Researchers are pondering why a once-promising strategy to prevent infection by the AIDS virus has recently run into problems despite success in earlier tests.

The use of antiretroviral drugs to prevent HIV infection — a strategy known as “pre-exposure prophylaxis” or PrEP — is the object of great hope in AIDS research. The strategy worked in a number of clinical trials but unexpectedly has begun to fail.

The abrupt closure last week of one part of a complicated study called VOICE marked the third time in eight months that anti­retroviral drugs did not prevent infection in those assigned to use them.

“People are upset. It’s a big head-scratcher as to why it didn’t work,” said Lynn A. Paxton, an epidemiologist at the Centers for Disease Control and Prevention who oversaw a study completed last summer that showed a strong protective effect of an antiretroviral pill.

Researchers were hoping that if a vaginal gel containing an AIDS drug was as protective in VOICE as it had been in an earlier study, it might be approved for use and on the market by late 2014. Now it won’t be available until 2016 at the earliest — and only if further testing proves it works.

“My first reaction was surprise. My second was disappointment,” said Sharon Hillier, a researcher at the University of Pittsburgh who heads the Microbicide Trials Network, which is overseeing the VOICE trial.

PrEP is considered especially promising for Africa, where 60 percent of infections occur in women, with heterosexual intercourse the main mode of transmission. Clinical experiments are looking at various routes and schedules for delivering the drugs, including pills, microbicide gels, and vaginal rings that could be used for months.

The leading theory as to why the gel failed in VOICE is that women didn’t use it as instructed. In the lingo of medicine, they were “non-adherent.”

About 2,000 South African, Zimbabwean and Ugandan women, most in their 20s, were told to insert a dose of gel every day. For half of them, the gel contained tenofovir, a long-lasting antiretroviral drug. For the other half, the gel contained no drug. The women were given condoms, tested for HIV each month and instructed to practice safer sex.

The assignments were random: Neither the women nor the researchers knew what each person was getting. But an outside monitoring board did, and it periodically reviewed the study’s progress.

In late November, it saw there were exactly the same number of HIV infections — 46 — in each “arm” of the study. The board recommended the trial, which was supposed to end next summer, be stopped.

The percentage of women who became infected over the course of a year — 6 percent — was 50 percent higher than researchers expected — a “heart-stopping rate to see 30 years into the epidemic,” Hillier said.

In October, another part of VOICE was similarly stopped because of “futility.”

About 3,000 women had been assigned to take a daily tenofovir pill; a combination pill containing tenofovir and another drug, emtricitabine, that is sold under the trade name Truvada; or a placebo pill. The tenofovir pill did no better than the placebo, and that arm of the study was dropped. The test of Truvada, however, is continuing.

A somewhat similar study called FEM-PrEP testing Truvada prophylaxis in 2,000 South African, Kenyan and Tanzanian women was stopped in April because it, too, showed no difference between the drug and the placebo.

What’s curious is that each of the strategies being tested had previously been shown to be effective.

A South African study called CAPRISA, whose results were announced in 2010, found 39 percent fewer infections in women assigned to use tenofovir gel. In a study completed last summer in Kenya and Uganda (Partners PrEP), men and women taking tenofovir or Truvada pills had more than 60 percent protection from infection. A study in Bot­swana (TDF2), also finished last summer, found a similar effect for Truvada.

Researchers are scrambling to figure out what explains the contradictory findings. The answers may lie in subtle differences between the groups being studied and the designs of the experiments.

For example, the volunteers in Partners PrEP were long-term couples in which one person was infected and the other not. It’s possible they may have been more motivated to take the pills every day. In CAPRISA, the women inserted the vaginal gel before and after sexual intercourse rather than every day — a targeted approach that may have helped them stick to the program.

“My personal opinion is that the reason the gel was not successful in VOICE is that the daily regimen just wasn’t acceptable for some reason,” said Henry Gabelnick, head of CONRAD, an Arlington-based nonprofit group that developed the gel in collaboration with the pharmaceutical company Gilead. “If it were being used according to instructions, I can’t see how you could have absolutely no difference between the groups.”

The VOICE researchers have twice-a-year vaginal swabs from women using the gel and monthly blood samples from those taking pills. Tests of these for the presence of the drugs may shed light on what percentage of women were actually using the products.

Other studies of vaginal microb­icides and pre-exposure prophylaxis have shown that few people use prevention tools as regularly as they say they do. There’s also evidence that the more “adherent” people are, the more protection they get.

“I think what we have to face up to is that everything in HIV prevention is based in human behavior,” said Robert M. Grant, a researcher at the University of California at San Francisco who oversaw a study that showed a protective effect of Truvada when used by gay men.

What seems clear is that this strategy, once viewed as the easiest and most certain, is going to require a lot of fine-tuning even if it works.