UNIT 5 INTRODUCTION TO BIOASSAY

Structure
5.1 Introduction
Objectives
5.2 Bioassay or Biological Assay
5.2.1 Principle of Bioassay
5.2.2 Components of Bioassay
5.2.3 Role of Statistics in Bioassay
5.3 Types of Bioassays
5.4 Direct Bioassay
5.5 Relative Potency
5.6 Fieller’s Theorem
5.7 Summary
5.8 Solutions/Answers

5.1 INTRODUCTION
In Block 1 of this course, you have studied about the vital statistics which deals
with the methods used to analyse the data pertaining to vital events of the The potency is
populations specially birth, death, marriage, migration, etc., along with the defined as a strength
various methods used for the analysis of these vital events. In this unit, we shall or power of the
discuss bioassay which helps the scientists, pharmacists or doctors to decide stimulus or substance
the amount or dose of a particular medicine or drug which is optimal for administered to the
treatment. It also helps to avoid the dose which may be harmful for a patient. It subject.
may also help agricultural scientists to decide the amount of chemicals or
pesticides, which can be used in the fields to kill the pests and increase the
yield of a crop, etc.
This unit will provide you a general understanding and make you familiar with
bioassay by giving a brief overview of basic concept and components of
bioassay. In Sec. 5.2, we explain about bioassay, components of bioassay and
role of statistics in bioassay. In Sec. 5.3, we briefly discuss about various types
of bioassays. We elaborate on the direct assay in Sec. 5.4. We explore how to
draw a statistically valid conclusion in Sec. 5.5 on the basis of relative potency
of the test preparation with respect to standard preparation, its variance and
fiducial limits. Sec. 5.6 deals with the computation of fiducial limits of relative
potency using Fieller’s theorem.
In the consequent units, you will study about other types of bioassays in detail.

Objectives
After studying this unit, you should be able to:
• define bioassay and its components;
• distinguish between different types of bioassays;
• explain the direct bioassay;
• compute the relative potency and its variance; and 87
Bioassay
• determine the fiducial limits of relative potency.

