Twice-a-Year HIV Prevention: What Long-Acting Injections Change

The question comes up at nearly every seminar I give, usually from a clinician in the back row, usually phrased some version of: "If a shot twice a year is close to a hundred percent effective, why are we still having this conversation?"

The answer I used to give was the tidy one. Efficacy isn't effectiveness. Trials are not programs. Access takes time.

All true, and all incomplete. What I've come to think, after a few years of watching this literature and arguing about it with people who run actual clinics, is that long-acting PrEP changes the shape of the scientific question rather than just the size of the effect. Adherence stops being the dominant variable and something else takes its place: supply, service delivery, diagnostics, and resistance surveillance. That's a different research agenda, and a lot of us are still writing grants for the old one.

What the agents actually do

Two drugs carry most of the weight here, and they're pharmacologically very different.

Cabotegravir is an integrase strand transfer inhibitor formulated as a nanocrystal suspension for intramuscular gluteal injection, dosed every two months after a loading schedule. It blocks integration of proviral DNA. The long action comes from slow dissolution at the injection depot, not from any unusual property of the molecule itself.

Lenacapavir is the more interesting one mechanistically. It's a capsid inhibitor, binding the HIV-1 capsid protein hexamer and interfering with multiple stages — nuclear import, capsid assembly and disassembly, and the orderly uncoating that reverse transcription depends on. Because it has no close relative in clinical use, cross-resistance with existing regimens is minimal. Its half-life and solubility profile allow subcutaneous dosing roughly every 26 weeks, which is why "twice a year" entered the vocabulary at all.

Both drugs share a property that matters enormously for prevention and that we under-discuss: a long pharmacokinetic tail. After the last injection, drug levels decline slowly through sub-protective concentrations over months. If someone acquires HIV during that window, they're being treated with functional monotherapy, which is exactly the condition under which resistance emerges. Hold onto that; I'll come back to it.

The strongest evidence: randomized trials

Start with cabotegravir. HPTN 083 enrolled cisgender men and transgender women who have sex with men and compared injectable cabotegravir against daily oral tenofovir disoproxil fumarate/emtricitabine. Incidence in the injectable arm was roughly two-thirds lower. HPTN 084 ran the same comparison in cisgender women across sub-Saharan Africa and found a substantially larger difference — on the order of a 90 percent reduction, with a hazard ratio around 0.12.

That asymmetry told us something important before the lenacapavir data arrived. The oral comparator performed much worse in the women's trial, consistent with a long history of low adherence to daily oral PrEP in that population. Long-acting injection didn't make the drug more potent. It made the drug present.

Then the PURPOSE program. PURPOSE 1, in adolescent girls and young women in South Africa and Uganda, reported zero incident infections among roughly 2,100 participants randomized to twice-yearly lenacapavir, with superiority over both background incidence and oral comparators. PURPOSE 2, in cisgender men and gender-diverse people who have sex with men across several countries, found two infections among roughly 2,180 participants, a reduction in the mid-nineties relative to background incidence.

Zero events in an arm of that size is rare enough in prevention science that I remember exactly where I was when a colleague forwarded the press release. I also remember my second thought, which was: what's the counterfactual?

Why the counterfactual deserves scrutiny

Placebo control is no longer ethical in HIV prevention, so PURPOSE used a background HIV incidence estimate derived from recency testing among screened-but-not-enrolled individuals. It's a defensible design and it was reviewed carefully. It's also not a randomized comparison, and the point estimate of efficacy depends on an externally constructed denominator.

I don't think this undermines the conclusion. Zero infections is zero infections, and the oral arms within the same trials provide an internal anchor. But when a student tells me the drug is "100 percent effective," I push back, because that number is an artifact of a specific comparator construction, not a biological ceiling. The honest statement is that twice-yearly lenacapavir prevented essentially all incident infection in two large randomized trials under study conditions.

If you want to follow the primary literature rather than the summaries, a running search for lenacapavir pre-exposure prophylaxis and one for long-acting cabotegravir PrEP will keep you current better than any review I could write.

What's established versus what's emerging

Established, in my reading:

• Both injectables outperform daily oral PrEP in head-to-head randomized comparison, with the margin largest in populations where daily adherence has historically been poorest.

• Efficacy is driven primarily by sustained drug exposure rather than by superior intrinsic potency at the exposure site.

• Regulatory and normative bodies have moved. Cabotegravir has been approved for prevention since late 2021; lenacapavir received US approval for PrEP in mid-2025, and the World Health Organization issued guidelines recommending twice-yearly lenacapavir as an additional prevention option later that year.

