Use of the Bergum Method and MS Excel to Determine

The Probability of Passing the

USP Content Uniformity Test
Pramote Cholayudth

O
Using the Bergum Method and the MS Excel ne of the most well - k n own basic principles of quality
software program, the author determines assurance is that qu a l i tycannot be inspected or te s ted
the probability of passing the USP Content into a finished produ ct . That is, reliable produ ct qual-
Uniformity test. ity means more than just meeting the specifications. If
100 random samples are taken and tested , then a partial nu m ber
of samples wi ll pass the test depending on the overall qu a l i tyvari-
a bi l i ty. The nu m ber of samples passing the test wi ll re ad i ly in-
crease and su b s equently approach 100 (= 100%) wh en the qual-
i ty para m eters are more and more con s i s tent and centra lly within
the specifications.
Companies want assu ra n ce that any sample taken from a
product lot has a high percentage (say 95%) of passing the of-
ficial tests. Such a high percentage—in terms of a high proba-
bility for a sample to pass the official tests—will be established
t h ro u gh set ting new tigh ten ed spec i f i c a ti ons. Th e s e
spec i f i c a ti on s , i.e., the new accept a n ce limits, wi ll be
Formula 1 for Excel calculation for tablet content uniformity the benchmark for each particular product quality.
Formula 1 James S. Bergum’s 1990 paper entitled “Construct-
= M AX((FDIST(10/((0.06^2)*(1 1 10* (Mean/Sigma)^2)),(1 1 10*(Mean/Sigma)^2)^2/ ing Accept a n ce Limits for Mu l tiple Stage Tests,” in-
(1 1 2*10*(Mean/Sigma)^2),10 2 1)) 1 ((NORMSDIST((Mean 2 8 5 ) / Sigma) 1 troduced the method by which to compute the prob-
NORMSDIST((115 2 Me a n ) / Sigma) 2 1)^10) 2 1 ,(FDIST(30/((0.078^2)*(1 1 ability of passing the multiple-stage tests for a sample.
30*(Mean/Sigma)^2)),(1 1 30* (Me a n / Sigma)^2)^2/(1 1 2*30* (Me a n / Sigma)^2),30 2 Bergum’s method is one of the most important tools
1)) 1 ((NORMSDIST((Mean 2 8 5 ) / Sigma) 1 NORMSDIST((115 2 Me a n ) / Sigma) 2 1)^30 for content uniformity assessment. In this article, the
1 COMBIN(30,1)*((NORMSDIST((Mean 2 85)/Sigma) 1 NORMSDIST((115 2 Mean)/Sigma) prob a bi l i tyof passing the USP con tent uniform i ty te s t
2 1)^29)*(((NORMSDIST((Mean 2 7 5 ) / Sigma) 2 NORMSDIST((Mean 2 8 5 ) / Sigma)) 1 for tabl ets and capsules according to Sampling Plan 1
(NORMSDIST((125 2 Me a n ) / Sigma) 2 NORMSDIST((115 2 Me a n ) / Sigma)))^1)) 2 1,0) (one assay is performed on each of L locations) in
The above formula may be downloaded from the Pharmaceutical Technology Web site (www.pharmtech.com).
Ber g u m’s paper and the corre s pondng MS Excel com-
putations are discussed.

Content uniformity of tablets

As de s c ribed in the Un i ted St a te s Ph a rm a copei a , the short pro-
cedure used to determine the content uniformity test result is
as follows:
● Acceptance cri teria stage 1. As s ay 10 tabl ets. Pass if both of
the fo ll owing cri teria are met: relative standard deviation
Pramote Cholayudth is the managing (RSD) < 6.0% and no va lue is outside 85–115% of claim. Fail
director of the Professional Conference
Center, Ltd., 6/756 Number One Complex,
if one or more values are outside 75–125% of claim. Other-
Bangkok-Ram 2 Road, Pravate District, wise, go to stage 2.
Bangkok 10250, Thailand, fax 662.740.9586, ● Acceptance criteria stage 2. Assay 20 more tablets. Pass if for
cpramote2000@yahoo.com. all 30 tablets the following criteria are met: RSD < 7.8%, no
m ore than one value is out s i de 85–115% of cl a i m , and no
86 Pharmaceutical Technology SEPTEMBER 2004 w w w. p h a rm t e c h . c o m
 Table I: Probability of passing the USP content uniformity test for tablets (%).
Sigma 5 2 Sigma5 4 Sigma 5 6
Mean Bergum Excel Simulation Bergum Excel Simulation Bergum Excel Simulation
86 2.5 2.5 2.5 0.0 0.0 0.6 0.0 0.0 0.2
90 98.5 98.5 99.1 26.8 26.8 36.8 0.0 0.0 9.0
94 100.0 100.0 100.0 94.8 94.8 96.8 34.7 34.7 53.7
98 100.0 100.0 100.0 100.0 100.0 100.0 88.4 88.4 92.0
100 100.0 100.0 100.0 100.0 100.0 100.0 93.4 93.4 95.7
102 100.0 100.0 100.0 100.0 100.0 100.0 89.6 89.6 92.9
106 100.0 100.0 100.0 94.8 94.8 97.0 38.8 38.8 55.8
110 98.5 98.5 99.1 32.4 32.4 36.4 0.9 0.9 10.0
114 2.5 2.5 2.5 0.4 0.4 0.5 0.0 0.0 0.3

