AIDSmap has a story up about a study just published about gay men, with the title "Having older sexual partners increases HIV risk for younger gay men." If you didn't read closely, you might think that the researchers had actually done a study that might buy them the cache to make such a claim. I thought I would take a moment here to dissect what the study actually found, and what we can(not) safely conclude based on those findings.

First, the study wasn't just about "gay men." Though many of the men involved did identify as such (74%), they recruited "men who have sex with men" - which includes a variety of other kinds of guys. Second, and more importantly in my book, they study involved a small sample of 74 men in North Carolina. That's right, 74 people are the basis for that rather startling headline. The meat of their claim is this statement here:

Men with PHI had partners on average 6 years older than themselves, whereas uninfected men's partners were 4 months their junior (P , 0.001). After adjusting for race, sex while intoxicated, and having a serodiscordant/serostatus unknown partner, a participant had twice the odds of PHI if his sex partner was 5 years his senior (odds ratio 2.0, 95% confidence interval: 1.2 to 3.3)

Let me try to translate that into slightly more clear English:

On average, the men that recently HIV-positive participants reported having sex with before they seroconverted were 6 years older than them. By comparison, HIV-negative men's recent partners were 4 months younger than them. When we take into account these guys' race, whether they reported being intoxicated while having sex, and whether they had sex with poz guys or guys whose HIV-status they didn't know, a participant was twice as likely to be in the HIV-positive group if he reported having partners five years older than him.

I hope that was somewhat more readable. In any case, what they're trying to say here is that because there is an association between having sex with older guys and being HIV-positive yourself, there may be causal relationship between those two things. But of course their data doesn't actually show this. Indeed, what there data merely says is that - of the 74 guys they interviewed - the ones who were poz were more likely to report having sex with older men before they were infected. The data they are relying on here is a survey asking men about their three previous partners before infection if poz or their three previous partners before their enrollment if neg.

But of course we can infer other possible theories here, and not just the causal link that they're alleging here. Or at least it could be missing pieces of the story. Now, importantly they did control for race, substance use, and sex with poz and unknown status guys. These are all important factors that should indeed be taken into account, so we don't mistake the difference in age between partenrs for the causal factor when it's actually other factors.

However, in a strange move that they don't explain, they actually don't include in their associative model condom use with their previous serodiscordant or serounknown partners, which actually varies considerably between the two groups (click to embiggen):

Just to clarify: This data isn't about condom use with any last partner, but only the previous partner who was either HIV-positive or for whom you did not know their status. That's a pretty important piece of data, and the different shown above is striking. But there's a bit of a statistical trick going on in their write-up. While the difference above is stark, non-condom use with a previous poz or HIV-unknown partner is not actually what they include in their model. Rather, for their odds ratio calculation, they include whether or not the participant reported having sex with a serodiscordant or serounknown partner at all. But this of course is the less meaningful datapoint -- what actually matters for risk is whether you used a condom with that last positive or unknown partner!

I don't like being duped, and seems to me to be what's going on here. They are using statistical trickery to manipulate our interpretation of their findings. What other conclusion can we reach? They have more meaningful data, and yet they willfully leave it out of the model and never address that exclusion. There is just no way that this was an accident -- they obviously ran the model with the condom use data and it was likely weaker than the one they decided on. Thus, I'm suspicious.

But beyond the statistical manipulation, this is a study of 74 men in North Carolina. Let's not get ahead of ourselves and start making any big claims about "gay men" universally or even across the United States.

Moreover, what is most troubling for me about this data is what they clearly want us to do about it: Prevent younger men from having sex with older guys. This to me is the next logical step in Public Health logic, and it's the one that had a bunch of us gay men's health activist up in arms a year or two ago when Michael Scarce reported on the Ning that STOP AIDS was working on an intervention research project that would attempt to do just that. It's important to know the ways in which HIV gets transmitted, and to be able to implement culturally tailored prevention efforts that come out of that knowledge. But dissociative mixing is not the answer.

Citation:

Hurt CB et al. Sex with older partners is associated with primary HIV infection among men who have sex with men in North Carolina. J Acquir Immune Defic Syndr, online advance publication, 2010.