The Most Probable Number Method and Its Use in QC Microbiology Scott Sutton

"Microbiology Topics" discusses var1ous topics in m1crobiology of practical use in validation and compliance. We intend this column to be a useful resource for daily work applications.

Reader comments, questions, and suggestions are needed to help us fulfill our objective for this column. Please send your comments and suggestions to column coordinator Scott Sutton at scott.suttonOm1crob1ol.org or journal coordinating editor Susan Haigney at shaigney@advanstar.com.

KEY POINTS The following key points are discussed:

• The most probable number (MPN) method ts famtliar to quality control (QC) microbiologists as part of the microbial limits tests. Its usefulness goes far beyond this one test, however.

• The theory behind the MPN method is central to the commonly used 0-value determination by fraction-negative method, and a variant of this method has been suggested for trending of environ­mental monitoring data from the aseptic core.

• MPN can be adjusted to provide a sensitive method to determine differences between two qualitative microbiological methods. As such. it can be used as a tool in validation of rapid microbiological methods and for growth promotion testing of broth media.

INTRODUCTION The most probable number (MPN) method is a useful, if underutilized, tool for the microbiologist. It is part of the harmonized compendia! chapter on bacterial enumeration (1) and has been part of the "Microbi­al Limits Test» chapter in the United States Pharmacopeia (USP) since the chapter's inception in USP XVITI (2). The test is a method to estimate the concentration of viable microorganisms m a sample by means of replicate liquid broth growth m 10-fold diluuons. It is particularly use­ful with samples that contain particulate material that interferes with plate count enumeration methods.

The has11. ac;sumpuons oltht' :-..1P:--: methL1d arc that bactena follow Pms-;on ~t<lltSttcs .. md thm a :.in­gle 'table L'Cll v.tth result in turoiduy uf the test me­dta undt:r th~ condnwns u<.~:d . :-\umcm bmth ''til support grl1\\ t h of organtsms and turn turbid Tht: basiL p.mcrn of ~mwth \':, no-gro\\t" l:.m provtde mformation , parttcularly at low number-; ol organ­ISms "11 h large numbt:r~ ol rcpltt att:s 0) llowe,·er. tht~ accutacr can be great!)' m.:rcascd b) dtlutmg the moculum and then companng the rccO\Wtes of all tuhe::. m tht: Jlluuon !--I.'Tles Thi~ then. ,..,the b.lsts oltht: i>.IP\l \also kno'' nand multtp!t: tube dtltuton LUhc, ot J duuon ILl he met hous) met hod

The mt:thoJ ofrero; real opponunlltes as a tool lor mtcrohtolo~tsls 1 n c;n u.n ions v. here, for one reason m anothc1.thc plate count method 1~ unsllltahlc It can also he employed for scm1-qu:mt ita live estt­maunn uf gro,,·th-promt)lllm L·apah11ity of hqu1J med ta tn estim.uion ol predswn fo1 alternate mi­nobtologK.t l mt:thoJ::. and a ... patt of the ··traction ncgaii\C method of dctnn11nin~ D-,:~lues

THE METHOD In the compcndt.tl ,·cr::.ton of the l'viP~ 1e..,t. the sample to be tc<.ted 1s ptcparcd m 10-lold dilution sencs, and then I ml sampk::. ol e,1(h dilutton arc !ll!Kubted into trtpltc:\lc broth nllturc tuhes f,1r incuh.1uon As tht: <.hlutlt'ns 1n<.:reasc, tht: pos::.tbtl-11~ that the broth tubes wtll btl to be llllltUbtcd w11h an) rn lro01~an1~tn 1!1lreasc-; At some pomt , therefore. \er} [c,, of thl n:phc.lle tubes wtll he tnOLUiatcd '' nh \ iablc mtcwor~ant~ms .

follow1ng lnLuh.llt<m. all tube-; arc cx.unined for LUrbtdlt}' .md the paucrn of grtm th m the rubes IS

scored .tgatnst a whlc of such values (4). rhe MPN t;rbk lrom the U~ l'ood and Drug Adm~ni~tratwn's Bmtawl AllLllytnal Mwllwl (BAM) 1::. prov1dcd in the Table (-1). At} pH.:al Jcs1gn li'>C::o three rcpltcatcs w1th a three log. un1t dtlutwn <.cries In this des1gn. tf all tubes showed ~mwth. then the re!-iuhs wtll be noted J~ 3B If nnl} om: lUht: ·n c.H.:h rt:phc:lle o::hows growth tl \\'llUitl he de noted .IS II I. rhe pattern ol growth tsthen read from the table Ill prtl\'tdc thl' mo ... t probJhle number and tJ')');, confidence tnt en al B) this, the r•:::-uh 111 2.1.0 would reflect an

M PN of l ) , and a rtstllt of 3.2.2 would be tnter­preted .rs an M P'\ of 210

hgurcs I .md 2 shuw thi:-; tn graphtc deprcuon. As the nlub.1teu tuhcs would he rc.1d 321. the MPN \\ouiJ be rc~.tlrdcd :JS 1 '50.

