WHAT HAPPENED TO HIV TRANSMISSION AMONG DRUG INJECTORS IN NEW HAVEN?

Edward H. Kaplan and Robert Heimer

(Chance: New Directions for Statistics and Computing - 1993 Springer-Verlag New York, Inc.)

----------------------------------------------------------------------

Roughly one third of the 242,000 AIDS cases reported to the U.S. Centers for Disease Control and Prevention can be associated in some fashion with injecting drug use. Not surprisingly, people concerned with preventing the further spread of HIV and AIDS are giving a great deal of attention to intervention programs targeting injection drug users (IDUs). One such intervention, needle exchange, may be as promising as it is controversial, according to data from a New Haven, Connecticut study.

Needle exchange programs allow IDUs to exchange their used needles for clean ones. The programs are aimed at both reducing the amount of needle sharing that occurs among IDUs and removing potentially infectious svringes from circulation. A secondary objective of many needle exchanges is to make contact with otherwise difficult to reach IDUs in the hope that many will consider entering drug treatmen! programs and perhaps will cease injecting drugs altogether. Critics of needle exchange contend that by distributing needles one is merely exacerbating existing drug abuse in the population by enabling drug injectors to continue their habits, and, perhaps worse, the availability of clean needles might even encourage noninjectors to initiate drug injection (though this latter concern is somewhat like suggesting that distributing spoons could lead to a rush on hot fudge sundaes at the local ice cream stand).

Although the concepts underlying needle exchange seem simple enough, until recently there was very little empirical evidence that such programs actually work. Past evaluation studies of needle exchanges focused on self reported accounts of risky behavior among program participants. Although it is noteworthy that in many such studies those surveyed did report reductions in needle sharing, critics dismiss the findings because of interview response bias. Drug injectors who are participating in a program dedicated to eliminating needle sharing and who are told repeatedly that needle sharing is a very bad idea owing to the potential transmission of HIV and other needle borne diseases are asked how often they share needles. Is it a surprise that, given the circumstances, program participants report reductions in needle sharing?

The New Haven Needle Exchange

Enter the New Haven study begun in November 1990. Due to the disproportionate number of AIDS cases among drug injectors in New Haven, Connecticut (more than 60%, compared to the national figure of 33%), local health officials, AIDS experts, and community organizers recognized the need for an intervention targeted toward IDUs. Cognizant of both the political objections to needle exchange programs and the methodological objections to needle exchange evaluation research, however, a unique research design emphasizing a statistical view of needle exchange was used.

The evaluation design is guided by a set of empirically testable conjectures regarding the impact of needle exchange on HIV transmission that do not require the use of self reported data. Rather, the evaluation design interviews needles to obtain the necessary information! Needles have a relatively low proclivity to please the interviewer. Before stating these conjectures and the data collection device invented for the New Haven evaluation, let us review a bit of theory based on the physics of needle exchange.

A Circulation Theory of Needle Exchange

First, ask what must happen for a person to become infected via needle sharing. An uninfected drug injector must select a previously used and infected needle. fail to disinfect this needle (with the use of bleach, for example), and inject uTith this needle, and then the infection must be transmitted. It follows that the rate with which HIV infections are transmitted via needle sharing is proportional to the level of infection in shared needles.

Second, ask what the impact of a needle exchange program will be on the behavior of needles. Prior to the needle exchange program in New Haven, it was illegal for IDUs to obtain needles without a prescription; thus, one would expect that scarce needles would be used for as long as possible. The needle exchange, however, allows the legal opportunity to receive new needles for used ones. Among those IDUs who choose to take advantage of this opportunity bv participating in the needle exchange program, one would expect a reduction in the length of time needles remain in use (and available for needle sharing). If the length of time needles remain in circulation is reduced, however, so is the number of times each needle is used; in effect, needles should be sharing fewer people. Consequently, the chance that needles become infected (and are able to infect others) should fall. The performance of a needle exchange over time, as measured by the fraction of needles in circulation that are infected, should thus re

flect the rapidity with which program clients remove used needles from circulation.

Imagine now that somehow one is able to determine how long program needles have remained in circulation over the course of the needle exchange. How might one infer w hether needle circulation times have been reduced significantly from preprogram levels? Consider the following very simple model: Prior to the needle exchange program, individual needles are removed from the population of needles with mean rate r per needle (e.g., due to needle wearout, breakage, resale, loss, or theft), and as a consequence, the average needle circulation time is given by 1/r. Given the opportunity to exchange used needles for clean ones, the per needle removal rate increases from r to r+e (here e is the per needle exchange rate); thus, the average needle circulation time falls to 1/(r+e). What is the percentage reduction in needle circulation times? The answer is [1/r 1/(r+e)]/(1/r) = e/(r+e).