5.2 BIOASSAY OR BIOLOGICAL ASSAY

Bioassays also known as biological assays are typically conducted on the living
organism to measure the effect of a biologically active substance under
controlled standard sets of conditions. The estimation of concentration or
potency of a specific substance by measuring and comparing the responses of
the test and standard preparations is called as bioassay. In other words, we can
define bioassay as a method for determination of the potency of a particular
substance by measuring its effect or response on the living organism.
The bioassays are considered to be essential during the development of new
drugs. It helps in analysing the efficacy as well as in determining the side
effects or degrees of drug toxicity, etc. The bioassays may be qualitative or
quantitative in nature while estimating the concentration or potency of a
substance by measuring the biological response produced by it. When the
physical effects of a substance may not be quantified, we consider qualitative
bioassays which involves quantitative bioassay. For example, to assess the
nonstandard development or abnormality.
Let us now discuss how bioassays are similar or different from usual biological
experiments.
Similar to biological experiments, bioassay also provides numerical assessment
of some of the properties of the substance which is to be assayed. An essential
part of this assessment is to take a measurement of growth or other
morphological changes in animals, plants, animal tissues, micro-organism, or
some other form of a living matter. We can say that a bioassay is a type of
biological experiment, but here the interest lies in comparison of the potencies
of the treatments on an agreed scale instead of comparing the magnitudes of
the effects of different treatments. So, we compare the potencies rather than the
magnitudes in a bioassay. The experimental technique may be same in case of
bioassay as we use in a purely comparative experiment. Since the purpose of
experiment is different in bioassay, the experimental design and statistical
analysis will also be different in this case. The following examples will explore
the difference more clearly:
• If we consider a study to investigate the effects of different doses of
insulin on blood sugar level of guinea pigs, this study will not necessarily
be a bioassay. We can consider it as bioassay if a researcher is interested
not only in the changes in blood sugar level but also in the estimation of
potencies of the doses on a scale with reference to a standard.
• In the same way, the yield of maize in a field trial due to various type of
fertilisers will not be generally considered as a bioassay until unless we
use the yield of maize to evaluate the potency of a fertiliser relative to a
standard one.
Before discussing the various components of bioassays, let us look at the
history and principle of bioassay.
In bioassays, we make use of biological organisms (Living beings which are
capable of responding to some external stimuli, e.g., animal, plant, fungi,
bacteria, etc.) to test the toxicity of chemicals, effect of a new drug, compare
the efficacy, etc. You may notice that the bioassays are employed in real life
88
and may be unknowingly sometime. You may have heard of coal miners Introduction to Bioassay
sending canaries to the coal mines and we may consider it as a traditional
example of a bioassay. The coal miners used to send the canaries down into the
mines to ensure a safe air supply before entering themselves. Since the canaries
are more sensitive to the methane gas in comparison to the humans, the miners
used to send them ahead of themselves to provide an advanced alarm of rising
levels of the methane or dangerous levels in the mines. The death of canaries
was considered as an alarm signal to the miners so that they can leave the mine
immediately.
Traditionally long back and area today, we considered various disorders and
diseases of the fishes as biological indicators of environmental problems which
usually indicate the presence of chemical and toxicity in the water. It can be
considered as a measure of the extent of the environmental pollution / waste
sinking and assimilating in river water, affecting flora and fauna. This can be
considered as an analogy to the miners’ canary used to monitor the rising levels
of methane in the mines.
The scientific work on biological assays can be noticed in late 19th century with
the standardisation of diphtheria antitoxin investigations of Paul Ehrlich. After
that it became a common practice to standardise the amount of dose (substance)
based on the responses of the living organism. This practice has become useful in
many branches of science like pharmacology, biochemistry, plant pathology, etc.
5.2.1 Principle of Bioassay
The basic principle of bioassay is to ascertain the potency of a biological
substance which can be any plant molecule/extract, drug, chemical, etc. in which
we compare the test substance with the standard preparation whose potency is
known and validated. It gives a fair approximation how much test substance is
required to produce the same biological effect, as produced by the standard
preparation. The bioassays are used to standardise chemical, drug, toxin, etc. The
standards are internationally accepted samples of drugs maintained and Paul Ehrlich (1854
recommended by the Expert Committee of the Biological Standardisation of –1915) was a
World Health Organization (WHO). They present the fixed units of activity German physician
(definite weight of preparation) for drugs. In India, standard drugs are maintained and scientist who
in Government institutions like Central Drugs Standard Control Organisation worked in the fields
(CDSCO), New Delhi; Central Drug Research Institute (CDRI), Lucknow; of haematology,
Central Drugs Testing Laboratories (CDTL) in Kolkata, Chennai, Tamil Nadu, immunology, and
Hyderabad, Mumbai, Kasauli; Regional Drugs Testing Laboratory (RDTL) in antimicrobial
Guwahati, Chandigarh, etc. chemotherapy. He
received the Nobel
It is to be noted that the problem of biological variation must be taken into Prize in 1908.
consideration as far as possible. For that we should keep uniform experimental
condition and assure the reproducibility of the responses.
In most of the assays, the standard preparation is likely to be a sample of either
international standard or a more readily available working standard whose
potency relative to the international standard has previously been carefully
evaluated.
We assay any test preparation for its biological activity or stimulus, which has an
unknown potency by finding the mean response of a selected dose, and equating
this dose to that of the standard preparation to standardise or validate the same
mean response. We should always perform the experiment with several different
doses of one or both preparations in order to accomplish reproducibility and
reliability. 89
Bioassay
5.2.2 Components of Bioassay
Let us now outline the components that constitute bioassays. The classical
bioassay consists of the followings:
i) Stimulus or substance or dose which can be a drug, fungicide, pesticide,
concentration of pollutant, radiation, plant extract, etc.
ii) Subject, which can be an animal, a plant, bacterial culture, human, insect,
living tissue, etc.
iii) Response which is a change in some measurable characteristic or visual
quantification of the subject whose magnitude will depend upon stimulus
applied to the subject, for example, change in the colours, pH, density,
blood sugar, bone ash percentage, death of the subject, muscular
contraction, weight of whole subject or a particular organ, etc.
In a bioassay, we apply the stimulus (or dose) on a subject and observe
response of the subject due to the applied stimulus (Fig. 5.1). We repeat the
process until enough observations have been made with the desired level of
precision.
For example, in a bioassay, insulin may be applied to mice and its body
response in term of the decreased blood sugar level can be measured. The
magnitude of this change in the blood sugar (decreasing blood glucose level) is
the response of the subject which depends upon the dose of insulin. We can
vary the concentration of the stimulus by varying the amount of the dose given
to the subject and this dose can be measured.
Usually, in bioassay, two preparations of stimulus or two different stimuli
The potency is the (both with quantitative doses): one of known strength which is called standard
amount or measure of preparation and another of unknown strength which is called test preparation,
the response due to are applied to subjects or a set of living organisms. The effect of test drug is
the standard correlated with the effect of standard drug. At the same time, a control group is
preparation also monitored / tested in which no test / standard drug is given.
equivalent in effect to
one unit of the test We can consider the bioassay as a planned experiment in which we apply two
preparation. stimuli: one with standard preparation (known) and another with test
preparation (unknown) to the subjects. We compare these two sets of doses
(standard and test preparations) in order to produce the similar desired
response. The measurement of potency makes it possible to compare the
standard preparations of various drugs with other test preparations on a fixed
scale under the standard sets of conditions.
While carrying out experiments / trials, the amount of the stimulus can be
varied in accordance with the objective and plan of the experiment. We
measure the response or reaction of the subject after application of the stimulus
in terms of change in some characteristics of the subject, for example, blood
pressure, blood sugar, body weight, kidney weight, occurrence of death,
recovery from a particular disease, etc. We measure the amount or dose of the
stimulus given to the subject in terms of the weight, volume or concentration.
Generally, the amount of applied dose is measured in milligram, microgram,
gram per kilogram of body weight, etc. The magnitude or the frequency of the
response depends upon the dose (quantity administrated). We access the
potency of a dose on the basis of the relationship between dose and response
induced by the given stimulus.
90
 Introduction to Bioassay

