Return on investment in needle and syringe programs in Australia: summary report

3. Effectiveness of NSPs for preventing transmission of HIV and HCV infection

Page last updated: 2002

3.1 Methodology
3.2 HIV seroprevalence
3.3 HCV seroprevalence
3.4 Discussion
3.5 Estimates of injecting drug users living with HIV/AIDS
3.6 Estimates of injecting drug users with HCV and HCV-related deaths

3.1 Methodology

In this study, NCHECR repeated the ecological study of change in HIV prevalence in cities with and without NSPs because several countries have introduced NSPs since the previous study (Hurley et al. 1997). The study also used a similar methodology to assess the effectiveness of NSPs for prevention of HCV infection.

The ecological study design was used to compare HIV and HCV infection among injecting drug users in countries with and without NSPs. Data recorded on HIV and HCV infection included both seroprevalence and seroincidence studies. NSPs were defined as programs distributing needles and syringes, either free or with minimal charge, irrespective of whether they operated from a fixed or mobile site, whether return of a used syringe was mandatory, or the range of other HIV and HCV prevention and treatment services provided.

Several sources were used to identify published reports of HIV and HCV prevalence and incidence among injecting drug users and the implementation of NSPs. All studies with sample size of at least 50 were included. Cities with HIV prevalence studies were only included if HIV was measured among injecting drug users in two or more calendar years. Studies of HIV or HCV among incarcerated injecting drug users were excluded because very few countries provide NSPs during imprisonment. Analysis compared change in HIV and HCV prevalence between cities with and without NSPs at the time of the surveys. For HIV prevalence, city-specific change in prevalence was used in the analysis. For HCV prevalence, however, it was not possible to use city-specific change because relatively few cities had more than one estimate of prevalence.

For each city, the annual rate of change of HIV seroprevalence was estimated by fitting a regression line on a logit scale, with calendar years centred to 1990. The annual rate of change of HIV seroprevalence was also estimated using regressions weighting the comparison of cities with and without NSPs according to one over the variance of the regression estimator (Hurley et al. 1997). The effect of NSPs was assessed by comparing the annual rate of change in HIV seroprevalence in cities that had ever introduced NSPs with cities that had never introduced NSPs. Analyses of HIV seroprevalence were performed comparing all cities, and also in the subset of cities which had an initial HIV seroprevalence of less than 10%, and had results from at least three surveys available over at least three years. Analyses were repeated using regressions weighted according to survey sample size, and also excluding cities in developing countries.

A random effects regression model was used for analyses of HCV seroprevalence because few cities had data points before and after NSPs were introduced, and to allow appropriately for within and between city effects. The effect of NSPs on HCV prevalence was estimated using all data from all cities, excluding studies that used blood stored since 1981, and for cities that introduced NSPs between the first and last available study. A random effects regression model was also used to estimate the effect of NSPs on HCV prevalence using data available for people reporting less than three years of drug injection.

Two sets of analyses were performed to assess the effect of NSPs on HCV incidence. In the first set of analyses, random effects and GEE negative-binomial models were used to compare cohorts in cities with and without NSPs, allowing for within and between city effects in the analysis and for over-dispersion effects. In the second analysis, an overall incidence rate was calculated for each city by summing the numbers of incident infections and person-years of follow-up. Straightforward negative-binomial regression models were then used to compare cities with and without NSPs. Top of page

3.2 HIV seroprevalence

For HIV, there were 778 calendar years of data from 103 cities with HIV seroprevalence measurements from more than one year and information on NSP implementation. Studies were from 67 cities without NSP, 23 cities that implemented NSP between the first and last study, and 13 cities that already had NSP when the studies were carried out.

The overall comparison of annual rates of change of HIV seroprevalence in cities that never introduced NSPs with cities that did introduce NSPs are summarised in Table 3.1.

The analysis found that cities that introduced NSPs had a mean annual 18.6% decrease in HIV seroprevalence, compared with a mean annual 8.1% increase in HIV seroprevalence in cities that had never introduced NSPs (mean difference –24.7% [95% CI: –43.8%, 0.5%], p=0.06). An analysis which weighted each city by one over the variance of the fitted regression line estimated the mean difference in annual rates of change in HIVseroprevalence between cities with and without NSPs to be –32.7% [95% CI: -37.5% to -27.6%] p<0.001. In cities with an initial HIV prevalence less than 10% and with sero-surveys over a period of at least three years, the mean annual decrease in HIV prevalence was 4.0% in cities that introduced NSPs, compared with a mean annual 28.6% increase in cities without NSPs (mean difference –25.3% [95% CI: -50.8%, 13.3%], p=0.2). In these cities, the weighted analysis estimated the mean difference to be –18.4% [95% CI: -32.0% to –2.0%] p=0.030. Because the unweighted results are qualitatively very similar and, for all cities, the point estimate is smaller than the weighted analysis, estimates of NSP effectiveness were based on the unweighted analysis, representing a more conservative approach.

