Kidney Function and the Risk of Cardiovascular Events in HIV-1 Infected Patients

Abstract

Objective:

Cardiovascular events are a significant cause of mortality in HIV/AIDS patients. The objective is to determine the correlation between kidney function and the risk of cardiovascular events in the HIV-infected population.

Design:

Nested, matched, case-control study design was employed.

Methods:

We performed a single-center study of 315 HIV-infected patients (63 cases who had cardiovascular events and 252 controls). Estimated glomerular filtration rate (eGFR), calculated by the CKD-EPI and the MDRD equation, and proteinuria were the primary exposures of interest.

Results:

Mean eGFR was significantly lower in the cases compared to controls (68·4 vs. 103·2 ml/min/1·73m², p<0·001 by CKD-Epi and 69·0 vs. 103·1 ml/min/1·73m², p<0·001 by MDRD). In univariate analysis, an eGFR of <60 ml/min/1·73m² was associated with a 15·9 fold increased odds of a cardiovascular event compared to an eGFR ≥ 60 ml/min/1·73m² (p<0·001). In multivariate analysis, a 10 ml/min/1·73m² decrease in eGFR was associated with a 20% increased odds of a cardiovascular event (odds ratio 1·2, 95% CI 1·1–1·4). The prevalence of proteinuria in the cases was approximately twice that of controls (51% vs. 25%, p<0·001). Proteinuria was associated with cardiovascular events both in univariate and multivariate analyses (OR of 3·6, 95% CI 1·9–7·0 and 2·2, 95% CI 1·1–4·8 respectively). Traditional cardiovascular risk factors like history of previous cardiovascular events, diabetes mellitus, and dyslipidemia along with low CD4 counts were also found as significant predictors of risk of cardiovascular events.

Conclusion:

Our study shows a significant independent association between decreased kidney function and increased risk of CVE in HIV-1-infected patients.

Introduction

There were about 571,378 people living with HIV/AIDS in the United States at the end of 2007 with 42,655 (21·1 per 100,000 population) new cases reported in that year [1]. With the introduction of highly active antiretroviral therapy (HAART), the proportion of deaths due to infectious causes in HIV/AIDS patients has declined from 80% to 43·6%, and a higher proportion of deaths has been attributed to non-infectious causes, with cardiovascular disease causing 21·8% of deaths in the HAART era, compared to 8·4% of deaths in the pre-HAART era [2].

In HIV-infected patients, contributors to cardiovascular disease may include traditional cardiovascular risk factors (e.g., age, sex, diabetes, hypertension, cigarette smoking, and hypercholesterolemia), direct or indirect effects of HIV infection itself (including inflammation and immune activation), or adverse effects of HIV therapy (which may be partially mediated by changes in traditional risk factors) [3-6]. In non-HIV-infected persons, an extensive body of literature demonstrates strong associations between reduced estimated glomerular filtration rate (eGFR) or proteinuria and subsequent cardiovascular events and mortality [7-10]. Despite the fact that chronic kidney disease (CKD) is highly prevalent in HIV-infected patients in the U.S. [11,12], the potential contribution of CKD to cardiovascular disease risk has been little studied and under-appreciated. For example, in the Data Collection on Adverse Events of Anti-HIV Drugs (D:A:D) study, a large prospective cohort study of cardiovascular events in HIV-infected persons living in developed countries, data regarding eGFR and proteinuria were not obtained [13]. Similarly, in the Strategies for Management of Anti-Retroviral Therapy (SMART) study, data on renal function was not provided in an analysis in which the association of antiretroviral agents with cardiovascular events was assessed [14].

Using a nested case-control design, we evaluated the association between markers of kidney disease (eGFR and proteinuria) and cardiovascular events in a well-characterized cohort of HIV-infected individuals in Baltimore, Maryland.

Methods

Setting and Cohort

The Johns Hopkins HIV Clinic in Baltimore, Maryland, provides care to a large number of HIV-infected individuals in the region. The Johns Hopkins Clinical Cohort includes data from over 6000 participants who have received primary care in the clinic from 1990 onward. Information from clinical records was reviewed and abstracted by trained technicians onto structured data collection forms, then entered into a relational database. The clinic medical records, the main hospital medical records, and various institutional computerized databases (e.g., laboratory, radiology, pathology, and hospital discharge summaries) were abstracted. Comprehensive demographic, clinical, laboratory, pharmaceutical, and psychosocial data were collected at times corresponding to enrollment in the clinic and at 6-month intervals thereafter. In 1998, information regarding possible cardiovascular events (coronary artery disease and stroke) was added to routine data abstraction procedures. The study protocol was approved by the Johns Hopkins Medicine Institutional Review Board and participants provide written, informed consent.

