In 1990 a diagnosis of HIV infection was a prediction of death. Today a twenty-year-old who starts antiretroviral therapy in a high-income country can expect to live into their seventies — a life expectancy approaching that of the general population [1]. That is one of the largest reversals in the history of medicine, and it happened in about fifteen years.
Two things about it are worth understanding properly. The first is that the transformation is incomplete and unevenly distributed: the same analysis that found life expectancy rising from 36.1 to 51.4 further years between 2000 and 2007 also found it lower for people with a history of injection drug use, for non-white patients, and for anyone starting treatment with a CD4 count below 350 [1]. The second is that the drugs turned out to do something nobody designed them for. Suppressing the virus in one person prevents transmission to another — which is the subject of the centerpiece, and the single most consequential thing to know about HIV today.
Start here: what HIV is and what it does
HIV is a retrovirus. It enters a cell, copies its RNA genome into DNA with a reverse transcriptase, and integrates that DNA into the host chromosome, where it becomes a permanent part of the cell. Its principal target is the CD4+ T lymphocyte — the cell type that coordinates the rest of the immune response.
Untreated infection follows a characteristic course. A brief acute illness accompanies very high viraemia. Then the immune system partially controls the virus — cytotoxic CD8 T-cell activity is associated with the fall in viraemia during primary infection [2] — and a years-long clinical latency follows during which CD4 cells are steadily destroyed and replaced, with the balance tipping slowly the wrong way. When the CD4 count falls far enough, the infections and cancers that a competent immune system suppresses begin to appear. That state is AIDS, defined since 1993 by a CD4 count below 200 cells per cubic millimetre or by the presence of an AIDS-defining condition [3], and its complications are catalogued in dedicated guidelines [4].
The virus crossed into humans from primates, and the origins of the pandemic have been reconstructed in detail [5].
The most important opportunistic infection globally is tuberculosis. The association was established early: a prospective study of injection drug users found that HIV infection dramatically increased the rate at which latent tuberculosis reactivated [6], and tuberculosis in people with HIV behaves differently enough to warrant its own clinical literature [7]. This coinfection remains the largest single cause of death among people with HIV.
Pillar 1: measurement and diagnosis
Testing, and the window period
Modern tests detect both antibody to HIV and the viral p24 antigen, which appears earlier. The practical point for anyone being tested is the window period: the interval between infection and a detectable result. A test taken too soon after an exposure can be negative in someone who is infected — and, because viraemia is extremely high in acute infection, that is precisely the period of greatest infectiousness.
Getting people tested at all remains the binding constraint in much of the world. Community-based approaches substantially increase uptake compared with clinic-based testing [8], and the reason testing is avoided is often not logistical: HIV stigma is a measurable, structured phenomenon with its own instruments and its own literature [9] [10], and it obstructs every subsequent step [11].
That sequence — diagnosed, linked to care, retained, on treatment, virally suppressed — is the care cascade, and framing it that way made visible how much loss occurs at each step [12].
The two numbers
Once someone is diagnosed, two laboratory values do almost all the work, and they answer different questions.
Viral load is the concentration of HIV RNA in plasma. It answers: is the treatment working? The target is "undetectable" — below the assay's limit of quantification, typically 50 or 20 copies per millilitre. Viral load responds within weeks of starting effective therapy and is the measure by which regimens are judged and failure is detected.
CD4 count is the concentration of CD4+ T cells. It answers: how much immune damage is there, and what is this person at risk of right now? It determines opportunistic-infection prophylaxis and it recovers slowly — over months to years — after viral suppression. Both matter for prognosis: the effect on life expectancy depends on the CD4 and viral-load response to therapy, not on either alone [13].
The distinction is worth stating plainly because it is routinely confused. Viral load measures the virus; CD4 count measures the damage. Treatment fixes the first quickly and the second slowly, and someone can have an undetectable viral load and still be immunologically vulnerable.
Resistance testing
HIV replicates fast and error-prone, so resistant variants arise readily when suppression is incomplete. Genotypic resistance testing guides regimen selection, particularly after treatment failure or in settings with substantial transmitted resistance; treatment guidelines have been built around this since the earliest combination era [14] [15].
Centerpiece: a simple simulatable model of viral load and transmission
The question underneath modern HIV policy is whether treating one person protects another. It has a quantitative answer, and the answer came from a cohort study before it came from a trial.
