This week in Infectious Disease

This AudioScholar briefing is generated by artificial intelligence for healthcare professionals and trainees. It is not medical advice.

This week: empirical tuberculosis treatment tested in African infants with HIV and severe pneumonia, doxycycline in early pregnancy linked to pregnancy loss, and cefepime set against piperacillin-tazobactam for bloodstream infections at risk of AmpC production.

Welcome to This Week in Infectious Disease. We're covering 10 papers this week: 3 in depth, 4 in brief, 3 quick signals.

The Lancet published EMPIRICAL, a multicentre, open-label, two by two factorial randomised trial led by Domínguez-Rodríguez and colleagues [1]. It enrolled 563 infants aged one to twelve months with severe HIV-associated pneumonia across 19 hospitals in six African countries. All infants received standard care for bacterial and Pneumocystis pneumonia, and this report asks whether adding six months of empirical four-drug tuberculosis treatment improves survival compared with treatment guided by Xpert Ultra, urine lipoarabinomannan testing and clinical judgement. The guided arm did not stay untreated for long. Just under half of infants in the guided groups eventually started tuberculosis treatment, at a median of four days, more often on clinical grounds than on a positive test. That crossover narrows any difference the trial could detect, and it is the most important limitation. The abstract text available ends mid-sentence at the twelve-month figures, reporting 132 of 276 infants in the empirical groups and 121 of 282 in the guided groups, so the adjusted comparison and the authors' conclusion are not in the text provided. No interaction was found between valganciclovir and tuberculosis treatment. What the trial adds is a randomised comparison of the two strategies under real-world crossover, and the full report is needed to read its direction.

Annals of Internal Medicine published an active-comparator cohort study from Winterstein and colleagues [2], using MarketScan commercial and Medicaid claims from the United States. It covered about 218,000 pregnancies with a respiratory tract infection and about 6,600 with acne, comparing first-trimester doxycycline with amoxicillin, macrolides or topical clindamycin after matching on propensity score and gestational age. In the respiratory cohorts, doxycycline was associated with roughly a forty to fifty percent higher risk of pregnancy loss than amoxicillin, in both databases, and a similar excess against macrolides. In acne, the excess was statistically significant in the commercial data but not in Medicaid. Where the associations were significant, absolute risk differences ran between 7 and 11 percentage points. The authors acknowledge that residual confounding and misclassification of exposure or outcome are possible, and congenital malformations were not assessed. The strength is consistency across two independent databases and two active comparators. The weakness is that a claims design cannot fully separate the drug from the reasons it was chosen. The authors frame the result as information for weighing risks and benefits in reproductive-aged women, and the evidence remains observational.

The Journal of Antimicrobial Chemotherapy published ROCKET, a retrospective multicentre study from Jawanda and colleagues [3], in 728 adults with bacteremia due to Enterobacter cloacae complex, Klebsiella aerogenes or Citrobacter freundii complex, the organisms at moderate risk of clinically significant AmpC production, treated between 2017 and 2025. About a third of patients were immunocompromised. Cefepime was the most common definitive agent, used in just under half of patients, with piperacillin-tazobactam in about a fifth. Patients given piperacillin-tazobactam were slightly sicker by Pitt bacteremia score, yet crude thirty-day mortality was similar, around 14 percent against 11 percent with cefepime. After inverse probability weighting, mortality did not differ significantly between the two drugs. Early relapse was rare in both groups, and treatment-emergent resistance to definitive therapy occurred in only one to three percent of patients across cefepime, piperacillin-tazobactam and carbapenems. The main limitation is the retrospective design with treatment chosen by clinicians, and the mortality comparison was imprecise, so a modest difference cannot be excluded. The authors describe it as the largest study of these organisms to date, and it adds observational support for non-carbapenem therapy without settling the choice between the two beta-lactams.

Now, four in brief.

Also in the Journal of Antimicrobial Chemotherapy, at the other end of the resistance spectrum, Dubey and colleagues [4] describe 85 patients in several countries who received cefepime-zidebactam through compassionate use for carbapenem-resistant Gram-negative infections. Just over half of patients were in intensive care, Pseudomonas aeruginosa was the main pathogen, and metallo-beta-lactamases, mostly NDM, were common where genotyping was done. Investigator-assessed clinical success at end of therapy was about 88 percent, on-therapy mortality was around eleven percent, and adverse events were infrequent. This is an uncontrolled case series with investigator-assessed outcomes, and the authors state it needs confirmation in controlled studies.