Participant Selection

One investigator (EG) reviewed the medical records of participants that were identified in the cohort database as having sustained a cardiovascular event. To be included as a case in the present analysis, participants had to meet established criteria for myocardial infarction [15] or cerebrovascular accident. Myocardial infarction was defined as 1) documented increase in cardiac biomarkers in combination with supporting symptoms, electrocardiogram findings, or cardiac imaging, 2) sudden death accompanied by symptoms suggestive of cardiac ischemia, or compatible electrocardiographic or angiographic findings, or 3) pathologic findings of acute myocardial ischemia. Cerebrovascular accident was defined as a focal neurological deficit lasting > 24 hours or imaging evidence of an acute, clinically relevant ischemic brain lesion that was associated with rapidly vanishing symptoms. Individuals with myocardial infarction or cerebrovascular accident were excluded from the analysis if the clinical history suggested that metastatic or primary cancer, vasculitis, drug overdose, interventional procedures, or infection (e.g., toxoplasmosis or septic embolization) were likely to have caused the event. Control subjects (four per case participant) were randomly selected from the cohort population using incidence-density sampling, with replacement, excluding individuals with any history of cardiovascular event. Controls were matched to cases by sex, race (black or non-black), and age (5-year intervals). Individuals who sustained multiple events during the observation period were included in the analysis only once at the earliest event.

Definitions

Laboratory values for cases and their matched controls were those closest in time, but not after, the event date for cases and the matching date for controls. GFR was estimated using the CKD-Epi formula [16] and the 4-variable Modification of Diet in Renal Disease (MDRD) Study equation [17] using serum creatinine, age, race, and sex. MDRD estimates that exceeded 200 mL/min/1·73m² were truncated at that value. The modified National Kidney Foundation classification of chronic kidney disease was used for GFR categorization: ≥90 mL/min/1·73m² (stage 1), 60 to 89 mL/min/1·73m² (stage 2), 30 to 59 mL/min/1·73m² (stage 3), 15 to 29 ml per minute per 1·73 m² (stage 4), and less than 15 ml per minute per 1·73 m² (stage 5). Stage 3, 4 and 5 CKD were combined due to small numbers of subjects in these strata. Proteinuria was defined as urine dipstick reading ≥ 1+.

Diabetes mellitus was defined as 1) prior diagnosis of diabetes mellitus, 2) prior/current treatment for diabetes mellitus, or 3) fasting plasma glucose >126 mg/dL (7·0 mmol/L) or random plasma glucose > 200 mg/dL (11·1 mmol/L). Hypertension was defined as 1) previous hypertension diagnosis, 2) blood pressure greater than or equal to 140 mm Hg systolic or 90 mm Hg diastolic on at least 2 occasions, or 3) use of antihypertensive pharmacological therapy [18]. Subjects were considered to have a history of cardiovascular disease if there was compelling history of possible MI or stroke not meeting the diagnostic criteria, peripheral arterial disease, angina, or intervention for coronary artery disease prior to the observation period. Family history was defined as cardiovascular disease in male first degree relative <55years or in a female first degree relative <65 years [19]. Dyslipidemia [19] was defined as total cholesterol >200 mg/dl, HDL cholesterol<40 mg/dl, serum triglycerides>150 mg/dl, or LDL cholesterol higher than the corrected goals determined by the risk factors on random blood lipid estimation. The major risk factors that modify LDL goals were smoking, hypertension, low HDL cholesterol (<40 mg/dl), and family history. HDL cholesterol >60 mg/dl removes one risk factor from the total count. The corrected LDL goals were <100mg/dl for coronary heart disease and coronary heart disease risk equivalents (Diabetes Mellitus), <130mg/dl for multiple (2+) risk factors, and <160 mg/dl for zero to one risk factor. Both current and previous cigarette smokers were included as smokers. A body mass index >30kg/m² was defined as obesity [20]. All above mentioned comorbidities were estimated from records closest in time, but not after the match date. Participants were included as HAART users if they had any exposure in the 60 days preceding the match date, similarly for the different subgroups of HAART drugs. Certain drugs like didanosine and abacavir which were shown to be associated with cardiovascular events in prior studies [13,14] were also analysed separately. Statin users were those who had used any statin drug for at least 6 months immediately prior to the match date.