Stimulus

Subject

Response

Fig 5.1: Bioassay.

Form a statistician point of view, bioassay can be defined as a planned

biological experiment for estimating the potencies of one or more test stimuli
corresponding to a standard stimulus using the information provided by
responses measured on the subjects. We can say that the bioassays are also
relevant to compare the strength of alternative stimuli applied to the subjects
which are followed by the responses.
5.2.3 Role of Statistics in Bioassay
Statistics plays an important role in formulating experimental designs using the
existing information which provides most valuable and reliable results. We
most often consider the advice on experimental designs as an important
function of the statistician. Needless to emphasise that a sound design is as
important as the statistical analysis of data.
We can also use various statistical principles and analyse the experimental data
making best use of all evidences in bioassays. The validity of an assay is
checked by the means of various significant statistical tests. An important
aspect of statistics is to help in planning of an experimental programme so that
the results obtained by taking into accounts all relevant information will have
maximum utility.
You may like to pause here and check your understanding about the bioassay,
its principle and components by answering the following exercises.
E1) Explain the following statement with example:
“Bioassay is a form of biological experiment but all biological
experiments need not be bioassays”.
E2) Choose the correct option from the following:
i) Application of statistics includes
a) Randomisation
b) Formulate a good experimental design
c) Analysis of the experimental data
d) All of the above
ii) Bioassay can be used for
a) Standardising the drugs
91
Bioassay b) Estimating the potency of test drug
c) Both (a) and (b)
d) Neither (a) nor (b)

