Oral and Systemic Health Correlates of HIV-1 Shedding in Saliva

Abstract

The relationship among oral and systemic health and HIV shedding in saliva is not well-understood. We hypothesized that oral and systemic health are associated with HIV shedding in saliva of HIV-infected women. Saliva from 127 participants enrolled in the Women’s Interagency HIV Study (WIHS) was collected at repeated visits over a 5½-year study period (October 1998 through March 2004) and was evaluated for HIV-1 RNA. Demographic, lifestyle, and systemic and oral health characteristics were evaluated as possible correlates of salivary HIV-1 shedding. Multivariate models showed significantly increased risk of HIV-1 shedding in saliva as blood levels of CD4 cell counts decreased (p < 0.0001) and HIV RNA increased (p < 0.0001). Diabetes (p = 0.002) and a high proportion of gingival bleeding sites (p = 0.01) were associated with increased likelihood, while anti-retroviral therapy (p = 0.0003) and higher levels of stimulated saliva flow rates (p = 0.02) were associated with a lower likelihood of HIV-1 RNA shedding in saliva.

Introduction

In 2007, about 33 million people globally were infected with HIV-1. The number of HIV infections worldwide has been stable since 2000; however, new cases continue to increase in resource-limited nations. The source of infection differs by country, and includes intercourse (vaginal or anal), injected drugs, transmission from mother to child (including breast milk), and unsafe injections or other nosocomial sources (Cohen et al., 2008). The estimated incidence of HIV in the United States in 2006 was 56,300, affecting primarily blacks (45%) followed by whites (35%), Hispanics (17%), Asian/Pacific Islanders (2%), and American Indian/Alaska Natives (1%). Of these, 15,000 (27%) occurred in females (Hall et al., 2008). HIV transmission in females is predominantly attributed to high-risk heterosexual contact, accounting for 80% of new infections.

Blood, semen, vaginal and cervical secretions, and breast milk are established sources of HIV transmission (Schacker et al., 1996). While HIV is detected in saliva at various levels with different techniques (Liuzzi et al., 1996; Shugars et al., 2000; Freel et al., 2003), the possibility of HIV transmission via saliva is remote (Scully and Porter, 2000), a finding explained by the relatively low viral load in saliva, hypotonic disruption of mononuclear leukocytes, low numbers of CD4-positive target cells, anti-HIV antibodies, and inhibition of virus by salivary carbohydrate moieties and proteins such as secretory leukocyte protease inhibitor (SLPI) (Yeung et al., 1993; Shugars and Wahl, 1998; Baron et al., 1999; Shugars et al., 2002; Lin et al., 2004; Habte et al., 2006). However, the possibility of transmission of HIV and other viruses from saliva remains feasible if oral and systemic health is compromised (Greenspan et al., 2000; Gandhi et al., 2004).

Possible sources for viral particles, HIV virions and proviral HIV-1 DNA, include serum exudate and migration of HIV-1-containing mononuclear cells from gingival crevicular fluid (GCF) (Maticic et al., 2000). Other potential sources include oral ulcerations or erosions, inflamed gingiva, and oral herpes (Campo et al., 2006). The impact of oral and systemic health on HIV shedding in saliva is not well-understood, and few studies have addressed this (Liuzzi et al., 1996; Maticic et al., 2000; Shugars et al., 2000). In addition, most studies have been cross-sectional in design with limited data for a short period of time, and there is a paucity of such data in HIV-infected women.

We hypothesized that oral and systemic health factors are associated with HIV shedding in the saliva of infected women. The objective of the study was to correlate the presence of HIV-1 RNA in saliva samples from participants of the Women’s Interagency HIV Study (Barkan et al., 1998; Mulligan et al., 2004) with demographic, lifestyle, and medical and oral health characteristics. We believe that this is the first longitudinal study to assess the effects of oral and systemic health on HIV-1 shedding in the saliva of HIV-positive women in the United States.