Under the assumptions of this simple competing risks model, needles are either removed from the population with rate r per needle for reasons unrelated to the needle exchange or needles are returned to the program with rate e per needle (so needle wearout, breakage, resale, loss, or theft compete with the needle exchange program as reasons for removing needles from circulation). Thus, the ratio e/(r+e) also represents the fraction of needles circulating among client IDUs that are ultimately returned to the program. Because the needle exchange is an exchange program, a law of conservation of needles applies: The rate with which needles are distributed to clients by the program is balanced by the rate with which needles are delivered by clients to the program. Thus, the fraction of all distributed needles that are returned equals the fraction of all returned needles that are program needles. Linking all of these arguments together, we see that the quantity e/(r+e) also represents the fraction of all distr

ibuted needles that are returned. Thus, the fraction of distributedneedles that are returned provides information regarding the reduction in needle circulation times from preprogram levels. Though it is, of course, possible that r itself would decrease following the introduction of the needle exchange, it still stands to reason that high return rates should be associated with relatively large reductions in needle circulation times and by implication with lower levels of HIV infection in needles.

Now for the conjectures. As clients adapt to the operations of the needle exchange over time, one should observe the following:

1. Needle circulation times should decrease.

2. The fraction of distributed needles returned should increase.

3. The level of infection in returned needles should decrease.

To test these conjectures, however, it is necessary to obtain estimates of needle circulation times, needle return rates, and the level of infection in circulating needles. One clearly requires data describing the behavior of the needles. How might such data be collected?

The Syringe Tracking and Testing System

In New Haven, we have implemented a syringe tracking and testing system (or STT) that operates as follows. First, all participants in the needle exchange receive unique code names, while all distributed syringes also receive unique tracking codes (the program distributes nondetachable syringe/needle pairs, so the terms "needle" and "syringe" are used synonymously). When needles are exchanged, outreach staff record the code name of the client receiving the needles, along with the date (and location) of the exchange, so it is immediately known who received each program needle, when, and where. On the inbound side of the operation, when needles are returned, they are placed in a cannister. Affixed to the cannister is a label displaying the code name of the returning client, as well as the date and location of return. The needle tracking codes are collated with the information on the cannister label, so that for each needle returned one knows how long it has been circulating, whether the client returning the nee

dle was the same client who received the needle, and so forth. Finally, a random sample of returned needles is tested for the presence of HIV proviral DNA using a technique called polymerase chain reaction, or PCR (PCR does not stand for "politically correct research"). A positive result indicates that the needle has been used by at least one HlV infected drug injector. This test is reasonably accurate: in controlled experiments, none of 64 known HIV negative needles tested positive, whereas 28 of 30 known HIV positive needles tested positive.

The STT has been in place since the start of the needle exchange program in November 1990. The needle exchange has served an average of 236 clients per month over the first 16 months of the program. Thus far, 2132 (or 6.4%) of the 33,567 needles distributed between the start of the program and the end of February 1992 have been tested for HIV using PCR. In addition, 160 nonprogram "street needles" collected at the start of the program have been tested.

We computed monthly averages for needle circulation times from the STT. Because clients are allowed to exchange up to five needles per visit and because most needles are, in fact, exchanged in batches, the monthly mean needle circulation times are defined as the average elapsed time between needle distribution and return for returned needles, divided by the average number of needles exchanged per client visit. In addition to the circulation times, the fraction of needles distributed that were ultimately returned to the program was derived for each month, as was the fraction of tested needles yielding positive test results (by month of distribution).

Show and Tell

Figure 1 reports the mean circulation times and return rates obtained from the STT by month. In addition to the raw data, the plots show smoothed versions of these two time series obtained from an algorithm known as "4523H, twice". This smoothing algorithm operates like moving averages but downweights outlying observations, highlighting trends in the data.

The story told by Fig. 1 is remarkably clear: As conjectured, needle circulation times have indeed fallen over time, whereas the fraction of program needles returned has grown. Application of our back of the envelope competing risks model described earlier to the data in Fig. 1 suggests that prior to the needle exchange program, needles were circulating among program cllents Ior roughlv one week before being removed. Because the STT reveals circulation times in the neighborhood of two to three days, it does, indeed, appear that needles are being removed from circulation more rapidly than was the case prior to the implementation of the needle exchange.

With the operational features of the needle exchange apparently working as intended, what about HIV transmission risks? Figure 2 reports the fraction of tested needles yielding HIV positive results

over time, along with the smoothed version of these same data. That the level of infection among the needles tested has decreased is evident from this plot. In fact, that the level of infection in needles has fallen from 67.5% prior to the program (108 of the 160 nonprogram street needles tested positive) to below 45% over the course of the program suggests a reduction in HIV transmission due to needle sharing on the order of 33%. Although a decrease in HIV prevalence among client IDUs over this same time period could also account for this result, such a decrease is extremely unlikely over such a short time period. Moreover, this alternative explanation does not account for the observed reduction in mean circulation times and increase in needle return rates.

In Conclusion

We have formulated a theoretical argument linking the operations of a needle exchange program to testable conjectures, designed a datacollection system that provides us with the necessary data to evaluate the hypotheses proposed, and discovered that the data are, indeed, consistent vith the theory, lending empirical support to the notion that needle exchange programs can slow the spread of HIV infection and, hence, AIDS, among injecting drug users. Absent from our argument is the other salient feature of the New Haven needle exchange, which is the successful placement of IDUs into drug treatment programs. Indeed, placement in drug treatment reduces HIV transmission risks beyond the protection offered by needle exchange, presumably due to the elimination of drug injection for the majority of those who enter treatment during the time that they remain. Because roughly one out of everv six clients who enrolled in the needle exchange has been placed in drug treatment programs, the combined HIV risk reductio

n attributable to the needle exchange is likely to be greater than the 33% figure we have claimed.