re sult is 85–115% of claim and p2 is the

Table II: Probability of passing the USP content uniformity test for capsulses (%). prob a bi l i ty that an assay re sult is 75–85%
Sigma 5 3 Sigma5 5 Sigma 57 of claim or 115–125% of claim.
Mean Bergum Excel Bergum Excel Bergum Excel Prob (C11) and Prob (C21) are assoc i a ted
86 6.6 6.6 0.0 0.0 0.0 0.0 with the noncentral F d i s tributi on of sam-
90 94.7 94.7 27.0 27.0 0.0 0.0 ple size / C V2 (n/CV2) and also the non-
94 100.0 100.0 97.6 97.6 27.2 27.2 central T distribution of
98 100.0 100.0 100.0 100.0 73.6 73.6
100 100.0 100.0 100.0 100.0 80.0 80.0
102 100.0 100.0 100.0 100.0 82.0 82.0 [4]
106 100.0 100.0 97.6 97.6 52.3 52.3 To com p ute the noncen tral F prob a bil-
110 94.7 94.7 43.6 43.6 0.1 0.1 ity, Bergum used Jo h n s on and Kotz’s
114 6.8 6.8 0.0 0.0 0.0 0.0 method to approx i m a te a non central F
d i s tri buti on from a cen tral F distri bution
va lue is outside 75–125% of claim. Ot h- cri tical va lue on the basis of the fo ll owi n g
erwise, fail. relationship (4):
Probability bo u n d. To determine the [1]
prob a bi l i ty of passing the USP c ri teria, the in which n is the sample size (10 tablets),
following variables are defined: C11 is the k is the number of tablets outside the [5]
event that cri teri on 1 in stage 1 is met , C12 85–115% range but still within 75–125% in wh i ch F(n, df, 0) and F(1, df, l) are cen-
is the event that cri terion 2 in stage 1 is of claim, n! is n factorial, p1 is the proba- tral and noncen tral F distri bution cri ti c a l
met, C21 is the event that cri terion 1 in bility that an assay re sult is within 85– va lu e s , respectively, with n, df, and 1 de-
stage 2 is met, and C22 is the event that cri- 115% of cl a i m , and p2 is the prob a bi l i ty grees of freedom and
terion 2 in stage 2 is met. that an assay re sult is within 75–85% of
Using these definitions, Prob(passing claim or is 115–125% of claim.
USP criteria) 5 Prob(first 10 tablets Using Equation 1,
meet USP stage 1 criteria or all 30
tablets meet the USP stage 2 criteria) >
MAX{Prob(first 10 tablets meet USP [6]
stage 1 criteria), Prob(all 30 tablets meet Th erefore, P RO B F (x, n1, n2, l) 5
USP stage 2 criteria)} > [2] P RO B F (x/(1 1 l), n, n2, 0) and PRO B F
MAX {Prob(C11) 1 Prob(C12) 2 1, If A is an individual assay re sult, then (n/ C V2, n1, n2, l) 5 P RO B F ( (n/CV2)/(1
Prob(C21) 1 Prob(C22) 2 1} Prob (C22) is a functi on of Prob(75 < A < 1 l), n, n2, 0). PROBF(x, n1, n2, l) is the
Pro b a b i l i ty of passing tablet co nte nt uni- 85) and Prob(115 < A < 125), sample probability density function for noncen-
fo rm i ty. Prob (C12) and Prob (C22) are as- s i ze , and the nu m ber of t a bl et(s) falling tral F distri buti on at va ri a ble x (5 n/CV2)
sociated with binomial distri buti on as fol- within 75–85% or 115–125% of cl a i m . with n1 and n2 degrees of freedom and a
lows: Let A be an indivi dual assay result. Prob(C22) can be determined as follows: non cen tra l i ty para m eter l. PROBF(x/(1
Prob (C12) is a functi on of Prob(85 < A < 1 l), n, n2, 0) is the prob a bility density
115), sample size , and the nu m ber of functi on for cen tral F distributi on at a
t a bl et(s) out s i de 85–115% of cl a i m . Th e variable x/(1 1 l) [5 (n/CV2)/(1 1 l)]
bi n omial prob a bi l i ty den s i ty functi on is with n and n2 degrees of freedom and a
defined as [3] noncentrality parameter of l.
in which p1 is the prob a bi l i tythat an assay
88 Pharmaceutical Technology SEPTEMBER 2004 w w w. p h a r m t e ch . c o m
 Table III: Comparisons of probability by Table IV: Comparisons of probability by Table V: Comparisons of probability by
Bergum’s CuDAL software and Excel Bergum’s CuDALsoftware and Excel for Bergum’s CuDAL software and Excel
for tablets (n 5 10). Acceptance tablets (n 5 30). Acceptance limits for capsules (n 5 10). Acceptance
limits for tablet content uniformity.* for tablet content uniformity.* limits for capsule content uniformity.*
Mean Critical Probability Mean Critical Probability Mean Critical Probability
(% claim) % CV by Excel (% claim) % CV by Excel (% claim) % CV by Excel
95.0 2.23 95.15 95.0 3.15 95.06 95.0 2.80 94.97
96.0 2.43 95.08 96.0 3.43 95.04 96.0 3.04 95.03
97.0 2.63 94.92 97.0 3.70 95.08 97.0 3.27 94.98
98.0 2.82 94.91 98.0 3.97 94.97 98.0 3.46 95.03
99.0 3.00 95.00 99.0 4.21 94.95 99.0 3.60 94.92
100.0 3.17 95.07 100.0 4.39 95.03 100.0 3.68 94.98
101.0 2.94 95.02 101.0 4.13 94.98 101.0 3.61 95.08
102.0 2.70 95.14 102.0 3.81 95.05 102.0 3.36 94.95
103.0 2.47 95.10 103.0 3.49 94.98 103.0 3.08 95.10
104.0 2.25 94.88 104.0 3.17 94.96 104.0 2.81 94.94
105.0 2.02 95.07 105.0 2.85 95.06 105.0 2.53 95.07
*Meeting limits guarantees with 90% *Meeting limits guarantees with 90% *Meeting limits guarantees with 90%
assurance that at least 95% (by Bergum assurance that at least 95% (by Bergum assurance that at least 95% (by Bergum
software) of all samples tested for content software) of all samples tested for content software) of all samples tested for content
uniformity will pass the USP test. uniformity will pass the USP test. uniformity will pass the USP test.