5.3 TYPES OF BIOASSAYS

So far you have studied the definition, principle and components of bioassays.
We now introduce the various types of bioassays. We can broadly classify the
bioassays into two categories.
1) Qualitative Assay
We use qualitative bioassay to assess the response or effect of the
stimulus (substance) which cannot be quantified or measured. However, a
visual correlation can be made and response can be graded. We are not
going to discuss these types of bioassays here in this course.
2) Quantitative Assay
In quantitative bioassay, we can measure the response of the stimulus
(substance) produced by the subject. These types of bioassays are often
involved in estimation of the concentration or potency of a substance and
can be analysed using various statistical methods. We will emphasise and
discuss these types of bioassays in this block.
The standard
These quantitative bioassays are further classified into two different
preparation is the
preparation with
categories:
known strength while i) Direct Assay
the test preparation is
the preparation with In direct assay, the response is considered as fixed and the dose as
unknown strength. random. So, we can directly measure the concentration of the standard
and test preparations which produces a specified response. The response
measured in direct assay is binary, e.g., the occurrence of an event or
none.
For example, if we specify the death of a subject as a desired response,
the dose required can be measured which will be responsible for the
death of a subject.
We will discuss direct bioassays in detail in Sec.5.4.
ii) Indirect Assay
In such types of bioassays, we first determine the relationship between
the dose and response corresponding to standard and test preparations.
Then the dose corresponding to a given response is obtained from the
dose-response relationship for each preparation separately.
These assays are very
similar in the form of In this case, the dose is considered as fixed for each subject in advance
required statistical while the response of each subject is random. Note that the responses can
analysis using dose- be quantal (binary) or quantitative (continuous). We will elaborate on the
response regression indirect bioassays in Unit 6.
relationships.
The indirect bioassay can be divided into two parts based on the measurement
scale of the responses, i.e., quantitative and quantal responses. When the
responses are measured on a continuous scale, e.g., growth of bacteria,
measurement of blood sugar level, these are considered as an indirect bioassay
92
based on quantitative responses. We generally determine the ratio of two Introduction to Bioassay
equally powerful or efficient doses from the dose-response curve which relates
quantitative responses and doses of two different preparations. The continuous
indirect bioassays based on the shape of the dose-response curve can be
divided into two parts (i) parallel-line assay and (ii) slop-ratio assay.
Let us discuss these categories of indirect bioassay in some extent.
a) Parallel-line Assay
In parallel-line assay, the response is linearly related to the logarithm of
the dose and we obtain two parallel straight lines with common slope but
different intercepts. We will discuss parallel-line bioassays in Unit 6 in
detail.
b) Slope-ratio Assay
In this assay, the response is linearly related to the power of the dose and
we obtain two straight lines with a common intercept but different slopes.
We will elaborate on slope-ratio bioassays in Unit 7.
c) Quantal Assays
When the responses of subjects are binary which show only occurrence/
non-occurrence of an event, e.g., death of a subject, positive/negative and
presence/absence of a particular symptom of a disease, these kinds of
bioassays are considered as indirect bioassays based on quantal
responses or simply, quantal bioassays. In this case, the relationship
between response and dose is assessed using logit or probit
transformation. We will explain the quantal bioassays in Unit 8.

Graphically, different types of bioassays are shown in Fig. 5.2.

Qualitative Assay
Bioassay

Direct Assay Parallel-line Assay

Quantitative Assay
Indirect Assay Slope-ratio Assay

Quantal Assay

Fig 5.2: Types of Bioassays.

We now describe the dilution assays.

