Is Truvada, the Pill to Prevent H.I.V., 99 Percent Effective? Don’t Be So Sure
Truvada, the once-a-day pill to help keep people from contracting H.I.V., ison the cover of this week’s New York magazine, and Tim Murphy’s cover story focuses on how the pill is changing sex by drastically reducing gay men’s fear of infection.
It’s not hard to see why: Mr. Murphy writes, “When taken every day, it’s been shown in a major study to be up to 99 percent effective.” This is a claim I hear thrown around a lot among gay men in New York. And it’s wrong. The 99 percent figure isn’t a study finding; it’s a statistical estimate, based on a number of assumptions that are reasonable, but debatable.
Here’s how the estimate was reached: A major study of men who have sex with men, called iPrEx, found that H.I.V.-negative men who were prescribed daily Truvada were 44 percent less likely to contract the virus than those who were given a placebo. But a great many of the subjects did not take their prescribed medication regularly, or at all. Of 48 iPrEx subjects who were assigned to take Truvada and contracted H.I.V. anyway, just four had any detectable level of the drug in their system when they were diagnosed, indicating a 92 percent reduction in risk for people who were actually taking the medicine.
But wait, there’s more: Those four subjects who took Truvada and became infected had its active ingredients in their blood only at levels consistent with taking the drug twice a week. That is, in the study, there were zero apparent cases of subjects taking their pills daily and contracting H.I.V.
The sample size (2,500 subjects, half taking a placebo) wasn’t large enough to establish that Truvada is 100 percent effective when taken daily, especially because only 18 percent of subjects who were given Truvada actually had the medication in their blood at levels that were consistent with daily use. But by looking at the handful of infections among people taking their pills less than daily, the iPrEx researchers were able to build a statistical model of how the risk of infection declines as the number of pills taken weekly rises. In 2012, they estimated that actually taking Truvada every day produces a 99 percent reduction in the risk of H.I.V. infection, despite not directly observing any such infections.
In an email, Dr. Robert Grant, the lead investigator for the iPrEx study, called the 99 percent figure “our best estimate of the H.I.V. risk reduction when men and transgender women who have sex with men use PrEP daily.” (PrEP, or “pre-exposure prophylaxis,” is the practice of using antiviral medication like Truvada to prevent H.I.V. infection, rather than to treat it.) Dr. Grant, a top H.I.V. researcher at the Gladstone Institutes and University of California at San Francisco, noted that lower estimates, such as the 92 percent figure, include results for people who were not taking the drug daily.
All of the H.I.V. researchers and experts I spoke with for this article agree that Truvada works well to prevent H.I.V. infection and that its effectiveness is closely tied to frequency of dosing. But they did not all agree about how much to rely on the 99 percent estimate. Dr. Susan Buchbinder, the director of the Bridge HIV prevention research unit at the San Francisco Department of Public Health and also one of the investigators involved in iPrEx, said, “I don’t think we can be quite so precise about the exact percentage.”
Dr. Buchbinder expressed particular concern that the estimate relies on a comparison of people who chose to take their pills daily with those who did not — unlike in an ideal study condition, where subjects are randomly assigned to take medication or not. “We know that people who take their pills regularly are probably different from people who don’t take their pills regularly on a number of levels,” she said; for example, they could use condoms more frequently or have fewer sex partners, and those behaviors could help them avoid H.I.V. independently of Truvada’s effects.
Dr. Peter Anderson, a professor of pharmaceutical science at the University of Colorado who led the development of the statistical model, responded that the iPrEx team did not see evidence of confounding factors like those hypothesized by Dr. Buchbinder. If subjects who failed to take their Truvada pills daily were at elevated risk for H.I.V. infection, you would have expected to see them become infected more often than subjects in the placebo group, which contained people who ranged widely in how faithfully they took their placebo pills. But that’s not what they found; infection rates were similar for subjects on a placebo and subjects prescribed Truvada but not taking it.