A community-based study in Rakai, Uganda followed 415 couples in which one partner was HIV-positive and one was not, for up to 30 months. Ninety of the initially negative partners seroconverted, an incidence of 11.8 per 100 person-years. The infected partners whose partners seroconverted had a mean plasma HIV-1 RNA of 90,254 copies per millilitre against 38,029 among those whose partners did not, and the relationship was graded: each log increment in viral load carried a rate ratio of 2.45 for seroconversion (95 percent confidence interval 1.85 to 3.26). There were no transmissions at all among the 51 subjects with viral loads below 1,500 copies per millilitre [16].
That gives the model directly. Transmission rate scales as 2.45 raised to the log₁₀ of viral load, with one free constant setting the vertical scale — and that constant is computed from the study's own numbers rather than chosen, by recovering the cohort's mean viral load as the seroconversion-weighted average of the two group means and anchoring the curve to the reported overall incidence there.
The dose-response is the mechanism behind everything that follows. But the figure's right-hand panel is the part worth dwelling on, because it is where the model stops being trustworthy in an instructive way.
Extrapolating the log-linear curve down to suppressed viral loads predicts a small but clearly non-zero transmission rate — about 1.4 per 100 person-years below 200 copies per millilitre. What the studies measured was less than that. The PARTNER study followed 888 serodifferent couples through 1,238 couple-years of condomless sex with the positive partner below 200 copies per millilitre, across roughly 58,000 sex acts, and found zero phylogenetically linked transmissions, with an upper 95 percent confidence limit of 0.30 per 100 couple-years [17]. A meta-analysis of serodiscordant cohorts likewise found zero transmissions below 400 copies on treatment, upper limit 1.27 per 100 person-years [18]. And the randomised evidence agrees: in HPTN 052, of 28 virologically linked transmissions, only one occurred in the early-therapy group — a hazard ratio of 0.04 [19].
The measured ceilings sit below the extrapolation, by more than fourfold in PARTNER's case. That matters for how the claim should be understood. "Undetectable equals untransmittable" is an empirical finding, not a consequence of extending this curve downward. The dose-response explains why suppression should help enormously; the trials are what established that the residual risk is not merely small but undetectable in practice.
That is the teaching point, and its consequence is the central fact of modern HIV medicine: because transmission scales with viral load and suppression drives viral load to zero, treatment and prevention are the same intervention. A person on effective therapy is protecting themselves and everyone they have sex with, simultaneously.
Three honest limits. The anchor uses arithmetic mean viral loads because that is what the source reports, while the model is linear in log viral load — an approximation, and the reason the anchor should be read as setting scale rather than as a fitted parameter. The Rakai cohort was heterosexual, and per-act risks differ by exposure route; PARTNER's upper limit for condomless anal sex was higher, at 0.71 per 100 couple-years [17]. And plasma viral load is a proxy for genital-compartment virus, which is what actually transmits and which can diverge.
Pillar 2: treatment and prevention
Antiretroviral therapy
Treatment is combination therapy, because monotherapy selects resistance almost immediately. The drugs attack distinct steps of the viral life cycle — reverse transcription, integration of viral DNA into the host genome, and protease-mediated maturation of new particles — and combining classes with different targets makes escape require several simultaneous mutations.
What has changed is not the principle but the burden. Early regimens meant many pills a day, strict timing, and serious toxicity; contemporary regimens are frequently a single daily tablet. Integrase inhibitors in particular combine potency with tolerability: dolutegravir was superior to efavirenz-based therapy in treatment-naive adults [20] and to raltegravir [21]. The clinical framework is set out in guidelines that have been revised continuously since the combination era began [14] [15] [22] [23].
Starting early is better, and this was proven rather than assumed. The START trial randomised 4,685 people with CD4 counts above 500 to immediate or deferred therapy and found immediate treatment reduced serious AIDS-related and non-AIDS-related events [24]. In a high-tuberculosis setting, early ART with or without isoniazid preventive therapy reduced severe illness in people with high CD4 counts [25]. In infants, early therapy reduced mortality [26]. The old strategy of waiting for immune decline is gone.
Therapy is not without cost. Weight gain following ART initiation is a recognised issue with modern regimens [27], cardiovascular risk is elevated in treated HIV [28] [29], neurocognitive disorders persist despite suppression [30], chronic inflammation continues at some level [31], certain cancers remain more common [32], and an ageing treated population brings its own challenges [33].
