform allocation of diagnostic and treatment resources. In theory, a symptom-based score would be more generalizable than risk behavior-based scores; further validation in populations with different demographics and risk behavior patterns is needed. |
return to publications listing
|