Dilution Assays
A bioassay may be generally labelled as a dilution assay in which one or more
preparations of the stimulus or drug at different dose levels are prepared by
dilution. These preparations are directed into the experimental subjects and the
measurable biological responses are recorded. The dilution assays can be direct
or indirect.
93
Bioassay In a direct dilution assay, the amount of dose needed to produce a specific or
fixed response is measured. The dose is a random (stochastic) variable in direct
dilution assay.
On the other hand, various doses (at fixed dose levels) are administered in
Dilution is making a an indirect dilution assay, so the response is a random (stochastic) variable in
liquid sample/solution this case. In many assays, the test preparations can be considered as a dilution
weaker by adding a or concentration of the standard preparations.
solvent. The drugs’ On the basis of chemical components of the doses, the indirect bioassays can also
effective concentration be divided into two categories: (i) analytical dilution and (ii) comparative
is decreased by
dilution.
dilution.
i) Analytical Dilution Assay
In this kind of assay, the test and standard preparations are treated as
identical with same components (or constituents), apart from concentration
or dilution. So different concentration or dilutions are considered in these
assays.
In other words, we can say that an analytical dilution assay is one in which
we analyse for the effective dilution of a test preparation against a standard
preparation which has all common components of the test preparation.
Thus, in analytical dilution assay, the only difference is dilution level of the
preparations, so the relative potency is equal to the reciprocal of the
dilution factor and is a constant.
ii) Comparative dilution assay
In this case, the test and standard preparations do not have same
components, e.g., we estimate the concentration of two different drugs,
let’s say Drug A (test) with Drug B (standard) for a given dose. In
comparative dilution assay, the components of two preparations may only
be qualitatively similar and so the value of the relative potency may not be
constant or same.
In other words, we can say that two preparations may look alike
qualitatively in a comparative dilution assay, even though they are not the
same. For example, two pesticides of related but different chemical
compositions in terms of their effects on a particular pest.
You may like to pause now and check your understanding of various types
of bioassays discussed in this section. Answer the following exercises.
E3) Differentiate between qualitative and quantitative bioassays with
examples.
E4) Explain the dilution assays with example.

So far, we have given you an overview of the various types of bioassays. We

now explain the direct bioassay in detail.

5.4 DIRECT BIOASSAY

As discussed earlier in Sec. 5.3, a direct bioassay directly measures the
doses of the standard and test preparations which are sufficient to produce
a specified response. In other words, we can say that direct bioassays are
those assays in which the dose required to produce a pre-assigned or
specified response is directly measurable for both preparations. Here, the
94
response is certain (fixed) and the dose is a non-negative random Introduction to Bioassay
(stochastic) variable.
The direct assays are not feasible for all stimuli and subjects, so it has limited
applications. These types of assays can be used practically only when it is
possible to administer the dose in such a manner that the minimal (optimum)
amount of dose required to produce a specified response can be measured
directly.
For example, if we administer the standard or the test preparation at a fixed rate
into the blood stream of a cat until its heart stops beating. The dose can be
measured by multiplying the total time of infusion and the rate which is given
as:
Dose = Rate × Time … (1)
We repeat this experiment on several cats for each standard and test
preparations and compare the administered mean doses.
Thus, in such types of assays, we assume the following conditions to be
satisfied:
• The response should be clearly defined and easily recognised.
• The exact amount of dose needed to produce the desired response, can be
measured without time lag or any other difficulty.
It is important to note that the amount of critical dose may generally vary from
one occasion to another on the same subject. Hence, we obtain only estimator
of the potency and the average doses over a number of trials which will be
required to estimate the relative potency.
The following example will make your idea clear about the direct assay.
Example 1: Suppose two preparations, say A and B of a lethal drug are
infused at a fixed rate, into the blood stream of guinea pigs until their hearts
stop beating (causing death). In this example, the dose can immediately be
measured. This experiment is repeated on several guinea pigs for both
preparations. We obtain the average doses of both preparations and then
compare them. Suppose that Preparations A and B are considered as standard
and test preparations, respectively.
Table 1 shows the doses or tolerances of two groups of guinea pigs for both
preparations of the lethal drugs. The doses were recorded as per the amount
needed per kilogram of body weight of guinea pigs for stopping their hearts.
Table 1: Dose tolerances of guinea pigs for two preparations of the drugs

Preparation A Preparation B
(in mg/kg) (in mg/kg)
2.18 1.40
1.67 1.42
1.80 1.54
Doses 2.04 1.30
1.53 1.12
1.32 1.70
1.98 2.11
1.76 2.16
1.94 1.71
95
Bioassay In the next section, you will learn how to estimate the relative potency of the
test preparation, its variance and fiducial limits.