Prevention
Treatment as prevention is the direct consequence of the centerpiece: HPTN 052 established it in a randomised trial [19], and PARTNER established the absolute risk at suppression [17].
Pre-exposure prophylaxis puts antiretroviral drugs in HIV-negative people at risk. Daily oral tenofovir-based PrEP was shown effective in heterosexual men and women [34] and in an open-label pragmatic trial in men who have sex with men [35]. Its effectiveness depends overwhelmingly on adherence, which was demonstrated directly by measuring drug concentrations: efficacy tracked measured emtricitabine-tenofovir levels rather than reported pill-taking [36]. Renal function needs monitoring [37].
Long-acting formulations address adherence at its root, by removing the daily decision. Injectable cabotegravir given every two months was superior to daily oral PrEP in cisgender men and transgender women [38], building on earlier safety work [39]; long-acting injectable cabotegravir with rilpivirine also works as maintenance treatment after oral induction [40]. The direction of travel is unambiguously toward less frequent dosing, and the frontier is agents requiring only a couple of doses a year.
Other established measures include post-exposure prophylaxis after occupational and non-occupational exposures [41]; prevention of vertical transmission, where antiretroviral therapy in pregnancy has made mother-to-child transmission rare where it is available [42] [43], though breastfeeding transmission remains a real problem where formula feeding is unsafe [44]; and voluntary medical male circumcision, which reduced HIV acquisition in a randomised trial [45].
There is no cure, and the reason is specific
Antiretroviral therapy suppresses replication. It does not remove integrated provirus from cells that are not replicating.
This was demonstrated in 1997: in 22 patients successfully treated for up to 30 months, replication-competent virus was routinely recovered from resting CD4+ T cells, at a frequency of 0.2 to 16.4 per million cells, and that frequency did not fall with longer time on therapy [46]. The reservoir is small, stable, and largely invisible to both the drugs and the immune system. Later work showed the problem is worse than the standard assay suggests, because replication-competent proviruses that fail to be induced in vitro add to the true reservoir size [47], and latency is actively maintained by host mechanisms [48] [49].
The practical consequence: therapy is lifelong. Stopping leads to viral rebound in almost everyone, and this is why "undetectable" must not be confused with "cured."
Pillar 3: what is unresolved
Cure. A handful of people have achieved durable remission, and the routes are instructive but not scalable — a small number after allogeneic stem-cell transplantation from donors with a CCR5 mutation, and rare cases of prolonged remission after very early treatment and interruption, one lasting more than twelve years [50]. The general problem remains as stated when it was posed [51]: an assay that measures the true reservoir, and an intervention that clears it.
A vaccine. Decades of effort have produced one trial with modest, non-durable efficacy — ALVAC-AIDSVAX in Thailand [52] — and no licensed product. The obstacles are the extraordinary sequence diversity of the envelope and the fact that natural infection does not generate protective immunity to imitate; understanding the immune response in acute infection has been argued as the route in [53]. Broadly neutralising antibodies are the most promising current direction, both as prevention and as a component of cure strategies.
Access and equity. The pharmacology is largely solved; the delivery is not. Treatment coverage, retention in care, key-population access, and stigma are the binding constraints, and the epidemic remains concentrated in populations that health systems reach worst [54] [55] [11]. The gains in life expectancy documented above were measured in North America [1]; the collaborative cohort analyses show the same trajectory internationally but with the same inequities inside it [56].
Dig deeper in lmmol
Tuberculosis is the essential companion to this review: HIV-associated tuberculosis is the deadliest coinfection in the world, HIV multiplies the rate at which latent tuberculosis reactivates [6], and the trial evidence for early antiretroviral therapy in high-burden settings was designed around exactly that interaction [25] [7]. HIV also reshapes the epidemiology of the other infections in this collection — see malaria and schistosomiasis for the co-endemic tropical diseases, and measles for what happens to vaccine-preventable illness when immunity fails. The elevated cardiovascular risk in treated infection connects to heart failure [29], the cancer associations to breast cancer and colorectal cancer as points of comparison in screening strategy, and the stigma literature to depression, which is both more common in HIV and a barrier to care [9]. The full collection is at health.