Prob(C11) 5
Formulae (parts 1 and 2) for Excel calculation for capsules FDIST(10/((0.06^2)*(1 1 10*
Part1 (100/6)^2)), (1 1 10*
=(FDIST(10/((0.06^2)*(1 1 10*(Mean/Sigma)^2)),(1 1 1 0 * ( Me a n / Sigma)^2)^2/(1 + 2*10* (100/6)^2)^2/(1 1 2*10*
(Mean/Sigma)^2),10 2 1)) 1((NORMSDIST((Mean 2 8 5 ) / Sigma) 1 NORMSDIST((115 2 Me a n ) / Sigma) 2 (100/6)^2), 10 2 1) 5 56.24%
1 ) ^ 1 01COMBIN(10,1)* ((NORMSDIST((Mean 2 85)/Sigma) 1 NORMSDIST((115 2 Me a n ) / Sigma) 2
1)^9)*(((NORMSDIST((Mean 2 7 5 ) / Sigma) 2 NORMSDIST((Mean 2 8 5 ) / Sigma)) 1 (NORMSDIST((125 2 Prob(C21) 5 FDIST(30/
Me a n ) / Sigma) 2 NORMSDIST((115 2 Me a n ) / Sigma)))^1)) 2 1 ((0.078^2)*(1 1 30*(100/6)^2)),
Pa rt 2 (1 1 30*(100/6)^2)^2/(1 1
5(FDIST(30/((0.078^2)*(1 1 30* (Mean/Sigma)^2)),(1 1 30*(Mean/ Sigma)^2)^2/ (1 1 2*30* 2*30*(100/6)^2), 30 2 1) 5
( Mean/ Sigma) ^2),30-1)) + ((NORMSDIST((Mean 2 8 5 ) / Sigma) 1 NORMSDIST((115 2 Mean)/Sigma) 2 98.81%
1)^30 + COMBIN (30,1)*((NORMSDIST((Mean 2 85)/ Sigma) 1 NORMSDIST((115 2 Mean)/ Sigma) 2
1)^29)*(((NORMSDIST ((Mean 2 75)/Sigma) 2 NORMSDIST ((Mean-85)/Sigma)) 1 (NORMSDIST ((125 2 Using Microsoft Excel
Mean)/Sigma) 2 NORMSDIST((115 2 Mean)/Sigma)))^1) + COMBIN(30,2)*((NORMSDIST((Mean 2 85)/Sigma) To compute the probability, one must trans-
1 NORMSDIST((115 2 Mean)/Sigma) 2 1)^28)*(((NORMSDIST((Mean-75)/Sigma) 2 NORMSDIST((Mean 2 form the locations at 75, 85, 115, and 125%
8 5 ) / Sigma)) 1 (NORMSDIST ((125-Mean)/Sigma) 2 NORMSDIST((115-Mean)/Sigma)))^2) 1 of claim into the corre s ponding Z score s
COMBIN(30,3)*((NORMSDIST((Mean 2 85)/Sigma) + NORMSDIST ((115 2 Mean)/Sigma)-1)^27)* (((NORMSDIST using the population mean m and standard
((Mean 2 75)/Sigma) 2 NORMSDIST((Mean -85)/Sigma)) + (NORMSDIST ((125-Mean)/Sigma) 2 devi a ti on s. To calculate the Z score:
NORMSDIST((115 2 Mean)/Sigma))) ^3)) 2 1
The above formulae may be downloaded from the Pharmaceutical Technology Web site (www.pharmtech.com).
[7]
in wh i ch x is the critical locati on va lu e
To compute Prob ( C11) or Prob ( C21), an Prob(C21) 5 PROBF((30/CV2)/ (i.e., 75, 85, 115, or 125% of claim) at
MS Excel formula is derived from (1 1 l), (1 1 l)2/(1 1 2l), 30 2 which the Z score is calculated.
PROBF((n/CV2)/(1 1 l), n, n2, 0) 1, 0) = FDIST(30/((0.078^2)* The probability at a given Z score may
as follows: (1 1 30*(Mean/Sigma)^2)), be obtained with the Excel formula:
Prob(C11) 5 PROBF((n/CV2)/ (1 1 30*(Mean/Sigma)^2)^2/
(1 1 l), n, n2, 0) 5 (1 1 2*30*(Mean/Sigma)^2), 5NORMSDIST(Z) 5
PROBF((10/CV2)/(1 1 l),(1 1 30 2 1) NORMSDIST ((x 2 m)/s)
l)2/(1 1 2l), 10 2 1, 0) 5
FDIST(10/((0.06^2)*(1 1 10* For example, if Sigma (the population Finding p1 and p2 is essen tial. For ex-