But again, there is a problem of sample size. Dr. Anderson’s model found that patients who never took their prescribed Truvada had a 3 percent chance of contracting H.I.V. in a given year, which is a little lower than what was observed among patients taking a placebo. But that estimate comes with a range of confidence; perhaps the true figure is as low as 2 percent or as high as 7 percent. If the true figure is on the high end, then subjects who chose not to take their pills really did have greater independent risk of contracting H.I.V., in line with Dr. Buchbinder’s concern.
Then there is the issue of the model itself, what statisticians call a regression analysis. It fits a curve to the outcomes for participants in iPrEx, including many infections among people not taking Truvada, a handful among people taking it sporadically, and none among people who appear to have taken it daily. This approach effectively forces the estimate into a box: Infection risk must approach zero as frequency of Truvada dosing increases, because none of the high-adherence subjects became infected. If rare circumstances make infections possible among people with high blood levels of the drug (and two such infections were observed in a study of heterosexual couples, called Partners PrEP), that chance won’t be reflected in the model.
All of which is to say, while there are good reasons to find the 99 percent estimate plausible, the model is not so bulletproof as to say the figure was “shown in a major study.” Tim Horn, H.I.V. Project Director at the H.I.V. think-tank Treatment Action Group, said in an email, “Though it is possible that efficacy can approach 99 percent — language in the lay literature seems to be growing increasingly casual as to what this really means — the data supporting this currently exist in a vacuum.”
As such, Mr. Horn prefers to point to direct study findings, such as the 92 percent figure. Dr. Buchbinder uses qualitative descriptions, such as saying PrEP is “highly effective” and can “drastically” reduce infection risk if taken daily.
Dr. Anthony Fauci, who has headed the National Institute of Allergy and Infectious Diseases since 1984 and was one of the leading researchers involved in developing antiretroviral therapy for H.I.V., has perhaps found the best way to split the difference. Dr. Fauci told me PrEP is “highly efficacious, in my mind easily over 90 percent if you adhere rigidly to it,” without specifically citing the 99 percent figure. He was quick to add, in line with the guidelines from all the United States government agencies encouraging the use of Truvada as a preventive measure, that it’s meant to augment the protection provided by condoms, not to replace them.
From a policy perspective, the difference between 92 percent and 99 percent is not necessarily very important: Either way, PrEP looks to be a highly useful tool for reducing the spread of H.I.V. Any reduction in the 90-percent range looks extremely impressive when you consider a C.D.C. study from last year finding that gay men who said they “always” used condoms were only 70 percent less likely to contract H.I.V. than those who said they never used them. With perfect use, condoms (like PrEP) may approach 100 percent effectiveness, but in the real world, people often use them improperly, or do not use them through the entirety of intercourse, or “forget” to use them despite an intent to do so. The failure of condoms as a prevention strategy is one reason Gov. Andrew Cuomo has made PrEP one of three planks in his strategy to end the H.I.V. epidemic in New York.
But while either level of effectiveness argues for the use of PrEP, “99 percent effective” is a message that differs psychologically in an important way from “90 percent effective.” Bear in mind, PrEP is about reducing the probability of an already low-probability event. If the participants in iPrEx not taking medication had a 3-4 percent chance of contracting H.I.V. in a given year, 99-percent effective daily use of PrEP would reduce their annual risk of infection to 0.04 percent, or 4 in 10,000. That happens to be equal to the risk of accidental death for the average American in a given year. If you don’t walk around preoccupied with the idea that you might be hit by a bus, it’s hard to blame a PrEP user expecting 99-percent effectiveness for effectively discounting the idea he might become infected with H.I.V.
The failure of a 90-percent effective method is terrible luck; the failure of a 99-percent effective method is verging on a freakish accident. Putting too much weight on the 99-percent figure may lead some PrEP users to perceive virtually zero H.I.V. risk when they should really be thinking about very low risk. As Dr. Buchbinder puts it, “I try to be cautious when telling patients what they should expect, just because I don’t want to overpromise something that isn’t the case.”