Emerging, and treated as settled far too often:

• Effectiveness in routine delivery. Early implementation cohorts for cabotegravir suggest that injection visit adherence is good but not perfect, and that late or missed visits cluster in exactly the groups with the least stable housing, transport, and clinic access. We have very little comparable data for a 26-week product yet.

• Resistance after breakthrough. In HPTN 083, several breakthrough infections involved delayed diagnosis — antigen/antibody assays can be blunted by partial drug suppression — and integrase resistance emerged in a subset. Lenacapavir breakthroughs in PURPOSE were too few to characterize well. The signal to watch is capsid mutation in people who seroconvert during the tail.

• Pregnancy, lactation, and adolescents. Data are accumulating, including from pregnancies occurring during the trials, but the evidence base is thinner than the enthusiasm.

• People who inject drugs. Essentially untested for injectable PrEP in randomized designs. This is a real hole.

Early and translational work worth tracking

Ultra-long-acting formulations are in development that would stretch dosing toward annual administration, including intramuscular lenacapavir formulations and implantable delivery. Broadly neutralizing antibodies continue to be tested for prevention, with efficacy so far dependent on viral sensitivity to the specific antibody — the antibody-mediated prevention trial literature is the place to start if you're new to it. Combination approaches pairing a capsid inhibitor with a long-acting antibody are plausible but currently theoretical.

I'd put all of this in the "interesting, unproven" column. A postdoc emailed me last spring asking whether she should pivot her modeling work toward annual dosing scenarios. My advice was to model the 26-week product properly first, including missed-visit distributions, because we don't yet have good empirical inputs even for the thing that already exists.

The access problem is the scientific problem

This is where I part company with a lot of commentary.

Gilead signed voluntary licenses with several generic manufacturers covering a large list of high-incidence, resource-limited countries, and procurement arrangements involving the Global Fund and PEPFAR have been announced to bring lenacapavir to lower-income settings. That's genuinely faster than the antiretroviral access timelines of the 1990s and 2000s, and the people who fought for it deserve credit.

But a license is not a vial, and a vial is not a program. Several things have to be true at once for twice-yearly PrEP to change population incidence: manufacturing scale-up at a price health systems can absorb, cold chain and injection capacity in clinics that are often already stretched, demand creation among people who may distrust the health system for good historical reasons, and surveillance infrastructure capable of detecting resistance in the tail.

Middle-income countries with concentrated epidemics — much of Latin America, parts of Eastern Europe and Asia — sit awkwardly between the licensed territories and the markets that can pay. I expect that gap to be the subject of a lot of health policy research over the next few years, and I'd encourage people entering the field to work there rather than adding to the pile of efficacy commentary.

There's also a design question I keep raising with trial groups and keep losing arguments about: we are still running prevention trials that enroll in communities and then withdraw the product at trial end. A twice-yearly injectable makes that transition sharper, not softer, because participants leave with months of declining drug on board. Protocols should specify the step-down plan as carefully as they specify the primary endpoint. Registered protocols on ClinicalTrials.gov vary widely on this point.

What I'd tell someone starting work in this area

Don't write another efficacy review. The trials are strong, well reported, and widely summarized. Marginal value there is close to zero.

Do go after the places where the evidence is genuinely thin. Visit adherence over multi-year horizons. Diagnostic performance in people with partial drug exposure — this is a lab question with enormous programmatic consequences, and it's underfunded. Resistance surveillance protocols that are feasible in settings without routine genotyping. Cost-effectiveness modeling with realistic, not aspirational, delivery assumptions. Implementation work with people who inject drugs, where we have almost nothing.

And be careful with your framing. I reviewed a manuscript last year that reported "near-perfect protection" in the abstract and buried the comparator construction in a supplementary table. The editor and I both flagged it. The authors were not being dishonest; they were writing the sentence that everyone else was writing. That's how a field drifts.

Practically: if you're building a literature base in a fast-moving area like this one, the mechanics of keeping up matter more than people admit. I've watched students lose weeks because their notes, their reading, and their drafts lived in three places that didn't speak to each other. An AI-native research workspace like PACR is designed around that problem — search, reading, summarizing, writing, and collaboration in one place, so the context you build while exploring the literature carries through to the manuscript instead of evaporating between tools. For a topic where a WHO guideline update or a new implementation cohort can land between your first draft and your revision, that continuity is worth something. You can set up a workspace and test it on a single review before committing to it.

The thing I keep coming back to is this: for the first time in the history of HIV prevention, the limiting factor isn't whether the intervention works. We know it works. Whether it reaches a nineteen-year-old in KwaZulu-Natal or a man who sells sex in Kyiv is now a question about procurement, clinic hours, and political will — which means it's a question researchers can study, and should.

That's a better problem than the one we had. It's still a problem.