5.5 RELATIVE POTENCY

We define the relative potency of the test preparation which is relative to the
standard preparation as an amount of the standard equivalent in extent to one
unit of the test preparation. Hence, we can say that the relative potency is
estimated by the ratio of these two critical doses corresponding to standard and
test preparations.
As discussed, the relative potency of the test preparation with respect to the
standard is the amount of the standard preparation corresponding to one unit of
the test preparation. If Ds and Dt are the parameter doses of standard and test
preparations, respectively, producing the same response, then the actual
relative potency ρ is defined as:

Ds
ρ= … (2)
Dt

Let d si and d ti are the ith doses of standard and test preparations, respectively. If
we obtain the total K and L observations of the doses of standard and test
preparations, respectively, the following table shows the doses of both
preparations:
Table 2: Doses of standard and test preparations
Standard Test
d s1 d t1
d s2 d t1
Doses . .
. .
. .
d sK d tL
K L
Total d
i =1
si d
i =1
ti

Mean ds dt

If ds and dt be the mean doses of standard and test preparations, respectively,

which produce the same effect such that each of the doses produces a pre-
assigned response on a subject, then the relative potency (R) of the test
preparation relative to the standard preparation is estimated as:
ds
R= … (3)
dt
If R  1 , we can say that the potency of the test preparation is greater than that
of the standard preparation, i.e., a smaller mean dose of the test preparation
produces as much response as produced by relatively larger mean dose of the
standard preparation. Similarly, when R  1, the potency of the standard
preparation is greater than that of test preparation.
The variance of the estimated relative potency (R) can be determined using the
standard result of the variance of the ratio of two independent quantities a and
96 b, i.e.,
 a 1  
2 Introduction to Bioassay
a
var   = 2  var (a) +   var ( b ) 
 b  b  b 
In the same way, the variance of R is defined as:
1
Var(R) =  Var(ds ) + R 2 Var(d t ) 
2 
… (4)
dt
It is usually assumed that the variances of the doses of both preparations are
approximately equal, hence we can use estimated pooled variance for
estimating the variances of both preparations separately. The estimated pooled
variance is given as:
(K − 1)Var(d s ) + (L − 1)Var(d t )
S2 = … (5)
K+L−2
1 K 2 
1 K
 ( s) 
2
where V ar(d s ) = d − d =  d si − Kds2  … (6)
K − 1 i =1 K − 1  i =1
si

1 L 2 
1 L
 ( ) 
2
and V ar(d t ) = d − d =  d ti − L d t2  … (7)
L − 1 i =1 L − 1  i =1
ti t

If we consider pooled variance to define the variances of mean doses of the
standard and test preparations, we get
S2 S2
Sss = Var ( ds ) = and Stt = Var(d t ) = … (8)
K L
The variance of R defined in equation (4) can be rewritten as:

1 S2  1 R 2 
Var ( R ) =  + 2
 = + … (9)
dt 2  K L 
S R S
dt 2 
ss tt 

Let R L and R U be the lower and upper fiducial limits of R, respectively, which
are defined as:

R L = R − Var ( R ) …(10a)

R U = R + Var ( R ) …(10b)

We can also calculate the (1 − α ) % fiducial limits based on the t-distribution as:

R L = R − t (K + L −2),α/2 Var(R) …(11a)

R U = R + t (K + L −2),α/2 Var(R)
…(11b)
where t (K +L−2),α/2 is a two-tailed t-variate based on ( K + L − 2 ) degrees of
freedom and α % level of significance.
Let us now take up an example to illustrate this method.
Example 2: Let us consider the data given in Example 1 and
(i) Estimate the relative potency of the test preparation.
(ii) Obtain the variance of the relative potency calculated in (i). 97
Bioassay (iii) Construct the 95% fiducial limits for the relative potency.
Solution: (i) For computing the relative potency, we compute dsi2 and d 2ti as
shown in Table 3.
Table 3: Dose tolerances of guinea pigs for two preparations of the drugs
Preparation A Preparation B
d si2 dti2
( dsi ) ( d ti )
2.18 1.40 4.7524 1.9600
1.67 1.42 2.7889 2.0164
1.80 1.54 3.2400 2.3716
Doses 2.04 1.30 4.1616 1.6900
1.53 1.12 2.3409 1.2544
1.32 1.70 1.7424 2.8900
1.98 2.11 3.9204 4.4521
1.76 2.16 3.0976 4.6656
1.94 1.71 3.7636 2.9241
Total 16.22 14.46 29.8078 24.2242

