Study design

We conducted a matched nested case-control study to assess the association between NMR and MI in people who use cigarettes with HIV. We used data from the Center for AIDS Research (CFAR) Network of Integrated Clinical Systems (CNICS) cohort. Initiated in January 1995, CNICS is a prospective clinical cohort of > 49,000 PWH receiving care at ten CFAR sites nationally [16]. Seven of the CNICS sites were included in these analyses: Case Western Reserve University, Johns Hopkins University, University of Alabama at Birmingham, University of California, San Diego, University of California, San Francisco, University of North Carolina at Chapel Hill, and University of Washington. CNICS captures clinical data and self-reported measures. Plasma samples in the CNICS have been stored at −80 C since its inception.

Participants were selected from individuals with self-reported regular cigarette use, defined as smoking on at least two consecutive time points, a minimum of one year apart, as well as at all assessed intervening time points. Additionally, they had to have aviremic plasma sample available for NMR testing. Cases included all individuals who had an MI and had a plasma sample when they reported cigarette use. Controls were selected using incidence density sampling. Controls were matched to the cases based on site, age, race, sex, viral load status (at time of MI), and calendar time (within the same 365-day period) when cases had MI. MI has been adjudicated in the CNICS cohort [17]. Individuals diagnosed with MI, which could not be confirmed or with a reported cardiovascular disease equivalent, which had not been validated in CNICS, were excluded as cases or controls.

Outcome and covariate measurement and definitions

Standard liquid chromatography-tandem mass spectrometry (LC-MS) was used to measure both cotinine and 3-HC [10]. A cotinine value > 10 ng/mL confirmed cigarette use [18]. The NMR was treated as a continuous measure, and the log of NMR was reported because the NMR was not normally distributed. We examined demographic characteristics, including age, race, ethnicity, and birth sex, clinical characteristics including CD4 count, history of hypertension, diabetes, and statin medication use. Given the association between the duration and level of exposure to HIV viremia with adverse outcomes, we estimated the HIV viremia copy years for each participant at the time of MI or censored using the trapezoid method applied to all available viral loads (See Supplementary Information for the trapezoid method) [19]. To account for the effect of efavirenz’s on the NMR, we calculated the NMR as a weighted average which was computed by summing the NMR on efavirenz x time on efavirenz + NMR off efavirenz x time off efavirenz divided by total cigarette use duration, which we assumed to be the individual’s current age minus 16 [5,15]. Cigarette use was imputed for all participants to have begun at age 16 [20]. We considered the pre-ART NMR to be the same as the non-efavirenz containing regimen NMR given the relatively small increases in NMR in individuals initiating non-efavirenz containing regimens [5].

Statistical analysis and sample size

Baseline characteristics were compared using chi-squared tests for categorical variables and t tests or rank sum tests for continuous variables. Normality was assessed using the Shapiro-Wilk and histograms. Conditional logistic regression was used to compare NMR between those with and without MI. Potential confounders were selected based on biological plausibility. We targeted a sample size of 145 cases with a goal of 2:1 control to cases for 80% power to detect a 0.1 value difference in NMR between cases and controls with a p-value of 0.05. Sensitivity analysis was conducted to assess the impact of excluding participants with missingness and test the assumptions used in the weighted NMR calculation.

Ethics statement

The study was reviewed and approved by the University of Pennsylvania and University of Toronto institutional review board. The required for written informed consent was waived by the Ethics Committee. All data and specimens were anonymized by CNICS and therefore, the authors had no access to identifiable personal. Data access for research purposes was obtained on August 01, 2019. Additional data were obtained afterward to ensure that the enrollment included individuals who had MIs in 2020.

Characteristics of cases and controls

We identified 135 participants (mean age 51; 76% male) with myocardial infarction (MI) occurring between 2003 and 2020. We identified 252 matched controls (mean age 50 years; 76% male) and were unable to find a suitable second matched control for 18 cases. Two controls samples were excluded due to missing NMR. There was little missingness in the data with diastolic and systolic pressure having <2% and total cholesterol with 22% missingness in cases and 23% in controls. Table 1 summarizes participants’ baseline demographic characteristics. Cases and controls did not differ significantly at baseline, except for cases having higher rates of hypertension, diabetes, and statin medication use, compared with controls.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Participant demographic characteristics.

https://doi.org/10.1371/journal.pone.0356296.t001

NMR and Myocardial infarction

Median (IQR) NMR was greater in those with MI [0.51 (0.36 0.73)] than controls [0.47 (0.30, 0.70)]. In the conditional logistic regression model, patients with an MI had higher NMR than controls, but the difference was not statistically significant (Table 2). Patients with an MI also had a greater history of statin use, hypertension, and diabetes than controls. Adjusting for these factors did not substantively change the NMR-MI relationship (OR: 1.4, 95% CI: 0.92–2.00).

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Conditional logistic regression analysis for factors associated with MI.

https://doi.org/10.1371/journal.pone.0356296.t002

The relationship between MI and NMR did not exhibit a simple and consistent pattern. Therefore, to further investigate this, we categorized NMR into four quartiles, with each group containing the same number of people and ranked from lowest to highest. The odds of having NMR in the third versus first quartile were 1.79 times higher among cases than controls (S1 Table), although not a statistically significant effect (OR: 1.79, 95% CI: 0.92–3.5).

Sensitivity analysis

To assess the robustness of our main findings, we conducted sensitivity analyses using an alternative definition of NMR (S3 Table). The effect size of NMR risk on MI remained similar (OR: 1.3, 95 CI: 0.9–1.9). Excluding participants missing total cholesterol did not change the study conclusions (S2 Table).