Table 3.1 Estimated annual rate of change in HIV seroprevalence according to weighting of analysis and sample selection for cities without and with NSPs

Table 3.1 is separated into 4 smaller tables in this HTML version for accessibility reasons. It is presented as one table in the PDF version.

No weighting of analysis - all cities

Cities without NSPs
Cities with NSPs
Number
63
36
Mean
8.1%
-18.6%
(95% CI)
(-2.8%, 20.1%)
(-42.6%, 15.3%)
Mean difference (95%CI): -24.7% (-43.8%, 0.5%), p=0.057

No weighting of analysis - cities with initial HIV prevalence <10% and three calendar years of data

Cities without NSPs
Cities with NSPs
Number
19
25
Mean
28.6%
-4%
(95% CI)
(-4.9%, 73.8%)
(-28.5%, 29%)
Mean difference (95%CI): -25.3% (-50.8%, 13.3%), p=0.165

Weighting of analysis - all cities

Cities without NSPs
Cities with NSPs
Number
63
36
Mean
5.1%
-29.2%
(95% CI)
(1.4%, 9.1%)
(-30.8%, -27.6%)
Mean difference (95%CI): 32.7% (-37.5%, -27.6%), p=<0.001 Top of page

Weighting of analysis - cities with initial HIV prevalence <10% and three calendar years of data

Cities without NSPs
Cities with NSPs
Number
19
25
Mean
32.1%
7.8%
(95% CI)
(22.1%, 42.8%)
(-4.8%, 22%)
Mean difference (95%CI): -18.4% (-32.0%, -2.0%), p=0.030

3.3 HCV seroprevalence

For HCV, there were 190 calendar years of HCV seroprevalence data from 101 cities. Data were from 41 cities without NSP, 9 cities that implemented NSP between the first and last study, and 51 cities that already had NSP when the studies were carried out. There were 71 cities with data available for one calendar year, 13 cities with data for two calendar years and 17 cities with data for three or more calendar years. In the 30 cities with HCV seroprevalence data available for more than one year, 60% had already implemented NSPs before the first year of measurement and 30% introduced NSP between the first and last year of measurement.

Median HCV prevalence was 75% (range 24% to 96%) in studies from cities without NSPs and 60% (range 17% to 98%) in cities with NSPs (NPtrend p=0.01). Overall the results indicated little change in HCV prevalence before NSPs were introduced, followed by a decline after the introduction of NSPs (Table 3.2).

If HCV prevalence was 75% or 50% respectively before NSPs were introduced, the results correspond to around a 1.5% or 2% decline in HCV prevalence per annum.

There were 48 studies, from 19 cities, with HCV seroprevalence estimated among people reporting less than three years of injecting drug use. Most studies were carried out in cities with NSPs (43 studies from 16 cities). Five studies were carried out in four cities without NSPs. Before and after NSP data were only available from one city. Sample size ranged from 14 to 303, median 53. Relevant results are presented in Table 3.3 and Table 3.4.

Median HCV prevalence was substantially lower in cities with than without NSPs (19% vs 71%; Table 3.3). On average, HCV prevalence in cities with NSPs was 37% lower than in cities without NSPs using random effects regression modelling (mean (sd): 25% (+18%) vs. 66% (+15%), p<0.001; Table 3.4).

Table 3.2 Estimation of the effect of NSPs on HCV prevalence per year using random effects regression

Table 3.2 is separated into 3 smaller tables in this HTML version for accessibility reasons. It is presented as one table in the PDF version.