Statistical Analysis

Odds ratios and 95% confidence intervals for the associations between explanatory variables and cardiovascular events were estimated using conditional logistic regression models for matched sets. P values below 0·05 were considered to be statistically significant. Estimates for the associations between the primary explanatory variables of interest (eGFR and proteinuria) were adjusted in multivariate models that included potential confounding covariates that were statistically significant in univariate analysis. Statistical analysis was performed using STATA version 10 software package (College Station, TX).

Results

A total of 117 participants were identified in the Johns Hopkins HIV Clinical Cohort as possibly sustaining a cardiovascular event between 1998 and 2008. Of these, 39 participants were found to have an alternative diagnosis, 15 participants had possible events, but diagnostic criteria were not met or supporting data was unavailable, leaving 63 cases that were included in the analysis. A total of 252 matched controls were selected for the 63 cases.

Baseline Characteristics

The baseline characteristics of cases and controls are presented in Table1. Cases and controls were closely matched by race, sex, and age and these covariates were not considered further in the analysis. The cases and controls were significantly different in the severity of HIV infection. The CD4 counts in the two groups were significantly different with 49·2% cases having a CD4 count < 200 cells/mm³ as compared to 24·6% of controls (p<0·001). Also, a greater proportion of cases (38·7%) had a HIV-1 RNA concentration >50,000 copies/ml as compared to controls (20·1%) (p=0·007). The cases and controls were not significantly different in the HIV transmission category. Subjects who experienced cardiovascular events were more likely to be diabetic (31·8% vs. 11·1%, p<0·001), hypertensive (63·5% vs. 36·5%, p<0·001), dyslipidemic (71·4% vs. 48%, p=0·001), and to have a previous history of cardiovascular event (41·3% vs. 6·7%, p<0·001). There was no statistically significant difference in the prevalence of obesity, cigarette smoking, family history of cardiovascular disease, or HAART use in the two groups. However, exposure to statins was different, being present in 21% of cases compared to 8·5% of controls (p=0·004).

Table 1.

Baseline Characteristics of participants

Variables	Cases (N=63)	Controls (N=252)	p
Demographic data
Age* [Mean (SD)]	49.5(10)	49.5(9.5)	NA
Male* N (%)	40(63.5)	160(63.5)	NA
Black* N (%)	53(84.1)	212(84.1)	NA
Transmission Category†
N (%)
Homo/Bisexual	8(12.7)	56(22.2)	0.105
Heterosexual	33(52.4)	138(54.8)	0.732
Injection drug use	29(46)	121(48)	0.909
Transfusion	4(6.3)	20(7.9)	0.386
HIV Status N (%)
CD4 count			<0.001
Median (IQR)	216(84-425)	375(210-559)
≤200 cells/mm3	31(49.2)	62(24.6)
201-350cells/mm3	13(20.6)	55(21.8)
>350 cells/mm3	19(30.2)	135(53.6)
HIV-1 RNA copies			0.007
Median (IQR)	6374(75-96826)	750(52-22131)‡
<1000/ml	26(41.9)	126(50.6)
1001-50,000/ml	12(19.4)	73(29.3)
>50,000/ml	24(38.7)	50(20.1)
Cardiovascular risk factors N (%)
Obesity	13(20.6)§	41(16.3)∥	0.777
Smoking	53(84.1)	189(75)¶	0.248
Diabetes	20(31.8)	28(11.1)	<0.001
Dyslipidemia	45(71.4)§	121(48)‡	0.001
Hypertension	40(63.5)	92(36.5)	<0.001
Family history	10(15.9)**	34(13.5)††	0.242
Previous cardiovascular events	26(41.3) ‡	17(6.7)	<0.001
Medications N (%)
Use of HAART	32(50.8)	124(49.2)	0.822
PI	17(27.4)	52(21.0)	0.275
NRTI	32(51.6)	113(45.6)	0.393
NNRTI	11(17.5)	61(24.2)	0.254
Abacavir	9(14.3)	52(20.6)	0.254
Didanosine	2(3.2)	12(4.8)	0.585
Statins	13(21.0)‡	21(8.5)¶	0.004