(Mean/Sigma)^2)),(1 1 standard deviati on) and Mean (the mean) ample, if s 5 6 and m 5 100 (maximum
10*(Mean/Sigma)^2)^2/(1 1 are substituted by 6 and 100 respectively, s required by USP content uniformity is
2*10*(Mean/Sigma)^2), 10 2 1) the results are 10), p1 is the area between 85% and 115%
90 Pharmaceutical Technology SEPTEMBER 2004 w w w. p h a r m t e ch . c o m
of claim, and its calculation in Excel is as No te that from the su b s ti tuti on above , ● Accept a n ce criteria stage 2. As s ay 20
follows: Prob(C11) 5 56.24% and Prob(C21) 5 more capsules. Pass if, for all 30 capsules,
98.81%. From the probability cri teria the fo ll owing cri teria are met: RSD <
5NORMSDIST ((115 2 100)/6) a bove , a com p l ete MS Excel formula for 7.8% and no more than three va lues are
2 NORMSDIST((85 2 100)/6) the tabl et con tent uniform i ty (CU) is de- out s i de 85–115% of claim and no va lu e
5 0.98758 rived in sidebar “ Formula for Excel cal- is out s i de 75–125% of claim. Ot h erwi s e ,
culation for tablet content uniformity” fail.
For p2, the calculation in Excel is (see opening page). After simulation, the Probability bo u n d. To determine the
re sults from both formu l ae (conven ti on a l probabi l i ty of passing the USP cri teri a , the
5(NORMSDIST((125 2 100)/6) 2 and Excel) are alw ays equal for the same following variables are defined: C11 is the
NORMSDIST((115 2 100)/6)) 1 s ets of pop u l a ti on mean and standard de- event that cri teri on 1 in stage 1 is met ; C12
(NORMSDIST((85 2 100)/6) viati on . Ta ble I shows the com p a ri s ons of is the event that cri teri on 2 in stage 1 is
2NORMSDIST((75 2 100)/6)) = the re sults from the Bergum met h od (in met; C21 is the event that criteri on 1 in
0.01239 Bergum and Simulation columns). s t a ge 2 is met ; and C22 is the event that cri-
terion 2 in stage 2 is met. Then
Therefore, Content uniformity of capsules
As described in USP, the short procedure Prob(passing USP) >
MAX{Prob(C11) + Prob(C12) 2 1, to determine the content uniformity test MAX{Prob(C11) 1 Prob(C12) 2 1,
Prob(C21) + Prob(C22) 2 1} result is as follows: Prob(C21) 1 Prob(C22) 2 1}
5 MAX{0.5624 1 p110 2 1, 0.9881 1 ● Acceptance cri teria stage 1. Assay 10 cap-
p130 1 30p129p2 2 1} sules. Pass if both of the fo ll owing cri- Pro b a b i l i ty of passing capsule co nte nt uni-
5 MAX(p110 2 0.4376, p130 1 30p129p2 teria are met: RSD < 6.0% and no more form i ty. The probability of passing capsule
2 0.0119) 5 MAX(0.9875810 2 than one va lue is out s i de 85–115% of con tent uniform i ty can be calculated as
0.4376, 0.9875830 1 30*0.9875829* claim. Fail if one or more va lues are out- follows:
0.01239 2 0.0119) 5 MAX(0.4449, side 75–125% of claim. Ot h erwise, go
0.9341) 5 0.9341 5 93.41% to stage 2.

