health · family: it worked in the lab
the test demands whatthe sickest can't produce
Tuberculosis point-of-care diagnostics cannot reach the patients who need them most because they depend on sputum
Problem statement
Tuberculosis remains one of the deadliest infectious diseases worldwide, and about 3 million of the 10.6 million people who developed TB in 2022 were not diagnosed. The current rapid molecular tests (such as GeneXpert) are run mainly on sputum and need an instrument, basic laboratory infrastructure, and staff with basic technical skills, which has made them hard to deploy at scale in primary health clinics where electricity may be unstable. Children, people living with HIV, and people with presumed extrapulmonary TB often cannot produce a good sputum sample. The WHO's 2024 updated TPP is the first to unify sputum and non-sputum approaches and define three test classes (point-of-care, near-PoC, low-complexity), but no test meeting these specifications yet exists.
Why this matters
The diagnostic gap of about 3 million cases a year means millions receive no treatment, continue transmitting, and face preventable death. The WHO TPP modeling shows that a less accurate test deployed at true point-of-care with non-sputum specimens can achieve comparable or better case detection than current GeneXpert because increased access compensates for reduced sensitivity — but no such test has been developed.
What’s been tried and why it hasn’t worked
Sputum smear microscopy remains the primary tool for TB diagnosis in some low- and middle-income settings. GeneXpert was substantial progress but remains infrastructure-dependent: it needs an instrument, cartridges cost ~$8 each, and the platform requires stable electricity. Urine lipoarabinomannan (LAM) tests offer a non-sputum alternative but the lateral-flow LAM test (LF-LAM) has suboptimal sensitivity outside people living with HIV, and its use has been limited to them. Oral/tongue swabs tested with GeneXpert Ultra are under study but require the same GeneXpert infrastructure. The fundamental gap remains: no instrument-free, non-sputum test exists that can confirm TB at the point of first clinical encounter.
What would unlock progress
The WHO TPP defines the target: a point-of-care test requiring no electricity, cold chain, or equipment, with >98% specificity and sensitivity sufficient to outperform current care cascades when access gains are modeled. The most promising approaches combine non-sputum specimen types (oral swabs, urine, exhaled breath) with isothermal amplification or novel biomarker detection on lateral flow platforms. A student contribution could address the measurement science: systematic comparison of oral swab vs. urine LAM vs. combined approaches against culture-confirmed TB in a high-burden clinic setting.
Entry points for student teams
The door that needs no laboratory is the modeling the WHO TPP itself ran: take published subgroup sensitivities for urine LAM, oral/tongue-swab Xpert and sputum Xpert, combine them with WHO's freely downloadable global TB data files (case notifications, HIV-coinfection shares, diagnostic service coverage by country) and compute what a combined-specimen algorithm would add to case detection in a specific care cascade — the complementarity question answered on the public record, which is where the field currently argues from intuition. On the bench, a team could build a prototype lateral flow assay combining FujiLAM-type urine detection with an oral swab molecular component and report analytical limits of detection per modality in spiked matrices: purified lipoarabinomannan into pooled urine, and synthetic IS6110 or IS1081 target DNA plus a BSL-2 surrogate mycobacterium (M. smegmatis or BCG) onto oral-swab eluate. Establishing true clinical complementarity between HIV-associated and general TB needs paired specimens from diagnosed patients and handling of live M. tuberculosis at BSL-3 — access owned by national TB reference laboratories and their ethics committees, not by a student team — so the clinical evaluation is a protocol written for such a laboratory rather than a semester result. Alternatively a team could build a portable, battery-free isothermal amplification device optimized for oral swab specimens and benchmark it against benchtop qPCR on the same spiked panels and against GeneXpert's published performance, rather than against a $17,000 instrument and its cartridges. Relevant disciplines: biomedical engineering, microbiology, global health, diagnostics design, epidemiological modeling.
Genes
Structural cousins: stuck the same way, in other fields
Sources
WHO, "Target product profile for tuberculosis diagnosis and detection of drug resistance," 2024, ISBN 9789240097698. Kohli M et al., "WHO target product profile for TB detection at peripheral settings: 2024 update," PLOS Global Public Health 5(6):e0004612, 2025. go to source 1 ↗ go to source 2 ↗
Working notes
Notes made while writing and checking this brief, shown as written. They use working shorthand.
- The 2024 WHO TPP is notable for using simulation-based modeling (South Africa, India, Kenya care cascades) to derive performance targets — a methodological innovation in TPP design.
- Structurally related to water-field-pathogen-detection (field-deployable molecular detection) and health-malaria-rdt-behavioral-compliance (diagnostic-to-treatment cascade failures).
- The diagnostic gap is driven by both technical (sputum dependence) and systemic (health system access) barriers — the TPP attempts to address both simultaneously.
