New Haven Needle Exchange Program Was it effective

New Haven Needle Exchange Program • Was it effective in reducing HIV transmission? • Was it cost-effective?

US History of HIV • 1981 CDC MMWR reports unusual pneumonia in 5 gay men in LA • 1982 CDC coins the name AIDS • 1983 HIV virus discovered • 1985 HIV test approved • 1986 AZT approved • 1987 US bans HIV+ immigrants and visitors • 1991 More drugs approved • 1997 Combination therapy becomes standard

Drug use and the spread of HIV • IDU = injection drug user • 1/3 of US AIDS cases can be traced to drug injection • 1/2 of new HIV infections can be traced to drug injection • Spread of HIV among IDUs in NYC – 1985: prevalence close to 0 – 1988: 40% of IDUs infected • Becomes clear by 1987 that IDUs are dominant mode of transmission in New Haven • Reducing spread among IDUs a priority!

Reducing spread of HIV among IDUs • Drug abuse treatment (e. g. , detox, rapid detaox, residential programs) • Maintenance treatment (e. g. , methadone, buprenorphine) • Bleach/education programs • Needle exchange programs

Politics around needle exchange • Proponents: – Reduce HIV spread – Doesn’t increase drug use – Helps vulnerable minority populations • Opponents: – No evidence they reduce HIV spread – Encourages drug use – Admits defeat in war on drugs

History of Needle Exchange • • • 1984 Implemented in Amsterdam 1988 First US program in Tacoma, Washington 1988 Use of federal funds is banned 1990 May Connecticut legislature allows New Haven needle sharing program Nov. Program starts 1991 March initial data reported 1992 Syringe possession decriminalized in Connecticut 1993 Paper wins Edelman Award 1998 Dept of HHS report “NEPs: Part of a Comprehensive HIV Prevention Strategy” Currently ~200 needle exchange programs in US

Early needle exchange studies • Relied on self-reported behavior about reduction in risky behavior • Did not incorporate quantity of needles exchanged

New Haven program • Used needles exchanged 1 -1 (up to 5) for new ones • Program clients and needles had IDs • Date, location, client ID, and needle IDs recorded at distribution and return of needles • Samples of needles tested for HIV

Initial data from random testing • % HIV infected 91. 5% (44/48) needles from “shooting gallery” 67. 5% (108/160) street needles at program start 50. 3% (291/579) program needles (first 15 mo) 40. 5% (147/367) program needles (next 12 mo) • But how does reduced needle prevalence translate into reduced HIV transmission?

Circulation Approach • Needle exchange… – Keeps number of needles in circulation constant – Increases needle turnover, thus reducing the time a needle is in circulation • Shorter circulation time reduces the number of uses (and users) per needle • Thus, decrease in number of infected needles

Notation and Parameter Estimates = 0. 674 = 0. 84 = 0. 1 = 0. 0066 = 20. 5 = 0. 1675 (t) C(T) shared drug injections / client / year probability a needle is bleached before injection removal rate / HIV-infected client / year Pr [ HIV transmission probability | infected needle] needle exchanges / circulating needle / year # clients / #circulating needles fraction of circulating needles infected with HIV (0)=0. 675 HIV prevalence among program clients (0)=0. 636 new infections over time period T / IDU

Model ´(t) = [1 - (t)] (1 - ) (t) - (t) ´(t) = [1 - (t)] (t) - (t)[ + (1 - (t))] C(T) = t=0. . T [1 - (t)] (1 - ) (t) dt • HIV spread: IDU -> needle -> IDU • Malaria spread: humans -> mosquitoes -> humans • Needle exchange and bleach ~ replacing infected mosquitoes with uninfected ones

Effectiveness • One year horizon • No needle exchange ( =0) – C(1)=0. 064 = 6. 4 HIV infections / 100 IDUs / year • With needle exchange ( =20. 5) – C(1)=0. 043 = 4. 3 HIV infections / 100 IDUs / year • Incidence reduced by 33%.

Other Outcomes • No evidence of increase in drug injection • 1/6 of IDUs in program enter treatment • Program attracts minority clients – Local drug treatment: 60% white – Program clients: 60% nonwhite.

Cost Effectiveness • Program cost: ~ $150, 000 / year • Lifetime hospital costs / infection ~ $50 k-100 k • 20 infections averted • Cost saving!

Sensitivity • Assume (t) constant • Approximately, I = • I decreases as – increases – decreases • Results robust (0) + (0)+1 - (0)]

Estimating rate of shared injections • Self-reported 2. 14 injections / client / day • Sharing rate – Self-reported 8. 4% – But 31. 5% of program needles returned by different client than originally issued to – Assume 31. 5% (conservative) • Thus,

Estimating probability of bleaching • Bleach outreach program begun in 1987 • Self-reported 84% use of bleach • Thus, =0. 84

Estimating IDU departure rate • Departures due to – Development of AIDS – Drug treatment (1/6 of clients) – Hospitalization, jail, relocation, … • Assume departures due only to AIDS (conservative) • Mean time to AIDS ~ 10 years • Thus,

Estimating initial conditions (0), (0) • From needle data (108 infected / 160 tested) – Thus, • Other studies on HIV prevalence among IDUs, – 13% seeking treatment – 36% at STD clinics – 67% of African American men entering treatment • Assume at equilibrium before program starts: – ´(0)=0, =0 – Thus, (0)= (1 - ) ] = 0. 636

Estimating infectivity per injection • Studies of accidental needle sticks – Chance of transmission ~ 0. 003 -0. 005 • Drug injection has higher probability • Assume at equilibrium before program start: – ´(0)=0, =0 – • Thus, = = 0. 066 (1 - )(1 -

Estimating the needle exchange rate • Random variable Tr = time until needle returned – Exponential dist with rate – = needle exchanges / needle / year • Random variable Tl = time until needle is lost – Exponential dist. with rate – = rate at which needles lost / year • Random variable L = 1 if needle is legible, else 0 – Bernoulli with probability l – l=0. 86 fraction of needles whose code is legible

Estimating the needle exchange rate • xi = 1 if the ith needle has been returned, else 0 • ti = observed (censored) circulation time of ith needle • If xi=1, then ti=Tr<Tl and L=1 – Likelihood = exp[-( )ti] · l • If xi=0, then Tl<Tr Likelihood: /( ) or (Tr<Tl and L=0) Likelihood: /( ) · (1 -l) or (ti<min{Tr, Tl} and Tr<Tl and L=1) Likelihood: /( ) · exp[-( )ti] · l

Estimating the needle exchange rate max log L = ∑i I(xi=1) log [ exp[-( )ti]l] + I(xi=0)log[ /( (1 -l) /( ) + l[ /( )]exp[-( )ti]] • Max likelihood estimates – = 20. 5 needle exchanges / needle / year – = 23. 1 lost needles / circulating needle / year

Estimating #clients / #needles • = D/N N = number of needles in circulation D = number of IDUs in the program • Assume number of needles constant, N = D = 20. 5 needle exchanges / needle / year = 122. 4 needles distributed / IDU / year • Thus, 20. 5/122. 4=0. 1675