• This summer, two major studies found that no HIV-positive participants transmitted the virus to their HIV-negative partners when they had a fully suppressed viral load, even though they often had sex without a condom. So does this mean there’s a guarantee that people who are undetectable cannot transmit?
  • A recent meta-analysis of various studies found that men who have sex with men (MSM) on pre-exposure prophylaxis (PrEP) have very high rates of sexually transmitted infections (STIs) compared with other MSM. So does this mean that taking PrEP causes people to contract more STIs?

Properly interpreting such research papers—understanding what they can tell us as well as what they can’t—hinges on understanding certain principles of scientific reporting. This doesn’t require an advanced degree by any means. But it is helpful to understand a few key scientific and mathematical concepts when you’re trying to wrap your head around a science-based news story.

Confidence intervals:

You may be aware of a recent social media campaign by a group called Prevention Access Campaign (PAC) touting the claim that “Undetectable = Uninfectious.” This slogan is based on the lack of HIV transmissions seen in the HPTN 052 study and the ongoing PARTNER study. None of the HIV-positive individuals with a fully suppressed viral load transmitted to their HIV-negative partners in either study. In PARTNER, MSM couples reported 22,000 sex acts while heterosexuals reported 36,000.

That’s a lot of condomless sex with no transmissions. However, researchers will never be able to say without a doubt that people who are undetectable are uninfectious. To understand why, consider an important concept in statistical math known as the confidence interval.

When the authors of scientific studies publish estimates about a particular outcome—in this case the estimated rate of HIV transmission over time when individuals have fully suppressed HIV—they provide an accompanying confidence interval. The confidence interval means that based on the available evidence, the study authors are 95 percent certain that the actual figure lies within a certain range. The greater the body of evidence, the more they can narrow the confidence interval around their estimate.

You cannot eliminate a confidence interval; you can only keep making it smaller. Its persistence, a hard-and-fast acknowledgement of the unshakable nature of uncertainty, means that any scientific statements about the risk of transmitting HIV with a fully suppressed virus will always come with at least a whisper of doubt. Researchers can only become more and more certain about their assessment of the risk.

Another way of putting this is to say that there are no 100 percent guarantees in science, and you cannot prove a negative. Even if participants in PARTNER engaged in a hundred trillion sexual acts while on fully suppressive HIV treatment without transmitting the virus, there is still a chance that one of them could do so during the next act.

The PARTNER study authors already have enough evidence to estimate that the risk of transmitting with an undetectable viral load is zero. But their confidence intervals for various sex acts vary quite widely based on the size of the data they’ve collected so far.

Looking at all types of sex in the context of a positive partner having a fully suppressed virus, the authors’ confidence interval tells us that if 10,000 couples were followed for one year, between zero and 30 of them would transmit the virus.

Because the study has less follow-up data for MSM, the comparable confidence interval for the risk of transmission for anal sex with ejaculation when the positive man is the insertive partner (the top) is between zero to 270 transmissions.

As PARTNER follows its participants for longer, these confidence intervals will narrow. And so with time, so long as there are no new transmissions, science will likely be able to say with greater and greater certainty that the risk of transmitting the virus is vanishingly small.

Correlation vs. causation and why it is hard to determine whether PrEP raises STI rates:

It can be very difficult for scientific studies to come even close to proving that one factor causes a certain outcome. In a recent meta-analysis comparing STI rates among MSM in studies of PrEP with STI rates of MSM participating in other studies, scientists found that those on PrEP were 25 times more likely to be diagnosed with gonorrhea, 11 times more likely to be diagnosed with chlamydia and 45 times more likely to be diagnosed with syphilis.

The paper could not say for certain that going on PrEP or entering a study about PrEP caused men to experience higher STI rates. They could say only that there was an association, or link, between PrEP use and high STI rates.

Another way of putting this is to say that overall, people who take PrEP contract STIs at high rates. Or, more accurately, people who enter studies about PrEP have high STI rates.

Any number of what are known as confounding factors could have contributed to the high STI rate among those participating in the studies of PrEP.

It makes perfect sense that groups of MSM on PrEP would tend to contract a lot of STIs, since the very high-risk sexual behaviors that make men good candidates for PrEP in the first place also put them at risk of gonorrhea, syphilis and chlamydia. And the good news is that higher-risk guys are in fact more interested in PrEP; so hopefully, their taking Truvada will indeed help slow the transmission of HIV.

Often authors of research papers will try to adjust, or control, their data to account for confounding factors and thus provide a clearer sense of cause and effect, or at least whether one variable can help predict another—say if obesity predicts heart attack risk. In the United States, there is considerable overlap between race and socioeconomic status, for example; so controlling for such factors can help researchers better tease apart how various influences play into the outcome they are examining.

The authors of the PrEP paper did not adjust the STI rate data for any confounding factors. They did theorize that the difference in STI rates between MSM in PrEP studies compared with those MSM in other studies could be in part a function of the fact that to be eligible to participate in the PrEP studies men had to be engaging in high-risk sex.

Their analysis was not structured to examine changes in the STI rate over time or to properly examine potential causes of the high rate. For a study to explore whether starting PrEP changed a group’s STI rate, it would first have to establish the STI rate before they started PrEP and then see how it changed over time.

But even then the study wouldn’t be able to answer whether the STI rate would have followed a similar trajectory regardless of PrEP’s introduction. Researchers would need a comparison group of similar people who did not start PrEP to see the pattern in their STI rate over time.

But if the members of such a group are not selected randomly making such a comparison could be especially likely to introduce confounding variables. The differences in the sexual behaviors influencing the choices of those MSM who decided of their own accord to take PrEP or enter a PrEP study, on the one hand, and those who decided not to could bias the results of a comparison between these two groups.

The PROUD study of PrEP among high-risk MSM in England did fashion such a comparison group—in order to determine how well PrEP prevented HIV—by having a portion of the study population receive Truvada on a deferred basis. That way all the participants entered the study under the same protocol, but some were randomized to receive PrEP later than others. Comparing the STI rates of those taking PrEP with those waiting for it—the rates for all were very high—the researchers did not see that starting PrEP was associated with an uptick in such infections.

Other studies of PrEP generally haven’t seen a changing STI rate among participants over time. However, a study of a non-daily dosing protocol did see a declining condom rate after participants switched from the placebo to the open-label phase. This is the most solid scientific proof to date that, at least among some MSM, going on PrEP does lead to greater sexual risk taking, or what is known as risk compensation.