Risk factors associated with incident sexually transmitted infections in HIV-positive patients in the Australian HIV Observational Database- a prospective cohort study

Abstract

Background

We established a sub-cohort of HIV positive individuals from ten sexual health clinics within the Australian HIV Observational Database (AHOD). The aim of this paper was to assess demographic and other factors that might be associated with an incident sexually transmitted infection (STI).

Methods

The cohort follow-up was from March 2010 to March 2013, and included patients screened at least once for an STI. We used survival methods to determine time-to first new and confirmed incident STI infection (chlamydia, gonorrhoea, syphilis or genital warts). Factors evaluated included sex, age, mode of HIV exposure, year of AHOD enrolment, hepatitis B or C coinfection, time-updated CD4 cell count, time-updated HIV RNA viral load, and prior STI diagnosis.

Results

There were 110 first incident STI diagnoses observed over 1015 person-years of follow-up, a crude rate of 10.8 (95% confidence interval [CI]:9.0-13.0) per 100 person-years. Factors independently associated with increased risk of incident STI included younger age (≥50 vs 30-39 years old, adjusted hazards ratio, [aHR]=0.4; 95% CI:0.2-0.8, p<0.0001); prior STI infection, (aHR=2.5; 95% CI: 1.6-3.8, p<0.001); and heterosexual vs men who have sex with men (MSM) as likely exposure (aHR=0.2; 95% CI: 0.1-0.6; p<0.001).

Conclusions

In this cohort of individuals being treated with anti-retroviral drugs (ARV), who are MSM, 30-39 years old, and with a prior history of STI, are at highest risk of a further STI diagnosis

Introduction

Over the last 20 years a wealth of evidence has accumulated to support the strong amplification effect of STI on the acquisition and infectiousness of HIV (1-3). Recently, it has been shown that treatment with anti-retroviral drugs (ARV) can reduce onward HIV transmission (5-14), but some debate exists whether this is true in the presence of an inflammatory sexually transmitted infection (STI). (15-18). Therefore, it is important to estimate the incidence and prevalence of STI in populations at potential risk of transmitting HIV.

In Australia, new HIV diagnoses occur predominantly among men who have sex with men (MSM), a population in whom STI rates have been increasing for several years (most dramatically for infectious syphilis, particularly among those who are HIV-positive) (19-23).

There is a broad consensus that the increased incidence of some STI may be due to the increased prevalence of condomless intercourse in both HIV-negative and HIV-positive MSM (23,41). By combining retrospective and prospective data we have previously shown high levels of four STI in a sub-cohort of HIV positive individuals enrolled in AHOD (24).

In the present study, we have collected additional prospective data enabling risk factors for new STI diagnoses to be assessed.

Methods

The current study is a prospective cohort, being a subset of clinical sites of AHOD. AHOD was originally designed as a study of changes in HIV treatment patterns, and has been described in detail elsewhere (25-27). In AHOD, data are collected on a core set of variables including sex, age, HIV exposure category, hepatitis B virus surface antigen (HBV), hepatitis C antibody status (HCV), CD4 and CD8 cell counts, viral load, ARV history, AIDS illnesses, date and cause of death. Data variables are sent electronically to the Kirby Institute.

AHOD data collection commenced in 1999 and currently 27 hospitals, sexual health clinics, and general medical practices throughout Australia contribute data twice annually. At March 2013, over 3000 patients had been recruited to AHOD, and 2,328 were under active follow-up, 90% of whom were receiving ARV for HIV infection.

For the current study ten sexual health clinics within AHOD were invited to provide prospective data from 2010. Besides the core AHOD data above, extra STI-specific data variables were extracted from each clinic database. They included confirmed diagnoses of STI, (infectious syphilis, chlamydia, gonorrhoea, and genital warts), site of infection, and STI treatment.

Statistical Methods

This analysis included all patients who were in active follow-up at the ten participating sites, screened at least once for an STI and had a least one CD4 and viral load test during the follow-up period. The cohort follow-up was from March 2010 to March 2013. Time-to the first incident STI infection (chlamydia, gonorrhoea, syphilis or genital warts) was summarised using Kaplan-Meier plots. The Log Rank test was used to compute p-values comparing the survival curves. Risk factors were assessed using Cox proportional hazard model (28). Covariates considered included sex, age at enrolment, mode of HIV exposure, decade of HIV diagnosis, prior STI diagnosis, hepatitis B or C coinfection, CD4 cell count and HIV RNA viral load. Co-variates that changed over time (CD4 cell count and HIV RNA viral load) were included as time-updated variables. Time to a diagnosis of an STI was the endpoint, therefore the latter co-variates were updated each time they were measured, and were thus used to predict the future risk of an STI. Our endpoint was time from when prospective follow-up began (March 2010) to first STI diagnosis. People who did not experience a STI diagnosis were censored at date of last follow-up.