Although the analysis reported in this article appears straightforward, it is possible to formalize the circulation theory described by constructing more complex mathematical models. Such a model was constructed much earlier in the life of the needle exchange program, and it is interesting to note that the predictions of that model are consistent with the more complete data described here. Recently, we have also pursued a completely different approach to evaluating the needle exchange that does not rely on circulation theory but rather attempts to estimate HIV infection rates directly from the needle testing data using a statistical technique known as change point modeling. This second approach supports the contention that needle exchange has reduced HIV transmission risks as well.

In closing, we note that although the concept of needle exchange continues to have its share of vocal detractors, there has never been a published scientific study, statistical or otherwise, that demonstrates the feared negati e consequences of needle exchange programs. In the absence of competing data suggesting that needle exchange programs cause more harm than good, and given the seriousness of the threat of HIV transmission via needle sharing in many communities, it seems prudent to allow needle exchange programs to operate under the watchful eve of responsible community organizations or public health authorities where warranted.

Whether one favors or opposes such programs, however, all surely agree that there is no benefit in sponsoring ineffective interventions, especially in an area as resource poor as primary AIDS intervention efforts. Therefore, it is important to continue evaluating these programs with the goal of determining their true efficacy. Needle exchanges are being operated and studied in other American cities such as Tacoma, San Francisco, Seattle, and New York. Evidence continues to be collected from programs in Amsterdam, Canada, Sweden, and the United Kingdom. In New Haven, we are continuing our study of the needle exchange for two more vears under grants from the National Institute on Drug Abuse and the Robert Wood Johnson Foundation. As we have tried to demonstrate, statistical reasoning and modeling play an important and central role in this process.

[Research was funded in part by Grant R01 DA07676 01 from NIDA; additional funding was received from the Robert Wood Johnson Foundation.]

======================================================================

Costs of Needle Exchange in New Haven

It cost approximately $150,000 to operate the New Haven needle exchange during its first year of existence. This includes outreach staff salaries, operating costs for the mobile van, and supplies such as needles, bleach, and condoms. Lifetime health care costs for people with HIV infection (including progression to AIDS) can easily exceed $100,000; this is a public expense. Even assuming only $50,000 in lifetime health care costs due to HIV/AIDS, however, the needle exchange is cost effective if it has averted three infections or more in its first year of operations.

======================================================================

Alternatives to Needle Exchange

As mentioned in the accompanying article, there are those who oppose needle exchange on the grounds that it might encourage drug abuse. Other alternatives that have been offered include:

* JUST SAY NO! This is the status quo policy of doing nothing. Drug injection is viewed as a criminal act, and HIV transmission is viewed as part of the package criminals can expect.

* TREATMENT ON DEMAND! This view advocates for the provision of a sufficient number of drug treatment slots to provide treatment when requested by current addicts. The risks of HIV infection can only be avoided if addicts cease injecting altogether; nothing else will do.

* JUST USE BLEACH! Why bother exchanging needles when ordinary household bleach kills HIV? Educating drug injectors to properly disinfect their injection equipment can be achieved via "bleachand teach" programs.

* DECRIMINALIZE SYRINGE POSSESSION! In most states it is in fact legal to purchase hypodermic needles and syringes without prescription at local pharmacies. This is not the case, however, in those states most associated with the twin epidemics of HIV/AIDS and substance abuse. Allowing legal possession of injection equipment reduces the need for needle sharing and as a result lowers HIV transmission.

It is worth noting that the needle exchange program incorporates many of these ideas. Clients are placed into drug treatment programs when such placement is desired and treatment is possible. Bleach kits are distributed to program clients, along with instructions for the proper use of bleach. The State of Connecticut did decriminalize possession of syringes without prescription as of July 1992, which was largely as a result of the New Haven needle exchange study.

======================================================================

Additional Reading

Des Jarlais, Don C., and Friedman, Samuel R. (1992), "AIDS and Legal Access to Sterile Drug Injection Equipment", Annals of the American Academv of Political and Social Science, 521,42 65.

Heimer, Robert, Mvers, Samuel C., Cadman, Edwin C., and Kaplan, Edward H. (1992), "Detection by Polymerase Chain Reaction of Human Immunodeficiency Virus Type I Proviral DNA Sequences in Needles of Injecting Drug Users", Journal of Infectious Diseases, 165, 781 782.

Kaplan, Edward H. (1992), "Needle Exchange or Needless Exchange? The State of the Debate", Infectious Agents and Disease, 1, 92 98.

Kaplan, Edward H., and O'Keefe, Elaine (1993), "Let The Needles Do The Talking! Evaluating the New Haven Needle Exchange", Interfaces, 23, 7 26.

Thompson, Dick (1992), "Getting the Point in New Haven," Time, 139, 55 56.