From Table 3, we obtain the average doses of both preparations as:

16.22 14.46
ds = = 1.80222 and d t = = 1.60667
9 9

We can say that 1.802 mg of Preparation A and 1.607 mg of Preparation

B on an average are required, to kill a guinea pig.
Hence, we can estimate the relative potency as:

ds 1.80222
R= = = 1.12172
d t 1.60667

We can conclude that 1 mg of Preparation B is estimated to be equivalent

to 1.12 mg of Preparation A. This result holds for the sample considered
here, however such statements are subjected to some statistical errors.
(ii) We will now calculate the variance and fiducial limits of the relative
potency computed in (i).
For Preparation A

1 K 2 
Var(d s ) =  
K − 1  i =1
d si − K ds2 


1
=  29.8078 − 9  (1.80222) 2  = 0.07197
9 −1

Similarly, for Preparation B

1 L 2 
Var(d t ) =  
L − 1  i =1
d ti − Ld t2 

1
=  24.2242 − 9(1.60667) 2  = 0.12398
9 −1

98
Thus, the pooled variance is calculated as:
 (9 − 1)  0.07197 + ( 9 − 1)  0.12398 Introduction to Bioassay
S2 = = 0.09797
9+9−2
From equation (9), we compute the variance of the estimated relative
potency as:

S2  1 R2  0.09797  1 (1.12172 )2 
Var(R) = 2  K + L  = (1.60667) 2  +  = 0.00952
dt    9 9 
We determine the fiducial limits using equation (10) as:
R L = 1.12172 − 0.00952 = 1.02413

R U = 1.12172 + 0.00952 = 1.21930

(iii) Since α = 0.05 and α / 2 = 0.025 , the tabulated t value at 5% level of
significance with (9+9−2=16) degrees of freedom will be t16,0.025 = 2.12 .

95% fiducial limits (equation (11)) based on t-variate can be calculated

as:

R L = 1.12172 − 2.12  0.00952 = 0.91484

R U = 1.12172 + 2.12  0.00952 = 1.32859

Now, we can say that 1 mg of Preparation B may be asserted to be not
less potent than approximately 0.915 mg and not more potent than 1.329
mg of Preparation A.
You can now check your understanding of the computation of relative potency,
its variance and fiducial limits by answering the following exercises.

E5) Let ds and dt denote the average doses of the standard and test
preparations, respectively, then relative potency R > 1 indicates that
a) Test preparation is less effective than standard preparation.
b) Test preparation is more effective than standard preparation.
c) Both (i) and (ii).
d) Neither (i) nor (ii).
E6) The doses of three preparations, say, A, B and C are recorded which
cause death of the subjects as a result of a direct assay. The recorded doses are
given in the following table:
Preparation A Preparation B Preparation C
(in ml) (in ml) (in ml)
18.90 16.20 30.60
23.40 11.70 32.22
14.40 12.15 34.65
17.10 10.35 33.30
19.80 13.50 30.60

If we assume Preparation A as a standard preparation and B and C as

test preparations, then
i) Estimate the relative potency of B and C.
99
Bioassay ii) Also find the 95% fiducial limits of the relative potencies of
Preparations B and C.
Having explained the relative potency and its fiducial limits, we now discuss
another method to determine the fiducial limits of the relative potency using a
theorem, called Fieller’s theorem.

5.6 FIELLER’S THEOREM

The theorem for obtaining the fiducial limits of the ratio of two mean doses
was first given in its general form by Fieller (1940), though the particular case
of zero covariance which had also been used earlier given by BLISS (1935).
Let us consider standard and test parameter doses Ds and Dt , respectively, and
the relative potency ( ρ ) which is the ratio of these two parameters is defined
as:
Ds
ρ= … (12)
Dt

We assume that ds and dt (mean sample doses of both preparations) are linear
functions of the observations and are normally distributed. Let ds and dt be the
unbiased estimators of Ds and Dt , i.e.,

E ( ds ) = Ds and E ( d t ) = D t … (13)

As discussed in Sec. (5.5), the relative potency R is an estimate of ρ and

defined as:

ds
R= … (14)
dt

According to Fieller’s theorem, the lower and upper fiducial limits of relative
potency are given as:

 Sst  t  S2 
R −g  Sss − 2RSst + R 2Stt − g  Sss − st 
 Stt  dt  Stt 
(R L , R U ) =
1− g

t 2Stt
where g = , Sss = Var(ds ) , Stt = Var(dt ) , Sst = Cov(ds , dt ) and
d t2
t t ( K + L–2), α/ 2 .