SD=Standard Deviation; IQR=Inter-Quartile range; HAART=Highly Active Anti-Retroviral Therapy; PI=Protease Inhibitors; NRTI=Nucleoside Reverse Transcriptase Inhibitors; NNRTI=Non-Nucleoside Reverse Transcriptase Inhibitors

^*Matched in case and control,

^†Transmission categories not additive since many patients had multiple risk factors for HIV acquisition,

^‡information missing in 3 patients,

^§information missing in 1 patient,

^∥information missing in 40 patients,

^¶information missing in 8 patients,

^**information missing in 20 patients,

^††information missing in 38 patients.

Kidney Function and Cardiovascular events

The kidney function of study participants are shown in Table 2. The mean serum creatinine (in mg/dl) was significantly higher in cases as opposed to controls (2·4 vs. 1·1, p<0·001). Using the CKD-Epi formula, the mean eGFR was significantly higher in controls (103·2 ± 27·6 ml/min/1.73m²) than in cases (68·4 ± 41·7 ml/min/1·73m²) (p<0·001). With eGFR >90 ml/min/1·73 m² as reference, eGFR 60–89 ml/min/1·73 m² was associated with an odds ratio of 3·9 (p=0·002) and eGFR <60ml/min/1·73m² with an odds ratio of 15·9 (p<0·001). When potential confounders in the relationship between eGFR and cardiac events including diabetes mellitus, hypertension, previous cardiac events, dyslipidemia, HIV viral load, and CD4 count were added to the model, eGFR 60–89 ml/min/1·73m² was associated with an odds ratio of 1·8 (p=0·175) and eGFR <60 ml/min/1·73m² with an odds ratio of 6·4 (p<0·001) as compared to eGFR >90 ml/min/1·73m². Use of the MDRD formula to derive eGFR estimates produced similar results. The presence of proteinuria was also found to be significantly different between cases (51%) and controls (25%) and was associated with an odds ratio of 3·6 (p<0·001) which remained significant in multivariate analysis (OR 2·2, p=0·038). When a composite variable was created that included both eGFR (CKD-Epi) and proteinuria, it was found that participants with eGFR of <60 ml/min/1·73m² and proteinuria had a 41-fold increased odds as compared to participants with an eGFR of ≥90 ml/min/1·73m² and no proteinuria (OR 41.4, p<0·001). This association remained highly statistically significant in adjusted analysis and with use of the MDRD GFR estimates. However, the confidence intervals were wide in this composite analysis.

Table 2.

Kidney Function of study participants using eGFR as a categorical variable

	Cases (n=63)	Controls (n=252)	Unadjusted Odds Ratio (95% CI)	p	Adjusted Odds Ratio (95% CI)	P
Serum Creatinine
(mg/dl)*				<0.001
Mean (SD)	2.4(2.9)	1.1(1.4)	-	-	-	-
Median (IQR)	1.2(0.9- 2.5)	0.8(0.7-1)	-	-	-	-
eGFR(By CKD-Epi Formula)*† (ml/min/1.73m2)
Mean (SD)	68.4(41.7)	103.2(27.6)	-	<0.001	-	-
Median (IQR)	65(33-105)	109(92-123)	-	<0.001	-	-
>90	21(33.3)	192(76.5)	1(ref)	-	1(ref)
60-89	13(20.6)	39(15.5)	3.9(1.7-9.1)	0.002	1.8(0.7-4.7)	0.175
<60	29(46)	20(8)	15.9(6.7- 37.5)	<0.001	6.4(2.6- 15.6)	<0.001
eGFR(By MDRD)*† (ml/min/1.73m2)
Mean (SD)	69(45.8)	103.1(32.5)		<0.001
Median (IQR)	62(32-99)	105(86-124)		<0.001
>90	19(30.2)	181(72.1)	1(ref)	-	1(ref)	-
60-89	15(23.8)	48(19.1)	3.3(1.5-7.3)	0.003	1.8(0.8-4.3)	0.175
<60	29(46)	22(8.8)	12.8(5.7- 28.6)	<0.001	5(2.1-11.8)	<0.001
Proteinuria‡
Absent/Trace	25(39.7)	158(62.7)	1(ref)	-	1(ref)	-
Present	32(51)	63(25)	3.6(1.9-7)	<0.001	2.2(1.1-4.8)	0.038
eGFR and Proteinuria(By CKD-Epi Formula)†
eGFR>=90 without proteinuria	14(41.2)	127(51.2)	1(ref)	-	1(ref)	-
eGFR<60 with proteinuria	22(34.9)	12(4.8)	41.4(5.5-312.1)	<0.001	28.5(2-400.1)	0.013
eGFR and Proteinuria(By MDRD)†
eGFR>=90 without proteinuria	13(20.6)	120(48.4)	1(ref)	-	1(ref)	-
eGFR<60 with proteinuria	22(34.9)	14(5.6)	18.5(4.2- 80.9)	<0.001	10.7(1.6- 69.5)	0.013

