MPN: Counting Bacteria With Statistics Instead of Colonies
The study: The most probable number (MPN) technique — classical quantitative microbiology (documented in FDA BAM, USDA MLG, and standard methods references).
Sometimes presence/absence isn’t enough — you need to know how many. When pathogens are sparse, injured, or in matrices where direct plating fails, the most probable number method estimates concentration statistically: inoculate replicate tubes at several dilutions, score each as positive or negative after incubation, and read the pattern against MPN tables to estimate the original concentration.
How it works, plainly
Dilute the sample, inoculate (typically 3 or 5 tubes per dilution across 3 dilutions), incubate, and score growth. The pattern of positive tubes (say, 3-2-1) corresponds to a statistical estimate of organisms per gram — the “most probable” concentration consistent with that pattern, with confidence intervals. It’s enumeration by probability rather than colony counting.
When MPN is the right tool
Low-level contamination where direct plating would show nothing. Injured organisms that need enrichment to recover. Regulatory and validation contexts where the reference method is MPN-based (some BAM chapters). The trade-offs: wide confidence intervals (it’s an estimate, not a count), labor-intensive, and slow. Modern alternatives (qPCR with proper calibration, improved plating) sometimes replace it — but MPN remains the reference in many standards. For validation studies: if your acceptance criterion references an MPN method, replicate it faithfully — the statistics only work with the specified tube numbers and dilutions. And report the confidence interval, not just the point estimate. A good MPN result is honest about its uncertainty.
The statistics behind the tubes
The MPN method — the dilution series, the pattern of positive tubes, the statistical table — is applied probability: given the pattern of growth across dilutions, what’s the most probable original concentration? The method’s elegance is its handling of the low-concentration problem: when the organisms are too few to count on plates, the presence/absence pattern across dilutions still yields a quantitative estimate. The confidence intervals (wide — the MPN’s notorious imprecision) are honest about the method’s limits.
The MPN’s niche in modern food microbiology is specific: the low-level quantification where plate counts fail (the detection limit problem), the regulatory methods that specify it (the BAM’s MPN procedures for certain organisms), and the situations where only presence/absence per dilution is feasible. The imprecision (the wide confidence intervals — a 3-tube MPN’s interval spans an order of magnitude) limits its use for tight specifications; it’s a rough-quantification tool, not a precise one.
The method’s persistence — still in the BAM, still in the standards, after a century — reflects its robustness: no equipment beyond tubes and media, applicable to any organism with a detectable growth endpoint, the statistics well understood. The modern laboratory uses it less (the plate counts, the molecular quantification) but keeps it for the applications where it’s specified or where nothing else works. The MPN’s lesson: the old methods survive where their statistical honesty (the wide intervals telling you exactly how rough the estimate is) serves the decision better than a precise-looking number from a method whose biases are hidden.
Source: FDA Bacteriological Analytical Manual (MPN chapters); USDA MLG; standard methods references.