The ~lPN table (at least tht: ~lPN table'> tntht: h.umnm:cd mtcrohrallimtts WSb) wtll normallr onl}' present re~uhs for three dtluuons tn sc<.juenu: (e.g. 10 • 10 ' 10 '), hut the d luuon ::.cnes tested m1ght ha'-c been lrnm the 10 to 10' LUhcs tsee the rnA HA \1 diSlli">SlOn on how to <.dcd appwpnatc tuhes to read) The worker ''til need to take tile dtluuon fallors tn the table anJ 111 the actual experi­ment tnto accoum to dcnvc the mo::.t probable num ber from thts Sludy The rec;ults of dm test should be expressed as M PN rather th:tn colonr forming unit ~CFU) to reflect the c.:apaht!Htcs of the method.

The method as-.umcc, J random dt::.tnhution of micronrgan1~111~ tn the s.1mple and an accurate dilu­tiOn of tht: ~ample thrt1ugh the dtluuon -;encs. It also as~umcs that the tnllroorgantsms .lrc separate and dn not alfc<:t each other (<mmct or repel) (thus an: descnbcd b}· the Pots~on dt~tnhuuon at ILlW numbers) In addltlon. n must be a~~umcJ that ev­ery tube (or plate. etc.) whose inoculum has <l stnglc \'t.lble oq~amsm w1ll result in \'i'\lhle gro'' th .

Ahh(.>ugh the wmpendtal ver ... ion uuli:c~ thr<'c replit<lh:S and a [(~n-foiJ d1luuon seric~ . there 1::. no theorcucal rca::.on lor these parameters In fact , 11 1s well kilO\\ 11 th.ll the :JCCUI.Il} or tht: TllClht>d In·

crease::. dramauull)' wah tnett:a~tng the number of replicates and decn:.t-,tng the tlllcn.rl of the dtlution senes (fl\·c-[old 01 tW<1-Iokl) (5, 6) 1 he rDA RAM wehsuc rdcrenccd pmv1des an Excel spreadsheet to :lsSt!>t m ncaung dtfkrcnt MPN tahlcs as needed ('I)

MPN AND D-VALUE DETERMINATIONS Pnug ct al. ( 7) Jcscrihe the usc oft he M PN method in D-vnluc detcrminatwn from a htstoncll perspcc­t ivc:. Rtgt:low and Est}' (8) dcc;t. nbc the concept ol tht:rmal <.k:.nh lime:· whKh I':> hasu.allv the numher

of :::.Jmples of SUT\' I\'111g )l!ologtca) indlC<llOrS at diller emumes The tht·rmal dt.ath time for a temperature was then bel ween the longest he .rung ume } lekhng <l p•l::.lll\l' untt .md the shortt:Stlln1l' when <lll were

TABLE: MPN table for a three-replicatt design from FDA's Bacterial Analytical Manual.

