Inflammatory and Coagulation Biomarkers and Mortality in Patients with HIV Infection

If stopping HIV treatment allows the virus to rebound, and that viral rebound triggers inflammation, which then drives clotting, then interrupting therapy — even briefly — should raise your risk of dying from causes that have nothing to do with AIDS. Hold that chain for one beat. That is exactly what Kuller and colleagues went and measured, and the numbers they found were striking enough to reshape how the field thinks about HIV treatment. The Strategies for Management of Anti-Retroviral Therapy trial, or SMART, enrolled five thousand four hundred seventy-two HIV-infected participants across thirty-three countries. All had CD4 counts above three hundred fifty cells per cubic millimeter. One arm, called viral suppression, received continuous antiretroviral therapy with the goal of keeping the virus undetectable. The other arm, called drug conservation, used an intermittent strategy: stop antiretroviral therapy until the CD4 count fell below two hundred fifty, restart it, then stop again once counts recovered above three hundred fifty. The drug conservation approach was tested because of real concerns about long-term side effects — it was meant to spare patients unnecessary drug exposure, not to harm them. What emerged was unexpected. The intermittent arm had more than twice the risk of opportunistic disease or death compared to continuous antiretroviral therapy, with a hazard ratio of 2.6. All-cause mortality was eighty-four percent higher in the drug conservation group. And here is the detail that made investigators lean forward: very few of those excess deaths were from AIDS-defining infections. Of fifty-five deaths in the drug conservation group, only four were attributable to opportunistic disease. The puzzle wasn't AIDS. It was something else killing these patients. Kuller and colleagues proposed a specific biological chain to explain it. When antiretroviral therapy stops and HIV-RNA rebounds, the viral surge activates tissue factor — the molecular trigger for the blood-clotting cascade. That activation leads to thrombosis and then fibrinolysis, the body's attempt to dissolve those clots. To test this hypothesis, the team measured six biomarkers in stored plasma specimens from the trial. Four were inflammatory markers: high-sensitivity C-reactive protein, or hsCRP; interleukin-6, an inflammatory cytokine; and serum amyloid A and amyloid P. Two were coagulation markers: D-dimer, a breakdown product of blood clots that rises when abnormal clotting and dissolving are occurring; and prothrombin fragment one plus two, a marker of thrombin generation. Together, these six covered both sides of the proposed mechanism — inflammation and clotting. To test whether these markers predicted death, the team ran a nested case-control study. They identified eighty-five participants who died before the trial was halted in January two thousand six and selected two controls per death, matched on country, age within five years, sex, and date of randomization. Stored plasma specimens were assayed in blinded batches. The lowest quartile of each biomarker served as the reference, and conditional logistic regression estimated odds ratios for each of the three upper quartiles versus that reference. They also ran a separate analysis using the most recent specimen available before death — not just the entry level — to capture how markers moved over time. To test whether stopping antiretroviral therapy actually moved the markers in real time, they drew on a pre-specified random sample of two hundred forty-nine drug conservation and two hundred fifty viral suppression participants with specimens at baseline and at one month after randomization. Stored samples made both analyses possible without running a new trial. Now the findings. At study entry, higher levels of hsCRP, interleukin-6, and D-dimer were each significantly associated with dying. The unadjusted odds ratio for the highest versus lowest quartile was two point zero for hsCRP, eight point three for interleukin-6 — with a ninety-five percent confidence interval of three point three to twenty point eight — and twelve point four for D-dimer, with a ninety-five percent confidence interval of four point two to thirty-seven point zero. That last number deserves a moment. Participants in the top quartile of D-dimer at study entry had roughly a twelve-fold higher odds of dying compared to those in the bottom quartile. Twelve-fold. And this is in a population where most people still had reasonably intact immune systems. These associations did not disappear when the team controlled for age, race, HIV-RNA level, CD4 count, smoking, body-mass index, prior cardiovascular disease, diabetes, cholesterol, hepatitis co-infection, and treatment group. They held in the drug conservation and viral suppression arms analyzed separately. They also held in the latest-level analyses: an interquartile-range increase in interleukin-6 at the visit closest to death was associated with an adjusted odds ratio of two point zero for mortality, and the same increase in D-dimer carried an adjusted odds ratio of two point two. The signal was consistent across every way they sliced the data. Then came the one-month biomarker data — the part that connects the dots. In the drug conservation arm, interleukin-6 rose by thirty percent and D-dimer rose by sixteen percent at one month after randomization. In the viral suppression arm, interleukin-6 changed by essentially zero and D-dimer by five percent. The treatment-group differences were significant at a p-value below zero point zero zero zero one for both markers. In the subgroup who had been on antiretroviral therapy at enrollment with HIV-RNA below four hundred copies per milliliter — the people for whom stopping therapy caused the sharpest viral rebound — the effects were even larger: interleukin-6 jumped forty-three percent in the drug conservation arm versus five percent in viral suppression, and D-dimer rose twenty-seven percent versus a slight decline. Those increases tracked directly with how high HIV-RNA climbed. The more viral rebound, the bigger the jump in interleukin-6 and D-dimer, at a p-value below zero point zero zero zero one. That relationship is the mechanistic spine of the whole argument: antiretroviral therapy interruption lets the virus replicate, the virus drives inflammation and coagulation, and those changes show up in the blood within a month. The team then asked a harder question: do these marker changes actually account for the survival difference between the two arms? In an expanded case-control analysis, the unadjusted odds ratio for dying in the drug conservation versus viral suppression arm was one point eight. When the model was adjusted for latest interleukin-6, that odds ratio dropped to one point five. Adjusted for latest D-dimer alone, it dropped to one point four. Adjusted for both markers plus CD4 count and HIV-RNA, it fell to one point three — and the confidence interval crossed one. The excess mortality in the intermittent therapy arm was partly, though not entirely, explained by these inflammatory and coagulation markers. They appear to sit in the pathway between viral rebound and death. The log-scale differences between arms at one month also translate into meaningful mortality estimates. The drug conservation group had a zero point zero eight log10 higher interleukin-6 level at one month. Kuller and colleagues estimated that corresponds to a sixteen percent increased risk of death. The zero point zero eleven log10 higher D-dimer in that arm was estimated to correspond to a twenty-four percent increased risk. These are not trivial numbers in a population that was not yet immunocompromised by conventional standards. Kuller and colleagues are careful about what they can and cannot claim. The number of deaths was small enough to produce wide confidence intervals, follow-up specimens were not available for all participants, and elevated biomarkers near death could partly reflect reverse causality — the body deteriorating for other reasons before the markers rise. These are real limitations. But they are limitations of degree, not of direction. The pattern is too consistent across methods, time points, and treatment arms to dismiss. What SMART ultimately demonstrated is that the relevant danger in HIV infection is not only the virus's direct assault on the immune system. Even patients with entry CD4 counts averaging around six hundred cells per cubic millimeter — people who, by older clinical standards, were doing reasonably well — harbored an inflammatory and procoagulant state that predicted death across non-AIDS pathways. And stopping antiretroviral therapy, even temporarily, made it worse within a month. The authors argue explicitly that therapies targeting HIV-related inflammation and lowering interleukin-6 and D-dimer warrant investigation, and that these two markers are strong candidates for endpoints in future intervention trials. The field took that argument seriously. The legacy of SMART is continuous suppression — not to prevent AIDS, but to keep the blood quiet. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.