Materials & Methods

Study Sample

The Women’s Interagency HIV Study was established to investigate the impact of HIV infection on women in the United States. It initially enrolled 2059 HIV-seropositive and 569 demographically similar HIV-seronegative women between October 1994 and November 1995. A second recruitment enrolled 737 HIV-seropositives and 406 HIV-seronegatives between 2001 and 2002. The current substudy is based on data on HIV-positive women from both recruitments. Oral and systemic health data and saliva were collected at baseline and at follow-up evaluations every 6 mos at Los Angeles and Chicago sites. A total of 127 participants donated saliva from October 1998 through March 2004. Of these 127 women, 32 contributed data on 1 visit, 34 had 2 visits, 25 had 3 visits, 14 had 4 visits, and 22 had between 5 and 7 visits, for a total of 354 subject-visits. A complete description of the Women’s Interagency HIV Study design and participants has been reported (Barkan et al., 1998). The study was approved by institutional review boards, and informed consent was obtained from all participants.

Study Outcome Measures

Samples of unstimulated and chewing-stimulated whole saliva were collected under standardized conditions (Navazesh and Christensen, 1982). The stimulated saliva collection involved women chewing on a standard-sized gum base for 2 min; saliva generated was discarded. Saliva was collected for an additional 3 min and saved for further analysis.

HIV-1 RNA in saliva was measured by a commercial NASBA/NucliSens^® assay (Nucleic Acid Sequence Based Amplification Assay; BioMerieux, Durham, NC, USA). The binary response variable used in data analysis was the detection of HIV-1 RNA in saliva, with the limit of detection ≥ 25 copies per mL. The NASBA/NucliSens^® assay’s software reports a quantitative result if the sample’s HIV-1 RNA is within the range of 25 to approximately 5 million copies per mL. The performance of the assay is linear from 51 to 5.39 x 10⁶ copies/mL HIV-1 RNA. The detection rate diminishes below 176 copies/mL (> 95% detection rate) to approximately 50% detection rate at 40 copies/mL and below (Nowicki et al., 2001).

Other Data Collection

Demographic, lifestyle, and medical and oral health characteristics were evaluated as possible correlates of salivary HIV-1 shedding. Sociodemographic variables included age, race/ ethnicity, and education. Lifestyle measures were alcohol use, smoking, recreational and illicit drug use, frequency of oral sex, and HIV exposure category. Health measurements included body mass index, self-reported diabetes, hepatitis C serostatus, AIDS status, CD4+ cell count, and HIV-1 RNA in plasma. Anti-retroviral therapy use was modeled in two ways: type of anti-retroviral therapy currently used (monotherapy, combination therapy, or highly active anti-retroviral therapy), and as an indicator of highly active anti-retroviral therapy use relative to prior visits (never on, initiated, continued on, or terminated highly active anti-retroviral therapy). Oral health characteristics included the following: enlarged parotid and submandibular glands; unstimulated and stimulated salivary flow rates; hard, tender, or > 1 cm lymph nodes; plaque; gingival banding; proportion of bleeding sites; papilla with cratering; pocket depth > 4 mm; proportion of teeth with loss of attachment > 2 mm; proportion of teeth with recession; and number of decayed/missing/filled teeth, decayed/missing/filled surfaces, and decayed/filled root surfaces. We evaluated proportion rather than absolute numbers of teeth/sites affected, since the number of teeth present for evaluation varied across participants. Stimulated and unstimulated salivary flow rates were also evaluated as correlates of HIV shedding. Independent variables except for age, race, education, and HCV status were time-dependent (based on data collected at each visit). Descriptions of methodologies for oral and systemic health parameters have been published (Mulligan et al., 2000, 2004; Navazesh et al., 2000, 2009; Gandhi et al., 2004; Greenspan et al., 2004; Phelan et al., 2004; Alves et al., 2006).

Statistical Analysis

Univariate and multivariate analyses used logistic regression with generalized estimating equations (GEE) to account for the correlated data arising from repeated measurements from participants over multiple visits. The dependent variable was the presence of detectable HIV-1 RNA in saliva. Independent variables with univariate p-values < 0.10 were included in multivariate modeling; variables significant at p < 0.05 on multivariate modeling were retained. Due to the possible co-linearity due to the high negative correlation between CD4+ cell counts and plasma HIV-1 RNA (r = -0.54, p < 0.0001), two multivariate models were evaluated, one with CD4+ cell counts included (CD4 model), and the other with HIV-1 RNA in plasma included (RNA model). To determine if associations with salivary shedding were constant over the study follow-up, we tested interactions among all variables in the final model and study visit. To consider the quantitative nature of the HIV RNA data, we also used an ordinal logistic regression approach, categorizing the salivary HIV RNA outcome at each visit as: not detectable, 26-499, 500-4999, and ≥ 5000 copies/mL. Since HIV RNA was not detected in saliva in 60% of subject visits, and very few displayed very high RNA levels, the results did not substantially differ from the dichotomous analysis. We therefore present only the dichotomous results. A two-sided alpha level for the final model was set at 0.05. All analyses used SAS Version 9.0 software.

