TB Test LoD: What 10 vs 100 CFU/mL Really Means

The numbers 10 CFU/mL and 100 CFU/mL are analytical limits of detection (LoD)—the lowest concentration at which a cartridge reliably detects Mycobacterium tuberculosis DNA or a rifampicin resistance genotype under defined laboratory conditions. They do not represent clinical sensitivity, nor do they guarantee detection in every patient. This guide explains how these figures are generated, why one test can have two LoDs, and what evidence a procurement team must demand before comparing cartridge performance, using the iFIND TBR as a transparent and well‑documented example.

Analytical LoD versus Clinical Performance: Why the Two Numbers Belong in Different Columns

The table below separates the metrics that are often mistakenly substituted for one another. Analytical LoD tells you the lowest concentration the cartridge can detect in a contrived, highly controlled sample. Clinical sensitivity tells you how often the test is positive in patients who truly have the disease according to a reference standard. These are not interchangeable.

Why One Cartridge Can Have Two Different LoDs (and That’s Not a Design Flaw)

MTB detection and rifampicin-resistance calling are different analytical tasks. The first asks whether the assay target is detected; the second interprets signals from resistance-associated regions under a separate decision rule. Target design, mutant and wild-type discrimination, mixture composition, input amount and calling thresholds can therefore produce different performance limits. Without the product protocol, the reason for a tenfold difference cannot be determined from the headline numbers alone.

The Gap Between Analytical LoD and Clinical Sensitivity

Analytical LoD is determined with defined reference material, a documented matrix and a prespecified protocol. The replicate count, concentration levels, hit-rate criterion, statistical model, confidence interval and handling of invalid results all shape the estimate. If those elements are absent, two CFU/mL figures are not directly comparable.

• Viscosity and inhomogeneity of native sputum.
• Presence of PCR inhibitors (heme, mucin, salts).
• Uneven distribution of bacilli in paucibacillary samples.
• Loss during liquefaction, decontamination, or DNA extraction.
• Different MTB lineages with variable target copy numbers.

The 2026 prospective iFIND‑TBR study (452 patients) reports clinical performance in its enrolled population. Clinical sensitivity is a different metric and cannot be predicted from the analytical LoD alone. A procurement decision must review both, alongside specificity, invalid-result rates and operational fit.

What’s Behind a CFU/mL Number: The Methodology Checklist

A single CFU/mL figure is meaningless until you can answer ten questions about how it was generated. Use this checklist during technical evaluation of any TB molecular test tender.

1. Confidence level: Was LoD defined as the concentration giving ≥95% detection, or a less stringent 50% hit rate?
2. Replicate count: How many replicates were tested at each dilution? Low replicate numbers inflate uncertainty.
3. Reference material: Which strain (H37Rv, a clinical isolate) and how was viability confirmed? Was the inoculum quantified by plate counting or optical density?
4. Matrix: Was LoD determined in simulated sputum matrix, pooled negative sputum, or transport medium? The matrix profoundly alters recovery.
5. Extraction and input volume: How many millilitres of sample were processed, concentrated, and how much of the eluate entered the cartridge? A “per mL” claim assumes the entire sample is captured; losses at any step reduce the effective LoD.
6. Invalid‑result handling: Were invalid or indeterminate results excluded, or repeated until a valid call was obtained? This can artificially lower the LoD.
7. Statistical method: Was the point estimate derived from probit regression (continuum) or from the lowest dilution with 100% detection? The two methods yield different numbers.
8. Independent verification: Has the claimed LoD been reproduced by a laboratory independent of the manufacturer?
9. Endpoint definitions: For resistance detection, what was the minimum mutant fraction required to call “resistant”? If the LoD was determined with a 100% mutant population, it may not reflect heteroresistant clinical specimens.
10. Alignment with intended use: Is the claimed LoD relevant to the expected bacterial load in the target patient population (e.g., smear‑negative pulmonary TB)?

Worked Reading Example: Interpreting 10 vs 100 CFU/mL in a Procurement Document

Suppose a laboratory is reviewing a tender that lists “MTB detection LoD: 10 CFU/mL” and “RIF‑resistance LoD: 100 CFU/mL”. The first reflex should be to request the verification protocol. The evaluator should request the detailed report. Until it is available, the evaluator can compare against publicly available independent data, such as the Ou et al. 2024 assessment of the same cartridge, which reported an MTB LoD of 13.34 CFU/mL and a RIF‑resistance LoD of 109.79 CFU/mL.

The numerical proximity of 10 and 13.34 CFU/mL does not by itself establish equivalence or inaccuracy. Without the manufacturer protocol, the evaluator cannot know whether the same matrix, reference material, replicate design, hit-rate criterion and statistical method were used. The correct action is to request the protocol and confidence interval, then compare like with like; the two headline numbers alone are insufficient.

For the RIF-resistance endpoint, the evaluator should ask which mutations and strain materials were tested, how the resistance call was defined, and whether mixtures of susceptible and resistant populations were evaluated. A resistance-call threshold depends on more than total bacterial concentration, so it must not be treated as a second version of the MTB-detection LoD.