SD=Standard Deviation; IQR=Inter-Quartile range; eGFR=Estimated Glomerular Filtration Rate

^*Information missing in 1 control,

^†Adjusted for diabetes, hypertension, previous events, CD4 counts and HIV viral load, and dyslipidemia,

^‡Information missing in 6 cases and 30 controls.

In univariate analysis (Table 3), eGFR (CKD-Epi) was found to be a significant predictor of cardiac events with an odds ratio of 1·3 (95% CI 1·2–1·5) per 10 ml/min/1·73 m² decrease in value (Table 3). A similar correlation was also demonstrated using the GFR estimates derived using the MDRD formula (OR 1·3, 95% CI 1·2-1·4). A locally weighted regression and smoothing plot demonstrates a linear increase in cardiovascular disease risk as eGFR declined from normal values (~ 120 ml/min/1·73m²) (Figure 1). Other factors found significantly associated with cardiovascular events included prior cardiovascular event, diabetes mellitus, hypertension, dyslipidemia, statin use, CD4 cell count >350 cells/mm³, and HIV RNA plasma concentration >50,000 copies/ml. Family history of cardiovascular disease, obesity, and HAART use were not found to be significantly associated with cardiovascular events.

Table3.

Univariate and Multivariate analysis for odds of cardiac events using eGFR as continuous variable.

	Unadjusted Odds Ratio (95% CI)	P	Adjusted Odds Ratio (95% CI){Model 1}	P	Adjusted Odds Ratio(95%CI) {Model 2}	P
eGFR value* (CKD-Epi formula	1.3(1.2-1.5)	<0.001	1.2(1.1-1.4)	0.009
eGFR value* (MDRD formula)	1.3(1.2-1.4)	<0.001	-	-	1.2(1.1-1.5)	0.004
Proteinuria	3.6(1.9-7)	<0.001	2.9(0.9-9)	0.070	2.7(0.8-8.9)	0.101
Diabetes	4.4(2.1-9.3)	0.010	3.6(0.9-15.3)	0.077	5.9(1.1-29.1)	0.037
Hypertension	3.3(1.8-6.2)	0.001	1.3(0.5-3.9)	0.593	1.9(0.6-6.1)	0.253
Obesity	1.1(0.6-2.3)	0.31	-	-	-	-
Previous cardiac event	13.7(5.6-33.6)	<0.001	29.8(6.1-145.7)	<0.001	91.3(11.7-711.7)	<0.001
Family history	1.9(0.8-4.7)	0.149	-	-	-	-
Dyslipidemia	3.6(1.8-7.1)	<0.001	4.1(1.1-14.9)	0.034	2.9(0.8-10.5)	0.108
Statin use	3.6(1.5-8.7)	0.005	-	-	-	-
Any HAART use	1.1(0.6-1.9)	0.857	-	-	-	-
NRTI use	1.2(0.7-2.2)	0.479	-	-	-	-
PI use	1.4(0.9-2.3)	0.197	-	-	-	-
NNRTI use	0.6(0.3-1.3)	0.209	-	-	-	-
Abacavir use	0.6(0.2-1.4)	0.215	-	-	-	-
Didanosine use CD4 Cell Count (cells/mm3)	0.6(0.1-3.1)	0.574	-	-	-	-
>350	1(ref)	-	1(ref)	-	1(ref)	-
201-350	1.5(0.7-3.2)	0.332	0.4(0.1-1.7)	0.227	0.5(0.1-2.1)	0.308
<200	3.4(1.8-6.6)	<0.001	2.3(0.8-7.1)	0.141	4.8(1.3-18.2)	0.019
HIV-1 RNA (copies/ml)
<1000	1(ref)	-	1(ref)	-	1(ref)	-
1001-50,000	0.8(0.4-1.8)	0.639	2(0.5-7.3)	0.293	2.4(0.6-9.5)	0.201
>50,000	2.4(1.2-4.6)	0.011	1.5(0.3-6.5)	0.596	1.7(0.3-8.7)	0.551