A multivariate model including all covariates was used to assess independent associations. Sensitivity analyses were performed assessing risk factors for each individual STI. All analyses were performed using the statistical packages SAS and STATA (29,30).

Additional ethics approval was sought for the current sub-study and approved by local Human Research Ethics Committees.

Results

At March 2013, there were 598 patients (that had been recruited to the parent cohort) with prospective data from the participating sites, of which 547 patients had been STI screened at least once. There were 489 remaining patients that had further CD4 and viral load test results available to be included in the analysis. All were receiving ARV. There were 110 first incident STI events observed over 1015 person-years (pys), a crude rate of 10.8 (95% CI: 9.0-13.0) per 100 pys. When considering multiple STI diagnoses during the follow up period, there were 202 incident STI events observed over 1071 pys, a crude rate of 18.9 (95% CI: 16.4-21.7) per 100 pys.

The 110 incident STI events consisted of 155 separate STI diagnoses, where 78 patients had one STI diagnosis and 32 patients had 2 or more STI diagnoses. Of the 155 STI diagnoses, there were 59 chlamydia, 29 syphilis, 39 gonorrhoea and 28 genital warts. Overall, the STI with known site diagnosis were distributed between rectal (31%), urethral (51%), pharyngeal (7%) and other (12%).

Table 1 shows patient demographics and characteristics by presence of incident STI infection.

Table 1.

Patient demographics and characteristics by incident STI.

Factor		STI N (%)	No STI N (%)
Total		110	379
Sex	Male	106 (96)	352 (93)
	Female	4 (4)	27 (7)
Age (years) at study enrolment (2010)	<30	24 (22)	23 (6)
	30-39	46 (42)	119 (31)
	40-49	30 (27)	128 (34)
	>50	10 (9)	109 (29)
Mode of HIV exposure	MSMa	97 (88)	271 (72)
	Heterosexual	4 (4)	55 (15)
	PWIDb	8 (7)	26 (7)
	Other	1 (1)	27 (7)
Year of cohort enrolment	1999-2009	71 (65)	295 (78)
	2010-2013	39 (35)	84 (22)
Hepatitis B Coinfection (ever)	Yes	3 (3)	14 (4)
	No	69 (63)	299 (79)
	Not Tested	38 (35)	66 (17)
Hepatitis C Coinfection (ever)	Yes	11 (10)	38 (10)
	No	94 (85)	325 (86)
	Not Tested	5 (5)	16 (4)
STIc prior to study enrolment (2010)	No	48 (44)	302 (80)
	Yes	62 (56)	77 (20)
CD4 cell count (cells/μL) at study enrolment (2010)	0-349	18 (16)	68 (18)
	350-499	31 (28)	92 (24)
	≥500	61 (55)	217 (57)
	Not tested	0 (-)	2 (1)
HIV viral load (copies/mL) at study enrolment (2010)	<399	68 (62)	308 (81)
	400-9999	18 (16)	25 (7)
	≥10000	24 (22)	42 (11)
	Not tested	0 (-)	4 (1)

^aMSM: Men who have sex with men.

^bPWID: People who inject drugs.

^cSTI: Sexually Transmitted Infection.

Table 2 shows factors associated with STI diagnosis in univariate and multivariate analysis. Factors independently associated with increased risk of incident STI included younger age (≥50 vs 30-39 years old, adjusted hazards ratio, [aHR]=0.4; 95% CI:0.2-0.8, p<0.001); prior STI infection, (aHR=2.5; 95% CI:1.6-3.8, p<0.001); and heterosexual vs MSM as likely exposure (aHR=0.2; 95% CI: 0.1-0.6; p<0.001).

Table 2.

Risk factors associated with incident STI.