Results

At baseline, 48 (38%) of the 127 women had detectable HIV-1 RNA in saliva. Over all visits, 141 (40%) of 354 specimens had detectable virus (Table 1). Independent variables associated on univariate analysis with salivary shedding at p < 0.10 (Table 2) were included in multivariate modeling (Tables 3 and 4).

Table 1.

Baseline Demographics and Frequency of Salivary HIV-1 Shedding (n = 127 WIHS participants)

Characteristic
All participants, baseline salivary shedding	48/127 (38)a
Age (median, 35 yrs; IQR, 29-43 yrs)
< 30	14/32 (44)
30-39	18/43 (42)
40-49	13/45 (29)
50+	3/7 (43)
Race
White	1/5 (20)
Black	22/55 (40)
Hispanic	25/66 (38)
Other	0/1 (0)
AIDS
No	35/98 (36)
Yes	13/29 (45)
CD4+ cells (median, 400; IQR, 246-551)
500+	7/38 (18)
350-499	10/38 (26)
200-349	19/30 (63)
< 200	12/21 (57)
HIV-1 RNA in plasma (median, 1055; IQR, 80-13,000)
< 1000	5/63 (8)b
1000-9999	16/30 (53)
10,000+	26/33 (79)
ARV therapy
None	25/42 (60)
Mono/Combo	5/16 (31)
HAART	18/69 (26)

^aNo. with salivary shedding/total no. in demographic category (percent with salivary shedding).

^bOne participant with missing viral load at baseline.

Viral load in plasma was censored at 80 copies/mL.

Table 2.

Univariate Logistic Regression of HIV Shedding in Saliva

Variable	Salivary HIV-1 Shedding	OR (95% CI)	P-value
Alcohol use in last 6 months
No	71/207 (34)a	1.00
Yes	68/143 (48)	1.57 (0.97-2.54)	0.07
Drug use in last 6 months (crack, cocaine, heroin, IDU)
No	117/318 (37)	1.00
Yes	22/34 (65)	2.87 (1.08-7.63)	0.03
Diabetes
No	131/338 (39)	1.00
Yes	9/15 (60)	2.75 (0.95-7.99)	0.06
AIDS
No	90/247 (36)	1.00
Yes	51/107 (48)	1.59 (0.95-2.68)	0.08
ARV therapy at study visit
None	57/96 (59)	1.00	0.0002b
Mono/Combo	17/45 (38)	0.32 (0.14-0.75)
HAART	66/212 (31)	0.28 (0.16-0.48)
CD4+ cell count
500+	37/139 (27)	1.00	< 0.0001b
350-499	27/84 (32)	1.38 (0.75-2.53)
200-349	37/65 (57)	3.56 (1.76-7.20)
< 200	39/62 (63)	4.72 (2.46-9.06)
HIV-1 RNA in plasma
< 1000	39/202 (19)	1.00	< 0.0001b
1000-9999	34/67 (51)	4.16 (2.15-8.08)
10,000+	67/81 (83)	19.6 (10.3-37.3)
Plaque
No or probe only	17/57 (30)	1.00
Visible	111/273 (41)	1.57 (0.93-2.68)	0.09
Average gingival banding
None	111/303 (37)	1.00
Present	17/27 (63)	2.22 (0.97-5.10)	0.06
Decayed/filled root surfaces
None	103/278 (37)	1.00
Present	29/57 (51)	1.64 (0.91-2.96)	0.10
Stimulated salivary flow rate, mL/minc
Positive for sheddingd	0.67 (0.67)d	0.67 (0.46-0.98)	0.04
No shedding	0.83 (0.75)
Proportion of teeth with plaque visiblec
Positive for sheddingd	0.25 (0.40)d	2.08 (0.93-4.66)	0.08
No shedding	0.19 (0.39)
Proportion of bleeding sitesc
Positive for sheddingd	0.20 (0.30)d	4.28 (1.59-11.5)	0.004
No shedding	0.14 (0.24)
Visitc		1.16 (1.07-1.25)	0.0003b

Table includes independent variables associated with salivary shedding at p ≤ 0.10. Parameter estimates and significance tests are from logistic regression with generalized estimating equations.