Inclusion criteria: all cities and all data points

logit(HCV)
Coefficient
Std. Error
p value
95% CI
Calendar year
-0.008
0.02
0.7
-0.05, 0.04
Years since NSP
-0.079
0.03
0.003
-0.13, -0.02
Constant
1.040
0.24
<0.001
0.56, 1.52
sigma_u
0.5637
sigma_e
0.8082
rho
0.3275
(fraction of variance due to u_i)
(fraction of variance due to u_i)
(fraction of variance due to u_i)
Top of page

Inclusion criteria: all cities and excluding data points before 1981

logit(HCV)
Coefficient
Std. Error
p value
95% CI
Calendar year
-0.0460
0.03
0.1
-0.10, 0.12
Years since NSP
-0.0576
0.03
0.05
-0.11, -0.001
Constant
92.775
59.3
0.1
-23.5,209.1
sigma_u
0.5627
sigma_e
0.8084
rho
0.3264
(fraction of variance due to u_i)
(fraction of variance due to u_i)
(fraction of variance due to u_i)

Inclusion criteria: all nine cities with data points before and after NSP

logit(HCV)
Coefficient
Std. Error
p value
95% CI
Calendar year
0.0446
0.04
0.2
-0.03, 0.11
Years since NSP
-0.1317
0.05
0.01
-0.24, -0.03
Constant
-87.17
70.8
0.2
-226, 51.6
sigma_u
0.2255
sigma_e
0.8245
rho
0.0696
(fraction of variance due to u_i)
(fraction of variance due to u_i)
(fraction of variance due to u_i)

Table 3.3 Summary of HCV prevalence rates among people reporting less than three years of drug injection according to availability of NSPs

Number of studies
Mean HCV prevalence
Standard deviation
Median HCV prevalence
Inter-quartile range
No NSP
5
66%
15%
71%
5%
With NSP
43
25%
18%
19%
21%

Table 3.4 Estimation of the effect of NSPs on HCV prevalence among people reporting less than three years of drug injection using random effects regression

HCV prevalence
Coefficient
Std. Error
p value
95% CI
NSP
-37.06
7.75
<0.001
-52.25, -21.86
Constant
64.50
8.41
<0.001
48.01, 80.98
sigma_u
22.74
(fraction of variance due to u_i)
(fraction of variance due to u_i)
(fraction of variance due to u_i)
sigma_e
8.70
(fraction of variance due to u_i)
(fraction of variance due to u_i)
(fraction of variance due to u_i)
rho
0.87
(fraction of variance due to u_i)
(fraction of variance due to u_i)
(fraction of variance due to u_i)
Top of page

3.4 Discussion

On average, HIV seroprevalence decreased in studies of injecting drug users in cities with NSPs whereas in studies from cities without NSPs, HIV seroprevalence increased. Seroprevalence of HCV also decreased annually in studies carried out after NSPs were introduced. HCV prevalence was substantially lower among people reporting less than three years of drug injection in cities with NSPs compared to cities without NSPs. There was also a non-statistically significant protective effect for HCV incidence in cities with NSPs when compared to those without NSPs.

There are several limitations associated with the ecological study design that should be considered when interpreting the findings from these studies. Seroprevalence data used in the analyses were collected according to different protocols and in diverse populations. It is unlikely that estimates of HIV and HCV seroprevalence in cities with NSPs would differ systematically from those in cities without NSPs, so any such sampling bias would underestimate the effectiveness of NSPs. Because cities were selected for analysis by the existence of published HIV and HCV serological surveys, bias may have been introduced by the decision to do a survey in a particular city at a particular time.

Data on NSPs used in the analyses were based on presence or absence of NSPs rather than on the extent and uptake of these services. Given the positive findings, however, it is likely that inclusion of these parameters would result in a dose response effect on HIV and HCV seroprevalence from NSPs. In addition, it is not possible to separate the effects of implementation of NSPs from the other HIV prevention strategies (Benedikt et al. 2000). In most settings, introduction of NSPs is one component of a broader harm reduction package to reduce the risk of transmission of blood-borne viruses and other harm associated with injecting drug use. Other components include education and counselling, drug dependency treatment strategies such as methadone maintenance therapy, and provision of clean injecting equipment through other outlets in particular pharmacies. Adequate data was not available on individual components of harm reduction strategies to allow an evaluation of the impact of components other than provision of clean injecting equipment (NSPs). Sensitivity analysis has been conducted to determine the outcome of lower rates of NSP effect on HIV.

The excess risk of HIV in people who inject drugs is not due solely to sharing needles, other injecting practices and sexual behaviour patterns increase HIV risk. In contrast to HIV, HCV infection is rarely transmitted through sex. (MacDonald et al. 1996).

