Evidence of the effectiveness of AIDS prevention efforts has been constrained by the use of poor study designs, small sample sizes, short evaluation periods and a lack of control groups [1]. However, ‘These same flaws are common to a surprisingly high percentage of research in all areas of medicine, and are not necessarily more problematic in the AIDS area’[1]. These flaws are evident and have been acknowledged widely in the literature on the effectiveness of needle and syringe programmes (NSPs) [2,3]. In a recent editorial, Amundsen [4] relies on these limitations to question the status of NSPs as a ‘superior tool’ for HIV prevention and suggests that more emphasis be placed on testing and counselling to reduce risk behaviour in injecting drug users (IDUs). However, the evidence base for voluntary testing and counselling (VTC) is relatively weak compared to that supporting NSPs. VTC is most effective in reducing risk behaviours among serodiscordant heterosexual couples and those testing HIV positive [5]. While some studies show reductions in risk behaviour following a positive diagnosis [6], many indicate ongoing risk behaviour [7,8]. At best the effects of VTC are variable, with some [8–10], but not all [11], randomized controlled trials reporting a dose–response relationship, with greater reductions in risk associated with more intensive counselling. Evidence of beneficial effects in particular populations is especially weak [12], and many IDUs continue to engage in risk behaviours despite knowledge of their positive serostatus [13,14]. While some studies indicate that HIV-positive IDUs report less risk behaviour than those not tested or infected [15], a randomized trial [16] and a prospective study [17] found that VTC was not associated with a reduction in injection risk behaviour. To date, no trials have evaluated the impact of testing and counselling on HIV transmission, compared with no testing and counselling. However, prevention research is increasingly shifting its focus towards group and structural interventions that address the social and environmental contexts of risk [18] and the role of social networks and group dynamics [19]. The evidence to date suggests that VTC is clearly necessary, but not sufficient, in preventing the transmission of HIV and other blood-borne pathogens. As in other areas of medicine, evidence of the effectiveness of public health interventions is rarely clear-cut and policy makers have traditionally rejected putting all the eggs into a particular basket in favour of a range of interventions which, in turn, have different costs and varying levels of effectiveness. A recent study which examined the allocation of resources to NSP, methadone maintenance and condom distribution programmes targeting different population groups found that best allocations varied by epidemic scenario. In a setting characterized by 40% HIV seroprevalence among IDUs, allocating approximately 75% of the budget to NSP and 25% to a condom availability programme maximized the number of HIV infections averted. In a lower prevalence setting (5%), the best allocation expended almost all funds on methadone maintenance for HIV-positive IDUs [20]. NSPs are clearly a critical component of an effective harm reduction response. Unfortunately, questioning the efficacy of this vital public health tool on the basis of a lack of gold standard evidence may serve to undermine current attempts to scale up NSPs in settings where they are needed most.