^aNo. of visits with salivary shedding/total visits (percent).

^bP-value for trend.

^cORs for continuous variables are based on a one-unit change in the predictor; for study visit, one-unit change = 6 months.

^dMedian (IQR) by salivary shedding status.

Table 3.

Multivariate Logistic Regression (CD4 model)^a

Variable	OR (95% CI)	P-value
Visitb	1.23 (1.11-1.37)	< 0.0001
Diabetes
No	1.00
Yes	3.93 (1.64-9.38)	0.002
ARV therapy
None	1.00	0.0003c
Mono/Combo	0.35 (0.12-1.01)
HAART	0.23 (0.12-0.46)
CD4+ cell count
500+	1.00	< 0.0001c
350-499	1.58 (0.78-3.21)
200-349	3.78 (1.65-8.58)
< 200	4.90 (2.17-11.1)
Stimulated salivary flow rate, mL/min	0.57 (0.36-0.91)	0.02
Proportion of bleeding sites	1.42 (1.10-1.84)d	0.01

^aLogistic regression with generalized estimating equations (319 visits, 119 subjects), using exchangeable correlation.

^bOR per one unit change in study visit = 6 months.

^cP-value for trend.

^dOR based on a one standard deviation change in the independent variable.

Table 4.

Multivariate Logistic Regression (RNA model)^a

Variable	OR (95% CI)	P-value
Visitb	1.37 (1.21-1.54)	< 0.0001
Diabetes
No	1.00
Yes	5.86 (2.45-14.0)	< 0.0001
ARV therapy
None	1.00	0.002c
Mono/Combo	0.67 (0.22-2.10)
HAART	0.30 (0.15-0.63)
HIV-1 RNA in plasma
< 1000	1.00	< 0.0001c
1000-9999	5.47 (2.61-11.5)
10,000+	43.4 (18.8-100)
Decayed/filled root surfaces
None	1.00
Present	2.84 (1.45-5.56)	0.002
Proportion of bleeding sites	1.65 (1.22-2.24)d	0.001

^aLogistic regression with generalized estimating equations (321 visits, 120 participants) using exchangeable correlation.

^bOR per one-unit change in study visit = 6 months.

^cP-value for trend.

^dOR based on a one standard deviation change in the independent variable.

In the multivariate model including CD4 cell count (Table 3), the risk of HIV-1 shedding in saliva increased significantly with decreasing levels of CD4 cell counts (p < 0.0001). History of diabetes (p = 0.002) and a high proportion of gingival bleeding sites (p = 0.01) were associated with increased likelihood of HIV-1 RNA shedding, while anti-retroviral therapy (p = 0.0003) and higher stimulated salivary flow rates (p = 0.02) were associated with a lower likelihood of HIV-1 RNA shedding.

In the multivariate model including HIV-1 RNA in plasma (Table 4), the risk of HIV-1 shedding in saliva increased significantly with increasing HIV RNA (p < 0.0001). The Spearman correlations between baseline plasma and saliva HIV RNA were 0.60 (unadjusted, p < 0.0001) and 0.61 (adjusted for variables in Table 4, p < 0.001). History of diabetes (p < 0.0001), high proportion of gingival bleeding sites (p = 0.001), and decayed/filled root surfaces (p = 0.002) were associated with increased likelihood of HIV-1 RNA shedding in saliva, while anti-retroviral therapy (p = 0.002) was associated with a lower likelihood of HIV-1 RNA shedding. In both multivariate models, study visit number (i.e., longer duration of follow-up) was associated with higher likelihood of HIV shedding in saliva (p < 0.0001), indicating increasing prevalence of HIV shedding over follow-up (baseline prevalence = 38%, median prevalence visits 1-3 = 30%; median prevalence visits 4-6 = 44%; median prevalence visits 7-9 = 62%).