Model 1 analyzed eGFR according to the CKD-EPI equation and Model 2 according to the MDRD GFR estimation formula

CI=Confidence Interval; eGFR=estimated glomerular filtration rate, HAART=highly active antiretroviral therapy; NRTI=nucleoside reverse transcriptase inhibitor; PI=protease inhibitor; NNRTI=non-nucleoside reverse transcriptase inhibitor

^*Odds Ratio for every 10 ml/min/1.73 m² fall in eGFR

Figure 1.

Relationship between estimated glomerular filtration rate (GFR) and cardiovascular event status in a nested, matched case-control study of HIV-infected subjects. The points correspond to individual subjects. The curve, generated by a locally-weighted regression smoothing function, represents cardiovascular disease risk according to estimated GFR.

When potential confounders in the relationship between eGFR and cardiac events such as diabetes mellitus, hypertension, previous cardiac events, dyslipidemia, HIV-1 RNA concentration, and CD4 count were adjusted for(Table 3), the odds ratio for every 10 units/ml/min/1·73m² decrease in eGFR (CKD-Epi) (Model1) remained almost unchanged with an odds ratio of 1·2 (p=0·009). In another multivariate model using eGFR (MDRD) (Model 2), identical results were obtained (OR 1·2, p=0·004).

In multivariate models, lower eGFR remained significantly associated with cardiovascular events (Table 3), whether GFR was estimated with the CKD-Epi formula (Model 1) or the MDRD formula (Model 2). Other factors that remained significantly associated with cardiovascular events included history of previous cardiac events, diabetes, dyslipidemia, and CD4 count of <200 cells/mm³ though results were not consistent in the two models. However, proteinuria, hypertension, and HIV-1 RNA plasma concentration were not significantly associated with cardiovascular events, after adjustment for other factors in these models.

Discussion

In an HIV clinic-based population, decreasing eGFR was associated with a significantly increased risk of cardiovascular events independent of traditional cardiovascular risk factors and HAART. While an eGFR of 60–89 ml/min/1·73m² was associated with a marginally increased risk which became non-statistically significant after adjustment for other factors, an eGFR <60ml/min/1·73m² was associated with a significant 5- to 6-fold increase in odds of cardiovascular events. The association between GFR and CVD events was similar whether GFR was estimated with the MDRD or CKD-Epi equations. Consistent with other studies in the general population, proteinuria was also a significant independent predictor of cardiovascular disease in this group of patients and, as shown in Table 2, individuals with both proteinuria and an eGFR <60ml/min/1·73m² were at markedly increased risk for cardiovascular events. However, proteinuria was not found to be a significant predictor of cardiovascular events in the continuous eGFR model possibly suggesting an eGFR-dependent additive effect of proteinuria, which amplifies the risk when associated with low eGFR, but not contributing significantly when associated with small changes in eGFR.

Our study highlights the existence of a strong link between kidney function and CVD risk in HIV-infected individuals, an association that is at least as strong, if not stronger, than that reported the general population. This finding is important because the 1) prevalence of kidney disease has been found to be 3- to 5-fold higher in HIV-infected than in HIV-negative persons [2, 21] 2) indicators of reduced GFR or kidney damage (e.g., albuminuria) have often been omitted in large multi-site studies addressing risk factors for cardiovascular events in HIV-infected persons [13,14]. The strength of the association we found is notable because only relatively crude measures of kidney disease were available to us in this retrospective study. Equation-based estimates of GFR are inaccurate at higher levels of kidney function [22]. In our analysis we included GFR estimated by both the widely used MDRD equation [16] and the recently published CKD-Epi equation [17], the latter of which is designed to function better in a population that does not predominantly have kidney disease.