PLOS Medicine published an antimicrobial resistance sub-study of LAKANA, the cluster-randomised trial of mass azithromycin for infants in Mali, from Juma and colleagues [5]. The parent trial found no mortality benefit. Across 59 villages and about ten thousand nasopharyngeal and rectal samples, twice-yearly azithromycin for two years did not increase macrolide resistance in pneumococcus or E. coli, in treated children or in untreated older contacts, and it met the noninferiority margin. Quarterly dosing could not exclude a modest, dose-dependent rise, and the authors note the study may have been too small to detect it. The result is reasonably firm for twice-yearly dosing and inconclusive for quarterly.

Staying with tuberculosis, The Lancet Global Health published SPOT-TB from Mahfooz and colleagues [6], a pragmatic stepped wedge cluster-randomised trial across all four provinces of Pakistan. Thirty mobile x-ray van teams switched from conventional site selection to camp locations chosen by the MATCH-AI software, across nearly 4,000 camps screening about 269,000 people. In the intention-to-treat analysis, bacteriologically confirmed tuberculosis per camp was 7 percent higher with AI targeting, a difference that was not statistically significant. In a per-protocol analysis of camps held within five kilometres of recommended sites, yield was about a third higher. The primary result is null, and implementation fidelity could not be fully assessed.

Clinical Infectious Diseases published a national prescription cohort from Zhu and colleagues [7] on long-acting cabotegravir for HIV pre-exposure prophylaxis in the United States from 2022 to mid-2025. The number of people prescribed the injectable rose from about 4,600 in 2022 to over 25,000 in the first half of 2025 alone, though it still accounted for only about five percent of prescriptions. Abandonment, meaning a prescription never filled, differed by sex. Women were about a third less likely to abandon the injectable than daily oral PrEP, while men were about a quarter more likely to abandon it. These are pharmacy data, so the reasons behind unfilled prescriptions are not captured.

Three quick signals.

In Clinical Infectious Diseases, de Prost and colleagues [8] followed 726 people admitted with influenza to 42 French intensive care units across two seasons. Despite the emergence of H3N2 subclade K, the 2025 to 2026 season showed no increase in day-28 mortality, and subtype was not associated with death, while age, frailty and organ failure were. The cohort is prospective but observational, with sequencing in only forty H3N2 cases.

In The Lancet Infectious Diseases, Serra and colleagues [9] analysed 45 pregnant women with Lassa fever in a prospective Nigerian cohort. Maternal mortality at 28 days was about nine percent, similar to non-pregnant women and lower than older case series, but fetal death occurred in 60 percent, highest with early infection. The pregnant sample is small and from a single centre.

And in a research brief in Clinical Infectious Diseases, Yang and colleagues [10] found that congenital syphilis among United States birth hospitalizations rose roughly sixfold between 2012 and 2022, across every demographic group, with a more than twentyfold rise among Native American newborns. The data are administrative coding, reported only briefly.

Stepping back across this week's papers, the firmest designs answered the narrowest questions, while the most consequential safety and treatment questions rest on observational data. The randomised work was either diluted by crossover or carried its positive signal only in a per-protocol analysis. The pregnancy, bacteremia and novel-agent studies depend on claims, retrospective records or uncontrolled series, where confounding by indication cannot be removed and controlled comparisons have not yet been run.

If you only have time for one paper this week, make it the doxycycline and pregnancy loss cohort [2]. It reopens the question of first-trimester safety for an antibiotic widely prescribed to women of reproductive age, with a consistent signal across two large independent databases.

That's your roundup for This Week in Infectious Disease. The full transcript and references are available on the episode page. See you next week.

If this weekly briefing is useful, follow the show in your podcast app so new episodes arrive automatically. And a quick rating on Apple Podcasts or Spotify helps other physicians discover the show.

This is an automated summary generated by artificial intelligence, which can make mistakes. Always review the original source materials.