Positive Tulles Pesltlve Tabes

D.l D.Dl 0.011 lPN 95% Confidence Ranae 0.1 0.01 0.001 IPII 95% Confidence Raqe

0 0 0 <3.0 0-9.5 2 2 0 21 4.5-42

0 0 3 0.15-9.6 2 2 28 8.7-94

0 0 3 0.15-ll 2 2 2 35 8.7-94

0 1 6.1 1.2-18 2 3 0 29 8.7-94

0 2 0 6.2 1.2-18 2 3 36 8.7-94

0 3 0 9.4 3.6-38 3 0 0 23 4.6-94

0 0 3.6 0.17-18 3 0 1 38 8.7-110

0 I 7.2 1.3-18 3 0 2 64 17-180

0 2 11 3.6-38 3 0 43 9-180

1 0 7.4 1.3-20 3 1 1 75 17-200

1 11 3.6-38 3 1 2 120 37-420

1 2 0 11 3.6-42 3 1 3 160 40-420

1 2 1 15 4.5-42 3 2 0 93 18-420

1 3 0 16 4.5-42 3 2 150 37-420

2 0 0 9.2 1.4-38 3 2 2 210 40-430

2 0 1 14 3.6-42 3 2 3 290 90-1000

2 0 2 20 U-42 3 3 0 240 42-1000

2 0 15 3.7-42 3 3 1 460 90-2000

2 20 4.5-42 3 3 2 llOO 180-4100

2 2 27 8.7-94 3 3 3 >1100 420-4000

negative. ln this procedure it was important to use APPLICATIONS OF MPN IN ENVIRONMENTAL more than one sample. A refinement on this method MONITORING DATA was proposed by Stumbo et al. (9), who suggested Environmental monitoring data are a problem for the use of the MPN equation on the surviving units. microbiology. We are urged to Mqualify" our control As there was only one possible dilution, the number levels and our sample sites. However, we are us-of replicates must be representative to provide some ing a technology (plate count) that is exceedingly precision to the estimate. By varying the heating imprecise at numbers of CFU below 25. However, times we start to develop the survivor curves using the aseptic core of a modern facility will commonly this method. There are two other commonly used yield counts of zero, with concern expressed if the methods for determining D-values based on this count is approaching three. These control levels MPN equation-the Stumbo-Murphy-Cochran and are m truth of little value despite their popularity the Holcomb-Spearman-Karber methods (9) (which in regulatory ctrcles (ll, 12, 13). One approach are both a bit complicated to do justice in this over- suggested to deal with this mismatch between vtew arucle See reference 10 for a review). regulatory expectations and plate count capabilities

has been to explore the possibility of looking at a

Fiaure 1: Three-tube design fof MPN (IJlincubated).

10 1

10''

frequency distribuuon models to establish control levels m these areas (14) or incident models (15).

A recent publication by Sun et al. (16) pointed out the possibilities in usmg MPN methods for evalua­tion of clean room monitoring data. The basic idea is to use the fundamental statistics as if only a single dilution were being considered. In this approach, the application is not dissimilar to fraction nega-tive studies of biological indicators for sterilization studies. Preliminary studies presented by this group look promismg and this is an approach that could be pursued with exiStent data for evaluation.

MPN IN QUALIFICATION OF BROTH MEDIA The compendia! chapters on microbial limits tests and sterility tests both place great emphasis on med1a growth promotion studies as a requisite qual­ity control acuvity. This has been reinforced by the revised 2010 USP chapter <1117> (17) discussion of the importance of media control in the lab. While the methods at hand to compare bacterial growth on solid media are quantitative in narure (recovery with-

figure 2: Three-tube desi!J1 fof MPN {klcubated).

in 50% or within 70% by CFU), the tools described in the compendia for broth growth promotion are qualitative at best. "liquid media are suitable if clear­ly visible growth of the microorganisms comparable to that previously obtained with a previously tested and approved batch of medium occurs" (17).

Weenk (18) reviewed dtfferent methods of demonstrating equivalence between two media. His study looked at the different methods for both solid and liquid media, but we will look only at the methods for liquid. He looked at the following to compare nutrient broth performance:

• Length of lag phase in each media from identical inocula

• Growth rate in each media • Endpoint determination

• Total biomass at stationery phase • Agar inoculation (number of CFU at

stationery phase) • Adding agar to 15 gil nutrient broth, then using

quantitative enumeration techniques • MPN.

lit: found the length of lag phase to be wtdely vari­Jblc, thr growth rat!! <.letermmation fur cath media prohtbttt\'d}' IJbor-mtens1ve. The endpoint determi­n.ltl()nS ~we klund to be too 1mpretise :md labor m­ten~t,·c . L)nc wonders whJt he would have concluded from I he harmom.:ecl method or JnOculallng wnh <100 CFU aml appm\'lng it tf grov. th occurs

The MP'\ method was recommended by Weenk t 18) to qualify media and is rcwmmendcd for its ac:tttracy. Cl1St, and flextbJIIt}' TI1e basic approach n!(ommcndcclts to approach the MP). method lmm the opposttc dtreuion than th:~t of the com­pendtal btoburdcn MPN In the b1C1hurden test. we h:we a :;ample wHh an unknown bioburclen ;md we are trymg to deduce the most probable number of cclb In the recommended growth pro­nwuon te!>L !nr ltqutd medtJ, we have t Wl1 batches d1spcnscd tn tubes The "sample" is a known in­oculum 111 a kn1.1wn diluttnn ::.cnes The inoculum dtluuon:; are -.ceded tntn the two med1a and then mcubatccl If llw medt.l cxhtbll tdentical growth promotmg properues, thl'n the 95% confidence mtenab llf the two MPK determinations should u\·erlap In lht::. manner, a \!>emD-quanmam·e growth rromollon -.rudy ma}' he performed for ltqu1d med1a