It is also possible that HIV seroprevalence may have remained low in some of the cities with NSPs, irrespective of their introduction. HCV infection, however, is universally high among drug injectors. In most countries HCV infection became endemic among this population before there was widespread publicity about transmission of blood borne viruses through injecting practices. Because HCV infection remains asymptomatic for longer than HIV infection, it is also possible that people with HCV infection remain in the population of injectors for longer than those with HIV infection, therefore increasing the prevalence of HCV infection in seroprevalence surveys of injectors.

If NSPs decrease the incidence of HIV and HCV, the rate of increase in seroprevalence will decrease eventually. It is likely that the lower effect of NSP on HCV than HIV seroprevalence can be attributed to the generally higher prevalence of HCV compared to HIV before the introduction of NSPs.

NSPs influence HIV and HCV transmission by increasing use of sterile syringes for injection and lowering the rate of syringe sharing thereby reducing contact with each virus. Some NSPs also provide referrals to drug treatment centres, condoms and education about minimising risk. The difference in rate of change of HIV seroprevalence between cities with and without NSPs and the decrease in HCV prevalence in cities after the introduction of NSPs may not be due solely to NSPs. Nonetheless, the study provides evidence that NSPs reduce the spread of HIV and HCV infection. Top of page

3.5 Estimates of injecting drug users living with HIV/AIDS

The results of the analysis of the effect of NSPs on HIV and HCV prevalence internationally were then applied to estimates of the Australian injecting drug user population to estimate the number of cases of HIV and HCV avoided as a result of NSPs over ten years during the 1990s.

Estimates of past HIV incidence and future AIDS incidence as a result of injecting drug use were obtained using back-projection methods. The method uses observed AIDS incidence data (adjusted for reporting delay), and knowledge of the rate at which HIV-infected people progress to AIDS, to reconstruct the likely pattern of past HIV incidence. It is then also possible to estimate future AIDS incidence. Because of the relatively small numbers of AIDS cases reported due to injecting drug use, back-projection analyses were applied to annual AIDS counts.

3.5.1 With NSP introduction

The number of injecting drug users living with HIV/AIDS is estimated to have peaked in the early 1990s at approximately 470 cases, with a peak in people living with AIDS of less than 100 in the late 1990s. The cumulative number of deaths from HIV/AIDS by 2010 is projected to be approximately 350.

3.5.2 Without NSP introduction

The number of injecting drug users living with HIV/AIDS is estimated to peak in 2000 at approximately 26,000, with a peak in people living with AIDS of almost 3,000 in 2010. The estimated cumulative number of deaths from HIV/AIDS by 2010 is projected to be approximately 5,000.

3.5.3 Prevented through NSP introduction

By the year 2000, approximately 25,000 HIV infections are estimated to have been prevented among injecting drug users since the introduction of NSPs in 1988, and by 2010 approximately 4,500 deaths are projected to have been prevented.

3.6 Estimates of injecting drug users with HCV and HCV-related deaths

The modelled estimate of HCV incidence in Australia that has occurred with NSPs corresponds to a gradual increase in HCV prevalence among regular IDUs until the mid- to late-1980s, followed by a gradual decline to around 52% HCV prevalence in 2000. NSPs were first introduced in Australia in late 1987. Hence, NSPs were assumed to have reduced HCV prevalence among IDUs from 1988 onwards. The pattern of HCV prevalence if NSPs had not been introduced was estimated by assuming that HCV prevalence would have remained constant at 1988 levels from 1988 onwards. From this, a pattern of HCV incidence if NSPs had not been introduced was derived. The model also excludes any reduction of HCV through secondary infection routes.

3.6.1 With NSP introduction

In 2000, the number of injecting drug users living with HCV was estimated to be approximately 200,000 (approximately 150,000 with chronic HCV infection). By 2010 an estimated 11,800 injecting drug users are projected to be living with cirrhosis, and estimated cumulative HCV-related deaths are projected to be 1,800.

3.6.2 Without NSP introduction

In 2000, the number of injecting drug users living with HCV is estimated to be approximately 220,000 (approximately 165,000 with chronic HCV infection). By 2010 an estimated 12,500 injecting drug users are projected to be living with cirrhosis, and estimated cumulative HCV-related deaths are projected to be 1,900.

3.6.3 Prevented through NSP introduction

By the year 2000, approximately 21,000 HCV infections are estimated to have been prevented among injecting drug users since the introduction of NSPs in 1988, (of which approximately 16,000 would have developed chronic HCV); while by 2010 approximately 650 fewer injecting drug users are projected to be living with cirrhosis and 90 HCV-related deaths would have been prevented.
Top of page