Vitamin D won't prevent TB. Cheaper detection might control it

The pill that doesn't prevent

Tuberculosis prevention has a candidate that sounds too easy: vitamin D supplements. An individual participant data meta-analysis of randomised controlled trials finds it does not work as prevention. In a preprint posted this week, vitamin D supplementation does not reduce the risk of incident tuberculosis infection, and does not reduce the risk of incident tuberculosis disease, and there is no evidence that the result differs across participant characteristics, including baseline vitamin D status below 25 nmol/L. m:…1112

The analysis draws on individual participant data, the raw records from each trial rather than published summaries, which let the authors test subgroups directly, and they found no evidence of effect modification by participant characteristics, including baseline vitamin D status below 25 nmol/L. That matters because the usual defence of a failed supplement is that it works in the deficient subgroup; here, even participants with very low baseline vitamin D showed no benefit. m:…1112

Typing, cheaper

If prevention is not the lever, detection has to be. HiSpoT, a one-tube, dual-barcode PCR method for strain typing of the tuberculosis bacterium, multiplexes 96 samples in a single tube and replaces commercial library preparation kits with a cheaper in-house step. The method targets 43 conserved spacer oligonucleotides in the tuberculosis bacterium's CRISPR region, and the authors present it as a low-cost alternative to commercial library preparation kits, aimed at rapid surveillance where resources are limited. Validated on clinical isolates, it agreed with whole-genome sequencing in 70% of cases, with most disagreements traced to single-spacer insertion elements disrupting primer sites. The method keeps compatibility with legacy spoligotype databases, and it ships with open-source tools for processing the sequencing data. m:…3406

What resistance looks like on the ground

Routine testing in northern Morocco shows why that detection capacity matters. A programmatic analysis of testing episodes using Xpert MTB/XDR found rifampicin resistance in 39.9% of interpretable episodes. Of resistant episodes with complete profiles, 71.7% were resistant to two or more drug classes. These are programmatic testing figures from one region, not a population survey, since the tested episodes are the ones clinicians chose to test, but they show a surveillance system systematically characterising multiclass resistance patterns in routine diagnostic networks. m:…8463

The Morocco figures also show what routine molecular surveillance is for. Rather than treating resistance as something discovered only after treatment fails, the testing produces a running picture of which drugs still work in a region, episode by episode. That picture is the raw material for treatment guidelines and for spotting multiclass resistance early, while it is still a surveillance signal rather than a treatment failure. m:…8463

The shift

Taken together, the three preprints describe a rebalancing. The cheap, scalable prevention idea fails a rigorous test; the cheap, scalable detection tools pass theirs, with stated limits. For public health programmes, the implication is to stop spending attention on vitamin D as TB prophylaxis and put it into the diagnostic network instead: faster typing for outbreak tracking, routine molecular resistance testing where the burden is highest. The open questions are practical ones, whether HiSpoT holds up outside the cohort it was validated on, and whether resistance figures like Morocco's generalise, and both are answerable with the same kind of programmatic data that produced them. m:…1112 m:…3406 m:…8463