lo ':>etthl!> up. nne de:.1gn m1ght be to prepare twelve 10 mltubc:s of each med1um lhe mocula thnwe\t:r many rmcroorgamsms used) are then pre­pared to an approxunate concentration of 50, 5, 0.5. and 0.05 LrlJ/tube (thl:. has to be approximate only). blch drlutton IS then added to three tubes of each meclta, and 1 he med1a m~.:ubatcd. After incubation, 1 he MPN ts scparatelr deduced lor ci!<.h media­hopdullr the eonftdencc mtervafs wlll overlap.

MPN IN QUALIFICATION OF ALTERNATE (RAPID) MICROBIOLOGICAL METHODS It shl1uld be olwious to the reader that the previous di<.<.:Usston em the U!>C of MP:--: tn growth promotwn ~tud rc-. ha:; 1mmedtatc applllallon for determinat1~1n ,,[ 1 he rclau,•e I 10111 ~)f Detection for 1 wu mlcrobru­logrcal rnc1huds (e.g .. a "traduional" method and an "ahcrn<ttc methl.ld) I hl5 ts in fact referen<.:cd in USP <.hap1er < 1223> (I IJ).

The author ltrst realized thrs dunng prcpara-uon to teach at the 200-1- PDA-TRI course ·Raptd 1\;llcrobtOlOglcal Methods'' (Ocwher 20lH) Usmg the method de:'.<.rthed for liquid medra . qualttati\-c methods t"ould lJSII}' be wmpared fnr "lim II of Dctecuon Th1~ me1hod evcmuallr 1ppe;ned m rhc fmalt:ed Lhapll:r t20)

The U'>P l haptcr recommends tht:. approath even for quantltall\c methods. Although thts mtght at f1rst seem counter-intuttive Lhc MP:-.J method (when used wnh ;1 dtlut10n senes~ L<m •Ktually be nlllrc accumtc than plate counts a1 low numbers. espcc1ally tf five repltcates arc used ,md the dilu­tiOn series 15 lcss1han Ill-fold (the 10-fold dilution series 1s common only because nf traduion and the :~vatiabJI 1 ty of !>l:tndnrd tabJcs-1 here IS nn thcoreti­LJI obstacle to a 'l-fold dilul 1nn sene.;,, 01 3-lnld, or a 2-fold, or others) The Lmly mochnc.llton that needs In be made ts Ill tgnl1rr the counts and treat every plate or membrane a!> a separate ·tube''-the Ml'N method Ins nght tnlo the expcnmcmal des1gn

SUMMARY The basK LOncept for the Ml't\ method ts to d1lute the sample to such a degree that tnOl'ttla in the t ubc:; '' 1ll someumc:. ! but not always) contam nabk llrganisms B} rl'phtates. and d1luuon scnes. thts \\ 1ll result m a btrh a\.:( Urate esnmate nf the most probable number ,,f cells m the sampk. Whtle Lh1s method ts nwst comrnonh· used m the personal produtt:;, medtL.ll dcvJce. and pharmaceutical QC mtcrob1olngy labs for water tcstmg or the compendia! bioburden te:>t , 11 has s1gntftc1m potcnllal for other apphcat ions. These possible applic,ll ions of MPN Include D-value dctermtnat lon, envmmmental mont­to1ring, media growth promotion Slttdtes. and aspel't<; ol the validauun uf rapid mlcrt,hiolL,gkal met hods.

REFERENCE S li'>P, t h.optn c:hb •MI• r,>ht"l''~ll.tl r-~~mon.ollt>n ul

t'h•nsll'lllo l'h'dlh" Mtt ro>h1.d Enun1t 1.\lh>n k~l~.'- l 'mr.-,1

5tllto Phtlllll(lto'f'<'hl \2 , vc•l I, pp i l 7'i ,l,ll\4,,

l U~l' "\lt•u•ha.aii.Jmll lc,b,' l ':->1' .\\'/ll , p }Hn. ll.li tl

J ll;ll\'<ll'<>n, II (,_1 .and N R Zl•·~l,•r "i\pph,·.llhlll <II 'o1JII5lk~

In Pruhl~~~~~ In 8.11 lcnt'lO)\V II A Cnn,aunatJon ulth" :\•-

curacy of Dilution Data Usin~ a Smglc Dilution," j llallcriol.