92 Pharmaceutical Technology SEPTEMBER 2004 w w w. p h a r m t e ch . c o m

 0.8253 1 p1 30 1 30p129p2 1 Plan 1 (one result from each loc a tion).
435p128p22 1 4060p127p23 2 1} Su ppose a con tent uniform i ty test re-
5 MAX(0.2813, 0.8004) 5 0.8004 sult from the first stage (n 5 10) has a
5 80.04% mean of 98% of label claim (LC) and RSD
[8] 5 2.5%. Using Bergum’s table (see Table
No te that from the su b s ti tuti on above , III), a cri tical RSD value of 2.81% is re-
Prob(C11) 5 32.30% and Prob (C21) = qu i red for a sample mean of 98% LC to
82.53%. Using the prob a bi l i ty cri teri a , a en su re with 90% assura n ce that at least
com p l ete Excel formula for capsule con- 95% of a ll samples te s ted for con tent uni-
tent uniform i ty is derived as Formula II: formity will pass the USP test. The prob-
Probability of passing capsule CU test 5 abi l i tyfor this particular test result to meet
M AX ( Part1,Part2,0), in wh i ch Part 1 and the USP accept a n ce cri teria can be com-
[9] Part 2 are as shown in the sidebar “ For- puted using MS Excel as follows:
in wh i ch p1 is the prob a bi l i tythat an assay mu l ae (Parts 1 and 2) for Excel calcula-
re sult is in the ra n ge 85–115% of cl a i m ; tion for capsules” (see page 90). Upper(0.90^0.5)*100% Confiden ce
p2 is the probability that an assay result is Limit (Mean1)
_
in the range 75–85% of claim or Evaluation of sample data = x 1 UB*NORMSINV(1 2 (1 2
115–125% of claim. The probability of passing the test normally 0.90^0.5)/2)/n^0.5
Using Microsoft Excel. For ex a m p l e , if depends on the population mean and SD.
s i gma 5 7 and pop u l a ti on mean 5 100, In fact, these population parameters are un- Lower(0.90^0.5)*100% Con f i den ce
p1 and p2 are 0.967876 and 0.031769, known and can only be estimated using the Limit (Mean2)
_
respectively. So sample data. To calculate the lower bound = x2 UB*NORMSINV(1 2 (1 2
on the prob a bi l i tyof a futu re sample pass- 0.90^0.5)/2)/n^0.5(0.90^0.5)*100 %
MAX{Prob(C11) 1 Prob(C12) 2 1, ing the USP content uniformity te s t , on e Confidence Limit for SD (UB)
Prob(C21) 1 Prob (C22) 2 1} 5 must calculate the sample mean, RSD, and 5 SD*((n 2 1)/(CHIINV(0.90^0.5,
MAX{0.3230 1 p1 10 1 p19 p2 2 1, sample size from a batch using Sampling n 2 1)))^0.5