Specimen Matrix and Preanalytics: Why the LoD Doesn’t Travel Well

LoD results are tied to the matrix and workflow used in the study. When comparing products, verify whether the claimed CFU/mL was established in buffer, contrived sputum, processed sputum or another intended specimen. Evidence from one matrix should not be transferred to another without supporting data.

Transport, storage, decontamination, concentration and input-volume steps can change how much target reaches the cartridge. Their effect must be evaluated in the intended workflow; no fixed percentage difference between analytical LoD and clinical sensitivity can be assumed. Always request clinical performance data from the intended specimen and setting.

Comparison Traps: CFU/mL, Copies/mL, and Other Numbers That Look Alike

• CFU/mL ≠ copies/mL. CFU counts viable, culturable bacteria; molecular tests detect DNA from both viable and non‑viable organisms. A copies-per-reaction result uses a different denominator and measurement chain. Never compare it directly with CFU/mL without a validated, study-specific conversion and full workflow volumes.
• Different dilution series, different results. If one manufacturer starts with a high‑concentration stock measured by McFarland and another uses carefully plate‑counted CFU, the same nominal “10 CFU/mL” can correspond to very different true bacterial loads.
• Instrument‑specific thresholds. A cartridge may have an in‑built algorithm that suppresses borderline signals to reduce false positives. That threshold alters the effective LoD, even if the underlying chemistry could detect fewer organisms.
• Overselling analytical LoD as clinical sensitivity. Marketing materials sometimes conflate the two. An analytical LoD of 10 CFU/mL does not mean the test detects 10 CFU/mL in all patient samples. Always ask for the clinical sensitivity figure from an independent, prospective study.

Ten-Question LoD Procurement Checklist

Use this checklist to structure a technical evaluation request to any manufacturer:

1. At what confidence level (hit rate) is the LoD reported?
2. How many independent replicates were tested at each dilution step, and across how many independent runs?
3. Which reference strain and viability quantification method were employed?
4. In what exact matrix was LoD determined, and how does that matrix compare to the specimen types the test will see?
5. What were the sample input volume, extraction protocol, and eluate volume entering the reaction?
6. How were indeterminate, invalid, or “grey‑zone” results handled in the LoD calculation?
7. What statistical model was used to derive the point estimate and the associated confidence interval?
8. Has the LoD been independently verified by a laboratory not affiliated with the manufacturer, and are the full protocols available?
9. For any drug‑resistance component, what was the mutant fraction of the challenge population, and was LoD determined for heteroresistance scenarios?
10. How does the analytical LoD relate to the clinical sensitivity reported in an intended‑use population study?

Frequently Asked Questions

Is a lower CFU/mL always better?

Not necessarily. The clinical value of a lower analytical LoD depends on specimen quality, intended population, clinical sensitivity and specificity, invalid-result rates and workflow controls. A smaller headline number should not outweigh independent clinical evidence.

Can I directly compare the LoD of two different cartridges?

Only if both LoDs were determined using an identical protocol—same strain, matrix, extraction, statistical method, and confidence level. Head‑to‑head comparisons require, at minimum, a method‑comparison study. Relying on brochure numbers alone is a common procurement mistake.

What if the manufacturer’s brochure LoD doesn’t exactly match the independent study?

Do not judge agreement from the headline difference alone. Compare the matrices, reference materials, replicate design, statistical method and uncertainty reported by each study. If those are not comparable, request clarification rather than declaring the estimates concordant or contradictory.

Does the cartridge require viable bacteria to produce a signal?

No. CFU is based on viable growth under specified culture conditions, while a molecular result is based on detected nucleic acid. The units are not interchangeable without a validated conversion for the exact method.

Where can I find more guidance on procuring TB molecular tests?

See our comprehensive procurement guide for the iFIND TBR cartridge here. For resistance‑focused cartridges, we also cover the iFIND IFQ cartridge here. Russian‑speaking laboratories can consult the Russian‑specific evidence summary here.

Sources and Limitations

This article is an FYZSXNB educational resource. It is not medical advice. Analytical performance data are drawn from the following sources. Manufacturer claims listed as “pending certificate” have not been independently audited by a regulatory body.

• iFIND TBR manufacturer brochure – on file; certificate status pending; no public URL provided.
• Ou X, et al. Preliminary assessment of a molecular tuberculosis cartridge. PMC 2024. PMC11554655.
• Prospective iFIND‑TBR study in 452 patients, China, 2026. PubMed 41767560.
• World Health Organization. WHO consolidated guidelines on tuberculosis. Module 3: Diagnosis – rapid diagnostics for tuberculosis detection, 2025. WHO/9789240107984.
• World Health Organization. Mutation catalogue for Mycobacterium tuberculosis complex and drug resistance. WHO/9789240082410.
• U.S. Food and Drug Administration. Statistical Guidance on Reporting Results from Studies Evaluating Diagnostic Tests. FDA guidance.
• FYZSXNB editorial interpretation of the above evidence.

Limitations: The independent LoD values cited come from a single published study under specific laboratory conditions. CFU assignment and sample preparation can vary, so the numbers remain tied to the methods used in that study. Clinical sensitivity data from the 2026 study should be interpreted in the context of the patient population enrolled. This article does not claim that the iFIND TBR is WHO‑endorsed, FDA‑cleared, or CE‑marked; it serves as a general methodological guide.