26HU11-339, 1933.

4. FDA, Bacterial Ana/vtica/ Manual, Appendix 2, hup://www.

ld a .goi'/Food/Suence Research/Laboratory Met hods/Bactc­

riologicalA nalyucalManual BAM/ucm 1 Cl9o'i6. hun

BAM Excel spreadsheet Gin be d,>wnk><tdcd fwm the WW\\'.

lda.gol' site at http://www.fda.gel\'/dmvnloads/Fo,,d/Sci­

ence Research/LaboratoryMet hods/Bactc riologica It\ na lyt i­

cal Manual BA~I/UCM 164329.zip

S. Eisenhart, C and W Perry, "Statistical Methods and Cnntn>l

In Bacteriology:· Bactcrio/ Rn·. 7:57-137, 1941.

fl. Cochran, W., "Estimation ,,f Bactenal Densities by

Means of the Most Probable Number," Bwmctrics. o: 105-

116. 1950.

7. Pflug. l), et aL, "Principles uf the Thnm<tl Dcstructi,>n of

MICroorganisms:· Disinfection, Sicnli~alion and Pttvn·ution,

'5th Eel. SS Block led), Lippincott Willi<tms & Wilkins

pp79- l 14, 2001.

tl . Bigelow, W.O. and JR Esty, "The Thermal Death Point In

Relauon to Time of Typical Thermophilic Organisms:· J ln{CCI D1s 27602-617, 1920.

9. Stumbo. CR .. jR Murphy and J Cc>chmn, ''Nature of

Thermal Death Time Curws ll>r P.A. 3679 and Ck>stridium

Botulmum:· Food Tcclmol., pp. 121-32o, 1950.

10. Takahashi, M, et aL, "Comparison uf D-V<tluc Accuracy

Among Limited Speannan-Karbcr Procedure. Stumho-Mur­

phy Cl>chran Procedure (SMCP1 and Limited SMCP," PDAJ

Pharm Sci Tc,h. 49(5):220-22'i , 1995.

11. Hussong, D andRE Madsen , "Analysis ,,f Enl'imnmcntal

Microbiology Daw From Clean room Samples," Pharm Tccil­

nol Aseptic P•oc:l0-15, 2004.

12 . Farnngton,jK., "[nnronmental Monitoring in Pharmaccu­

ttcal Manufacturing-A PrcHluct Risk Issue." Amcr Ph11n11 Rn·

tl\4):26 -30, 2005.

13. Agalloco, J and Akers, )., "The ·Myth ' Called Sterility,"

PhtJrm Tecil, Bioproccss Sterile Mmnij(I<I:s44-s4'i, 2010.

1-l. Caputo, RA and A Huffman, "Environmental Monitoring:

Data Trending Using a Frequency Model ," PDAJ Pharm Sci

Tech. 5H(5):254-260, 2004.

15. USP, Chapter <1116> "Microbiological Cont rol and Moni­

lOring Environments Used for the Manufacture of Health­

care Products," Pharm Forum 33(3), 2007.

lo. Sun, X et al , "The Expanded Application of Most Probable

Number to the Quantitative Evaluation of Extremely

Low Microbial Count," PDAJ Pharm Sci Tccil 60(2U 24-

1 H, 2006.

17. USP, Chapter <1 117> "Microbiological Best Laboratory

Practices," USP 33 Reissue Second Suppkmcnl, pp. R-

1100-R-1105, 2010.

1H. Weenk, G. H .. "Microbiolu~ical Assessment ol Culture

Media: Comparison and Statistical haluation of Methods ,"

Inti) Fvod Miuobiol. 17:159-181, 1992.

19. USP, Chapter <1223> "Validation of Alternative Microbio­

logical Methods,'' USP 32 vol 1 pp730-3, 2009b.

20. USP, Chapter< 1223> "Validation of Alternative Microbio­

logical Methods," USP 29 Suppl 2 pp. 3807-3810, 2006.

GXP

ARTICLE ACRONYM LISTING BAM CFU FDA MPN QC USP

Bacterial Analytical Manual

Colony Forming Unit US Food and Drug Administration

Most Probable Number Quality Control

United States Pharmacopeia

ABOUT THE AUTHOR Scott Sutton , Ph.D., is owner and operator of The Microbiol­ogy Network lwww.microbiolorg), which provides se rvices to m icrolllolo~y-relatecl user's groups. Dr. Sutton may be reached hy e-mail at scott. snlllln@microbiolorg.