94 Pharmaceutical Technology SEPTEMBER 2004 w w w. p h a r m t e ch . c o m

 _
in wh i ch x is the sample mean 5 98% LC ; SD is the sample
_
standard deviation = RSD* x /100 5 2.5 3 98 4 100 5 2.45%
LC; n is the sample size 5 10; (0.90^0.5)*100% 5 94.87% (see
below); and UB is the upper bound.
Substituting the above parameters, the results are

Mean1 5 100.47% LC
Mean2 5 95.53% LC
UB 5 4.01 % LC

Substituting these values into Formula I gives

Prob1 2 (Substitution: Mean1 5 100.47, UB 5 4.01)

5 100.00%
Prob2 2 (Substi tution: Mean2 5 95.53, UB = 4.01) 5 99.23%

Th erefore , the ex pected prob a bility of passing the USP te s t

will be MIN(Prob 1 , Prob2) = 99.23%. No te that because the
sample RSD is 2.5, which is less than 2.81, the lower bound on
the probability will be greater than 95%. The 90% assurance is
simultaneously based on 94.87% confidence on the mean and
SD (i.e., 94.87 3 94.87 4 100 5 90%).
Wh en a 95% overa ll assu rance is requ i red , the fo ll owing for-
mulae are used:

Upper (0.95^0.5)*100% Confidence Limit (Mean1)

_
= x 1 UB*NORMSINV(1 2 (1 2 0.95^0.5)/2)/n^0.5

Lower (0.95^0.5)*100% Confidence Limit (Mean2)

_
= x 2 UB*NORMSINV(1 2 (1 2 0.95^0.5)/2)/n^0.5
(0.95^0.5)*100% Confidence Limit for SD(UB)
5 SD*((n 2 1)/(CHIINV(0.95^0.5,n-1)))^0.5

The ex pected probability (95% assu ra n ce) of passing USP

test is MIN(Prob 1 , Prob2) = MIN(99.95%, 93.62%) = 93.62%.
The calculati on is con du cted the same way for capsu l e s : Find
Mean1, Mean2, and UB, then substitute these values into For-
mula II and finally MIN(Prob1, Prob2).

Conclusion
To have reliable product quality, one must set a benchmarking
(i . e . , ti gh ten ed) accept a n ce limit that ref l ects a high prob a bi l-
ity of passing the official standard te s t . The USP con tent uni-
form i ty test is a mu l ti p l e - s t a getest. These types of tests contain
different accept a n ce cri teria for each stage . Determining the
prob a bi l i ty of passing these tests requ i res knowing the pop u-
lation para m eters that are unknown in actual applications. How-
ever, one can estimate the population parameters on the basis
of samples re su l t s . With its stati s tical and mathematical func-
ti on s , MS Excel sof t w a re is a useful tool for com p uting these
estimates as well as computing the probabilities of passing the
official test using the Bergum method.

Acknowledgment
This article is solely inten ded to support James Bergum’s method.
Although his met h od is ava i l a ble on CD, the author’s intenti on
to use MS Excel to compute the probability of passing the test
96 Pharmaceutical Technology SEPTEMBER 2004 w w w. p h a rm t e c h . c o m
 is just to provi de an altern a tive of com p ut a ti on . F i n a lly, the au-
thor is very gra teful to the revi ewer of this paper for his very
helpful comments.