				Univariate Analysis		Multivariate Analysis
Factor	Level	events / pysa	Rate per 100 pysa (95% CIc)	Hazard Ratio (95% CIc)	p-valueb	Hazard Ratio (95% CIc)	p-valuea
Year of cohort enrolment	1999-2009	71 / 829	8.6 (6.8, 10.8)	1.0 (ref)	<0.001	1.0 (ref)	0.074
	2010-2013	39 / 187	20.9 (15.3, 28.6)	2.4 (1.6, 3.6)		1.6 (1.0, 2.6)
Age (years) at study enrolment (2010)	<30	24 / 76	31.7 (21.3, 47.4)	2.2 (1.3, 3.6)	<0.001	1.5 (0.9, 2.6)	<0.001
	30-39	46 / 341	13.5 (10.1, 18.0)	1.0 (ref)		1.0 (ref)
	40-49	30 / 326	9.2 (6.4, 13.2)	0.7 (0.4, 1.1)		0.8 (0.5, 1.3)
	≥50	10 / 273	3.7 (2.0, 6.8)	0.3 (0.1, 0.6)		0.4 (0.2, 0.8)
Sex	Male	106 / 941	11.3 (9.3, 13.6)	1.0 (ref)	0.090	1.0 (ref)	0.479
	Female	4 / 75	5.4 (2.0, 14.3)	0.4 (0.2, 1.1)		1.4 (0.5, 4.0)
Mode of HIV exposure	MSMd	97 / 748	13.0 (10.6, 15.8)	1.0 (ref)	0.005	1.0 (ref)	0.008
	Heterosexual	4 / 139	2.9 (1.1, 7.7)	0.2 (0.1, 0.6)		0.2 (0.1, 0.6)
	PWIDe	8 / 65	12.3 (6.1, 24.6)	0.9 (0.4, 1.9)		1.1 (0.5, 2.5)
	Other	1 / 63	1.6 (0.2, 11.2)	0.1 (0.1, 0.9)		0.2 (0.1, 1.4)
STIf prior to study enrolment (2010)	Yes	62 / 241	25.8 (20.1, 33.0)	3.9 (2.7, 5.8)	<0.001	2.5 (1.6, 3.8)	<0.001
	No	48 / 775	6.2 (4.7, 8.2)	1.0 (ref)		1.0 (ref)
Hepatitis B coinfection (ever)	Yes	3 / 36	8.3 (2.7, 25.6)	0.9 (0.3, 2.9)	0.001	1.2 (0.4, 4.2)	0.061
	No	69 / 786	8.8 (6.9, 11.1)	1.0 (ref)		1.0 (ref)
	Not reported	38 / 193	19.7 (14.3, 27.1)	2.1 (1.4, 3.2)		1.7 (1.1, 2.6)
Hepatitis C coinfection (ever)	Yes	11 / 104	10.5 (5.8, 19.0)	0.9 (0.5, 1.7)	0.923	0.8 (0.4, 1.7)	0.400
	No	94 / 869	10.8 (8.8, 13.2)	1.0 (ref)		1.0 (ref)
	Not reported	5 / 42	11.8 (4.9, 28.3)	1.1 (0.5, 2.8)		1.7 (0.7, 3.8)
CD4 cell count (cells/μL) (time-updated)	0-349	14 / 150	9.3 (5.5, 15.8)	1.0 (0.6, 1.8)	0.418	1.1 (0.6, 2.0)	0.520
	350-499	34 / 214	15.9 (11.3, 22.2)	1.6 (1.0, 2.4)		1.3 (0.9, 2.0)
	≥500	62 / 651	9.5 (7.4, 12.2)	1.0 (ref)		1.0 (ref)
HIV viral load (copies/mL) (time updated)	<103	79 / 882	9.0 (7.2, 11.2)	1.0 (ref)	<0.001	1.0 (ref)	0.095
	103-104	8 / 45	18.0 (9.0, 35.9)	1.6 (0.8, 3.4)		1.3 (0.7, 2.6)
	104-105	15 / 64	23.4 (14.1, 38.7)	2.6 (1.5, 4.5)		1.5 (0.8, 2.9)
	≥105	8 / 25	31.8 (15.9, 63.6)	3.4 (1.6, 7.1)		1.9 (0.9, 4.2)

^apys: Person-years

^bGlobal p-values for year of cohort enrolment, age, CD4 cell count and HIV viral load are test for trend. Other global p-values are test for heterogeneity.

^cCI: Confidence interval.

^dMSM: Men who have sex with men.

^ePWID: People who inject drugs.

^fSTI: Sexually transmitted infection.

Figs 1-3 illustrate the associations with age, prior STI infection, and exposure to HIV by homosexual contact.

Figure 1.

Kaplan-Meier estimates of the probability of remaining STI free by age category.

Figure 3.

Kaplan-Meier estimates of the probability of STI diagnosis by HIV exposure category.