Reconciliation 2026-08-21: the C37 realism triage (score 2) flagged the first entry point for asking students to establish specimen-type complementarity across HIV-TB and general-TB subgroups, which requires paired clinical specimens and live M. tuberculosis at BSL-3; the flag holds. Checking the whole section rather than the flagged sentence, the second (unflagged) entry point failed the same test — "test it against GeneXpert" presumes the $17,000 instrument, its cartridges and TB-positive specimens, so the brief in fact had no reachable door. The section now leads with a facility-free modeling door, keeps the bench assay but scopes it to analytical limits of detection on spiked matrices with BSL-2 surrogates and synthetic targets, states the BSL-3 and clinical-specimen access line with the clinical evaluation as a handed-over protocol, and benchmarks the isothermal device against benchtop qPCR and GeneXpert's published performance. Resources verified by fetch on 2026-08-21: WHO Global Tuberculosis Programme data downloads (https://www.who.int/teams/global-tuberculosis-programme/data), free CSV files (burden estimates, case notifications and treatment outcomes, drug-resistance testing, laboratory diagnostic services) with no registration beyond accepting WHO's data terms. Reference material for the spiked panels — inactivated mycobacteria and purified antigens from BEI Resources (https://www.beiresources.org/) — is named here rather than in the entry points because its catalog is public but ordering is registration-gated and subject to institutional eligibility; the specific LAM catalog entry was not confirmed, so no item number is cited. Body sections and Genome Tags are unchanged.
Reconciliation 2026-09-29: Fact repair after a site review read the whole write-up against both cited sources. The WHO document was fetched on 2026-09-29 as the full PDF linked from the cited WHO page (https://iris.who.int/server/api/core/bitstreams/0f3962a0-d382-4a17-b86f-6eb6208c9f87/content), converted to text and searched for every figure and name in this write-up. Kohli et al. 2025 was fetched the same day from the cited PLOS address and read in full. (1) The text said TB "kills 1.25 million people per year — more than any other single infectious agent". Neither source gives that figure or that ranking. The WHO document calls TB "one of the deadliest infectious diseases worldwide", and the sentence now says that. No death figure is given: Kohli et al. print 1.6 million deaths for 2022, but the WHO Global Tuberculosis Report 2023 factsheet, fetched on 2026-09-29 (https://cdn.who.int/media/docs/default-source/hq-tuberculosis/global-tuberculosis-report-2023/global-tb-report-2023-factsheet.pdf?sfvrsn=f0dfc8a4_3), gives 1.30 million for that year. (2) The text said "an estimated 4 million cases go undiagnosed annually", and the same figure appeared in Why This Matters and in the notes above. The WHO document says 10.6 million people developed TB in 2022 and 7.5 million were diagnosed. Kohli et al. say the gap rose to 4.2 million during the COVID-19 pandemic and was back to just over 3 million in 2022. The text now says about 3 million. (3) The text said GeneXpert is confined "to district-level facilities or higher". Kohli et al. class GeneXpert as a low-complexity test that can be placed in peripheral laboratories and in some clinics with basic laboratory infrastructure. The WHO document says such tests have been hard to deploy at scale in primary health clinics where electricity may be unstable. The sentence now follows the sources, and "best available" was dropped because neither source ranks the tests. (4) "Elderly patients", "those with non-productive cough" and "20% of TB cases in prevalence surveys are completely asymptomatic" are not in either source and were removed. The WHO document names children, people with HIV and people with presumed extrapulmonary TB as the groups for whom a good sputum sample is hard to get, and the sentence now names those groups. (5) The sentence giving 98% of cases in low- and middle-income countries and naming five countries as over half the global burden is not in either source and was removed. (6) Smear microscopy "for over a century" and "misses 40-60% of cases" are not in either source. The sentence now gives what the WHO document says, that smear microscopy remains the primary tool in some low- and middle-income settings. (7) "GeneXpert (2010)", "dramatically improved sensitivity" and "each instrument costs $17,000+" are not in either source. The year and the instrument price were removed from this sentence, which now uses the WHO document's words, "substantial progress". The entry points are advice and keep "a $17,000 instrument". That price was checked by fetch on 2026-09-29 against the MSF Access Campaign technical brief "Time for $5: GeneXpert Diagnostic Tests", December 2019 (https://msfaccess.org/sites/default/files/2019-12/MSF_Access_TechnicalBrief_GeneXpert-Time-for-5_0.pdf), which gives US$17,000 for the four-module GeneXpert with a desktop computer at the price set for eligible low- and middle-income countries. (8) "Alere LAM has low sensitivity except in advanced HIV" now follows the WHO document, which calls the test LF-LAM and says its sensitivity is suboptimal in other groups, so its use has been limited to people with HIV. (9) The sentences on FujiLAM and on chest X-ray with AI are not in either source and were removed. (10) "87-92% sensitivity in Guinea" for tongue swabs is not in either source. Kohli et al. list swab studies, including one on tongue swabs tested with Xpert Ultra, without giving results, so the sentence now says these are under study. (11) ">=98% specificity" now reads ">98%", as both sources give it. Confirmed against the sources and left as written: both entries on the Source line (the WHO title as shown on the cited page, the ISBN, and the PLOS volume, issue and article number); the three test classes; sputum and non-sputum tests under one profile; a point-of-care test that needs no electricity, cold chain or equipment; the modelling on care cascades in South Africa, India and Kenya; the finding that a less accurate but more accessible test can match or beat current case detection; the cartridge price of about $8 (the WHO document gives US$ 7.97 for Xpert Ultra); oral swabs, urine and breath as sample types; and the absence of a true point-of-care test. Not checked: "isothermal amplification" as a promising approach and the laboratory details in the entry points (IS6110, IS1081, the surrogate organisms, the BSL levels). Neither source discusses them, they are this write-up's own suggestions, and they are left as written. One thing to know: the PDF's cover gives the WHO title in the plural, "Target product profiles", while the cited web page gives it in the singular. The Source line follows the web page and was not changed. Genes untouched.