References
1. “^9 0 5& Uniform i ty of Do s a ge Units” in USP 2 4 –NF 19 (The Un i ted
States Pharmacopeial Convention, Inc., Rockville, MD, 1999).
2. Mi l i t a ry Standard 105E (MIL-STD-105E), Sampling Pro cedu res and
Tables for Inspection by Attributes (Department of Defense, Washing-
ton, DC, 10 May1989), pp. 1–7.
3. S. Bo l ton , “ In trodu cti on to Probability” and “Single Unexplained Fa i l-
ing As s ay,” in Ph a rm a ceutical St a ti s ti cs: Pra ctical and Clinical Appl i c a-
tions (Ma rcel De k ker, In c . , New York , NY, 1997), pp. 62–95 and pp.
685–695.
4. S.C. Ch ow, and J. P. L iu , “USP Tests and Spec i f i c a ti on s ,” in St a ti s ti c a l
Design and Analysis in Ph a rm a ceu tical Sci en ce: Validation, Pro cess Co n-
trol s , Pra ctical and Clinical Appl i c a tions (Ma rcel De k ker, Inc., New
York, NY, 1995), pp. 158–165.
5. J.S. Bergum and M.L. Utter, “Statistical Methods for Uniformity and
Di s s o luti on Testi n g,” in Ph a rmaceutical Pro cess Validation: An In ter-
national T h i rd Ed i tion, Revi sed and Expa n d ed , R.A. Nash and A . H .
Wach ter, Ed. (Ma rcel De k ker, Inc., New York , NY, 2003), pp. 667–697.
6. J. Berman, and J.A. Planchard, “Blend Uniform i ty and Unit Dose Sa m-
pling,” Drug Dev. Ind. Pharm. 21 (11), 1257–1283 (1995).
7. G.J. Mer gen , “ Eva lu a tion and Va l i d a ti on of Bl end Uniform i ty: Ap-
proaches for Establishing Acceptance Criteria,” J. Validation Technol.
7 (2), 102–115 (2001).
8. P. Cholayudth, “Process Validati on Acceptance Cri teria for Solid Do s a ge
Forms,” Pharm. Engin. 23 (4), 102–114 (July–Aug. 2003).
9. FDA Center for Drugs and Biologics and Center for Devices and Ra-
diological He a l t h , “Gu i deline on Gen eral Principles of Process Va l i-
dation,” May 1987.
10. Bl end Un i form i ty Working Gro u p, Produ ct Quality Re s e a rch In s ti-
tute, “The Use of Stra ti f i ed Sampling of Bl end and Do s a ge Units to
Dem on s tra te Adequ acy of Mix for Powder Bl en d s ,” Final Report on
Bl end Un i formity Recommendations, 30 Decem ber 2002, www.pqri.org
/datamining/imagespdfs/123002sam.pdf.
11. FDA Cen ter for Drug Eva luati on and Re s e a rch, “ D raft Gu i d a n ce for
In dustry: Powder Bl ends and Finished Do s a ge Units—Stra ti f i ed In -
Process Do s a ge Unit Sampling and As s e s s m en t ,” O ctober 2003,
www.fd a . gov / cder/guidance/ 5831dft.pdf and www.fd a . gov / cder/
guidance/5831AttmtR.pdf.
12. Pa ren teral Drug As s oc i a ti on, PDA Technical Report No. 25: Blend Uni-
formity Analysis—Va l i d a tion and In - Process Testing, pp. S15–S17,
S44–S57 (1997).
13. Email message dated 11 June 2001, from Dr. Larry Paul, G en eral Po l i-
cies and Requ i rements, In form a ti on and Standards Devel opm en t ,
Un i ted States Pharm acopeial Conven ti on , rega rding the “Ori gi n a ti on
of RSD Requirem ents for the USP Content Un i form i ty Test,” www.usp.
org.
14. J.S. Bergum’s SAS Programs, “Appendix E: Lower Bound Calculations,”
Con tent Uniform i ty and Di s s o luti onAccept a n ce Limits (CUDAL),
version 1.0, dated 26 July 2003. PT

Please rate this article.

On the Reader Service Card, circle a number:
348 Very useful and informative
349 Somewhat useful and informative
350 Not useful or informative
Your feedback is important to us.

98 Pharmaceutical Technology SEPTEMBER 2004 w w w. p h a rm t e c h . c o m