Time-updated CD4 cell count was not independently associated with incident STI (p=0.520). There was an increasing risk of STI with increasing HIV viral load, (V_L≥10⁵ vs V_L<10³, aHR =1.9; 95% CI: 0.9-4.2), with a dose-response trend albeit not reaching statistical significance, (overall test for trend p=0.095). There was also a suggestion that recent recruits to AHOD (2010-2013 vs 1999-2009) were associated with increased risk (aHR=1.6; 95% CI: 1.0-2.6, p=0.074).

Sensitivity analyses assessing risk factors for the four separate STI were broadly similar for independent risk factors (namely, younger age, prior STI, and being MSM), with some differences in the trend for viral load as follows: chlamydia (aHR=1.4; 95% CI: 0.4-4.2, p=0.496), syphilis (aHR=2.6; 95% CI: 0.7-9.5, p=0.108), gonorrhoea (aHR=3.8; 95% CI: 1.0-14.4, p=0.020) and genital warts (aHR=1.8; 95% CI: 0.3-10.8, p=0.642). However, the separate numbers of incident STI were low, limiting statistical power.

Discussion

We have previously found a high incidence of STI diagnoses in AHOD (24). Routine STI screening of HIV+ve patients is recommended (31), and pretty universally implemented over the period from 2010 onwards – especially in the STI clinics included in this analysis. These are HIV+ patients being seen regularly at STI clinics for their HIV treatment. So most patients also had CD4 and viral load measurements.

In the current analysis from 2010, predictive factors for an STI diagnosis included age less than 30 years, being MSM, and a prior history of STI. There was also a suggestion that recent recruits to AHOD might be at increased risk. We also found a non-significant trend towards prior higher viral loads being associated with an STI diagnosis.

A number of other studies, in different populations world-wide, have examined prevalence of STI amongst HIV-infected persons, and a recent review estimated the median point prevalence was 12% (32). However, most studies were hampered by non-standardised procedures, and were unable to examine risk factors for STI. There have been very few studies that designed to measure incidence of STI accurately, or that have collected data on sexual behaviours, drug use, and the use of ARV (33-38)

Overall, the available data support our conclusions that younger age and being MSM are important predictors of subsequent STI. A history of prior STI being a risk factor for future STI diagnosis has not previously been described in HIV-positive populations, although it is an intuitive finding.

Another result in our study was the trend towards association with higher prior HIV-1 viral loads (V_L). Although there is compelling evidence associating increased V_L at time of diagnosis of syphilis or herpes simplex infections (but not other STI), these have generally been considered concurrent measurements, and consistent with biological reactivation of HIV-1, and none were on effective ARV (39-42). The study designs have not enabled differentiation of cause from effect. However, regardless of the possible mechanisms, the association between higher HIV viral load and an STI diagnosis is of concern. STI are associated with unsafe sexual behaviour, and this combined with the association with higher viral load raises the possibility of an increased risk of onward transmission. The question of whether the effects of STI on viral load are mitigated by successful treatment with ARV is currently a topic of intense interest, and increased risk of STI diagnosis at times when patients have raised HIV viral loads is one possible way in which this strategy might fail. However, our findings require verification in other cohorts.

There are a number of factors that need to be considered when interpreting our results. The strength of our study is its carefully defined characteristics. The prospective data are amenable to precise calculations of STI incidence, and predictive factors. However, there are limitations to our study design. First, although our sample is representative of the parent cohort, (24), the latter may not be representative of the general population with HIV infection. Second, although the pharynx accounted for only 7% of known site diagnoses, this does not necessarily mean that oral transmission is rare. It may simply be that a sexually transmitted pathogen at other sites may be more likely to be symptomatic and/or diagnosed.

Finally, even though STI are (by definition), epidemiologic markers of sexual behaviour, we lack data on the likely most important confounders, namely, the number of sexual contacts, and specified sexual behaviours (23,43). Variations in sexual behaviour strategies in MSM such as serosorting and strategic positioning, which aim to minimise HIV transmission risk, may account for the differential effects upon HIV and STI transmission. It is quite possible that these behavioural strategies are effective at reducing HIV transmission risk but have no effect on STI transmission. In addition, we did not collect detailed information on injecting drug use. The increased use of recreational drugs in certain groups of MSM may be an important factor (44,45).

Our results suggest that among HIV positive patients on ARV in Australia, young MSM with a prior history of an STI are at highest risk of a further STI diagnosis.

Figure 2.

Kaplan-Meier estimates of the probability of STI diagnosis by prior STI status.