Discussion

Forty-eight (38%) of the 127 women had detectable HIV-1 RNA in saliva in all study visits, whereas, overall, 141 (40%) of 354 specimens had a detectable salivary viral load. This was comparable with results from an earlier study, where 42% of participants had HIV-1 RNA shedding in saliva (Shugars et al., 2000). However, this is not comparable with results from other studies that reported very high (96%) (Liuzzi et al., 1996) or very low (1%) (Barr et al., 1992) prevalence of salivary HIV-1 RNA shedding. The use of different collection methods and/or molecular techniques for measuring salivary viral shedding may explain these differences.

Higher HIV-1 RNA viral load in plasma was strongly associated with a higher likelihood of salivary viral shedding. This was noted in earlier studies (Liuzzi et al., 1996; Shugars et al., 2000) and reinforces the idea that saliva may be a useful, non-invasive source for estimating plasma viral load (Shugars et al., 2000). In addition, decreased CD4+ cell count and positive AIDS status (in univariate models only) increased the risk of HIV-1 shedding in saliva, suggesting an effect of immunosuppression on salivary viral shedding.

The advent of highly active anti-retroviral therapy has significantly reduced HIV viral load systemically in plasma and other body fluids, including saliva (Shugars et al., 2000). This was evident in the Women’s Interagency HIV Study, where participants on highly active anti-retroviral therapy were less likely to have detectable HIV-1 RNA in saliva than participants not on anti-retroviral therapy.

Diabetes increased the likelihood of salivary HIV-1 shedding. Alteration in saliva composition is noted in diabetics with otherwise normal oral and systemic health. The mechanism by which diabetes increases HIV-1 shedding in saliva is unknown and deserves further investigation (Yavuzyilmaz et al., 1996).

We previously reported the association among salivary gland hypofunction, HIV-1 infection, and highly active anti-retroviral therapy in the Women’s Interagency HIV Study (Navazesh et al., 2003, 2009). The current investigation revealed a significant association between lower volume of chewing-stimulated saliva and higher risk for HIV-1 shedding in saliva. This could imply that endogenous viral inhibitors suppress HIV-1 shedding in saliva.

Interestingly, we observed an increase in prevalence of HIV-1 shedding in saliva with longer follow-up (higher subject visits). The mean plasma viral loads were elevated beyond the baseline level in later study visits (beyond visit 5) in some participants, which could increase the HIV-1 shedding in saliva (results not shown here). We speculate that this could be due to non-compliance with HIV medication intake or development of resistance to medications.

We observed a significant association between the proportion of bleeding sites and the risk of HIV-1 shedding in saliva. Similarly, increased linear gingival banding and visible plaque (only in the univariate model) were associated with increased HIV shedding that was similar to a previous observation (Maticic et al., 2000; Shugars et al., 2000). Inflammatory conditions such as linear gingival banding or erythema may increase viral shedding through ulcerated surfaces or via gingival crevicular fluid (Shugars et al., 2000). However, these findings conflict with earlier reports indicating no significant association between infectious HIV-1 in whole saliva and periodontal disease (Barr et al., 1996). These differences could be attributed to differences in populations and different sensitivities of the HIV assays. Among the other dental variables, the presence of decayed/filled root surfaces increased the risk of HIV-1 shedding in saliva.

Limitations of this study include the fact that some independent variables were based on participant self-report and hence could be subject to higher measurement error. Despite the fact that multiple systemic and oral health measures were obtained from the same cohort, the comprehensive nature of measures examined, as well as the longitudinal study design, make it more challenging for objective assessments to be obtained for all the variables. Given the large number of associations tested, it is possible that some of our findings may have arisen by chance; however, the highly significant associations noted in our multivariate models make this possibility unlikely. Nonetheless, these findings should be replicated in an independent sample.

To our knowledge, this is the first comprehensive evaluation of HIV-1 shedding in saliva of HIV-infected women. We observed specific oral and systemic health parameters to be associated with HIV-1 RNA shedding in saliva of HIV-infected women in the Women’s Interagency HIV Study.