Other traditional factors found to be significantly associated with cardiovascular events in our analysis include diabetes mellitus, hypertension, previous cardiovascular events, dyslipidemia, statin use, CD4 counts, and HIV-1 RNA concentration. The higher odds for a cardiovascular event associated with statin use could be accounted for by associated dyslipidemia and poor lipid level control achieved with statin therapy in HIV patients as opposed to non-HIV population [23-25]. The significant risk associated with hypertension, and higher HIV-1 RNA viral loads on univariate analysis was not found on multivariate analysis suggesting overlap of causal pathway with the other variables. Family history was not found to be a significant risk factor in this study. However, it is possible that family history of cardiovascular events was not rigorously collected in this HIV based cohort. We found no significant associations between cardiovascular events and HAART use, individual drug class use, or abacavir or didanosine use, as have been found in other studies [13,14,26]. However our study was not sufficiently powered to detect effect sizes in the range found in prior studies of associations between antiretroviral drug use and cardiovascular events.

Many mechanisms have been proposed for the cardiovascular risk in kidney disease. Renal insufficiency is known to be associated with elevated levels of apolipoprotein B [27], fibrinogen [27], homocysteine [27], C-reactive protein [27] and other inflammatory and procoagulant biomarkers [28], and decreased apolipoprotein A1 [27]. Anemia associated with decreased erythropoietin production in CKD also contributes to cardiovascular risk by causing elevated levels of markers of endothelial activation and left ventricular hypertrophy [29-31]. Secondary hyperparathyroidism associated with CKD is also known to be an independent risk factor for cardiovascular events [32]. Hypercalcemia and hyperphosphatemia play a major role in the occurrence of vascular calcification in patients with CKD, together with endocrine disturbances including vitamin D, fibroblast growth factor-23, and klotho [33].

There has been much interest in studying cardiovascular risk factors, morbidity and mortality in HIV-1 infected patients. Traditional cardiovascular disease risk factors [4,5], lower CD4 counts [4] and duration of NRTI [4] and PI [5] use have been reported to be associated with cardiovascular events. Traditional risk factors, inflammatory effects of HIV and the metabolic complications of antiretroviral therapy [34] are hypothesized to underlie the pathophysiology of cardiovascular diseases in HIV patients. No previous study has correlated kidney function with cardiovascular events in HIV patients. In 2004, Go et al performed a longitudinal study in 1,120,295 non-HIV-infected adults and observed a graded increase in cardiovascular events with decreasing GFR [7]. Using >60 ml per minute per 1·73m² as reference, the study demonstrated increasing odds ratio for any cardiovascular event with worsening stages of CKD.

This study has several limitations. A single measurement of serum creatinine prior to the event might not be an accurate measurement of kidney function. Neither the MDRD nor the CKD-Epi study equations are validated in the HIV-1 infected population. The sample size of our study was small, limiting our ability to detect potentially clinically-important associations. Proteinuria was determined by semi-quantitative measure of urine protein concentration (dipstick), which is inferior to quantitative measures such as protein-to-creatinine ratio from a random or 24-hour sample [35]. Additionally, data on proteinuria were not complete since urinalyses, unlike serum creatinine measurements, were not routinely performed in this cohort. Other potential confounders that were not accounted for include diet, physical activity, duration of HIV infection, duration of and adherence to HAART medication and severity of co-morbidities like diabetes mellitus and hypertension. HAART drugs other than abacavir and didanosine were analysed only within their subgroups and not as individual drugs. Lipoatrophy, a known complication of HAART therapy, could not be distinctly defined and adjusted for. Abnormalities in the individual components of lipid profile were not analysed separately for effect on cardiovascular events. All current and previous smokers were considered together and quantification of total exposure was not attempted. Considering the population differences and the hospital-based nature of the study, the results may not be generalizable to all HIV cohorts. Nevertheless, our study lays the groundwork for future prospective studies that involve a larger sample size to further explore the potential role of kidney function on cardiovascular disease in HIV-1 infected individuals.

In conclusion, we found an independent association between decreasing GFR and the risk of cardiovascular events in HIV-1 infected patients. This risk was prominent at eGFR below 60 ml/min/1·73m². Similarly, the presence of proteinuria in this study was an independent predictor of cardiovascular events in addition to the traditional risk factors and its effect was amplified by low GFR. Our findings require further confirmation but suggest the potential value of early screening and treatment of CKD in HIV-1 infected patients particularly those with other cardiovascular risk factors.