Proof: Let us consider a linear combination ( ds − ρ d t ) which is normally

distributed with mean:
E ( ds − ρ d t ) = Ds − ρDT = 0 … (15)

and estimated variance:

V ( ds − ρ d t ) = Sss − 2ρSst + ρ 2Stt … (16)

For determining (1−α)100% fiducial limits, we consider the probability given

100 as:
 Prob ( ds − ρdt )  t 2 Var ( ds − ρdt ) = 1 − α
2 Introduction to Bioassay
 

 Prob ( ds − ρdt )  t 2 (Sss − 2ρSst + ρ 2Stt ) = 1 − α

2

  ...
(17)
We can expand the expression written inside the square bracket as an
inequality for a quadratic expression given in terms of ρ as:

(d − ρd t )  t 2 (Sss − 2ρSst + ρ 2Stt )

2
s

ds2 − 2ρds d t + ρ 2 d t2 − t 2 (Sss − 2ρSst + ρ 2Stt )  0

i.e., ρ 2 ( d t2 − t 2Stt ) + 2ρ ( t 2Sst − ds d t ) + ( ds2 − t 2Sss )  0 … (18)

Suppose a quadratic equation in x is given as:

ax 2 + bx + c = 0

The solution of the quadratic equation will be

−b b 2 − 4ac
( xL , xU ) = … (19)
2a

We can obtain two estimated values of ρ after solving the quadratic equation
(18) same as defined in equation (19) as:

 
 −2 ( t Sst − ds d t ) 4 ( t 2Sst − ds d t ) − 4 ( d t2 − t 2Stt )( ds2 − t 2Sss ) 
2 2

(R L , R U ) = 
 2 ( d t2 − t 2Stt ) 

 

 ( t 4Sst2 + ds2 dt2 − 2t 2Sst ds dt ) 

 ( st s t)
 −2 t 2S − d d 2 
− ( d t ds − t Sss d t − t Stt ds + t SssStt ) 
2 2 2 2 2 2 4

=  
 2 ( d t2 − t 2Stt ) 
 
 
 

 ( d d − t 2S ) t t 2S2 − 2S d d + S d 2 + S d 2 − t 2S S 
= 
s t st st st s t ss t tt s ss tt
… (20)
 d t − t Stt
2 2

 
After dividing both numerator and denominator of equation (20) by d t2 , we
obtain:

  d t 2S  t t 2Sst2 2Sst ds ds2 t 2SssStt 

  s − 2st  − + Sss + Stt − 
d t2 d t2 d t2
=   t 
d dt  dt dt
t 2S 
 1 − 2tt 
 dt 
  101
Bioassay
  d t 2S  t 2S d d 2 t 2S S t 2S2 
  s − 2st  Sss − st s + Stt s2 − ss2 tt + 2st 
=   t
d dt  dt dt dt dt dt 
t 2S 
 1 − 2tt 
 dt 
 
  d t 2S  t 2S d d 2 t 2S  S2  
  s − 2st  Sss − st s + Stt s2 − 2tt  Sss − st  
  dt dt  dt dt dt dt  Stt  
=  … (21)
 t 2Stt 
1− 2
 dt 
 
2
t Stt d
After Substituting 2
= g and s = R in equation (21), we obtain
dt dt
 S  t  Sst2  
  R − g st  Sss − 2RSst + R Stt − g  Sss −  
2

 Stt  d t  Stt  
=  … (22)
 1 − g 
 
 
Hence (1−α)100% lower and upper fiducial limits of the relative potency can be
obtained as:
 Sst  t  Sst2 
R −g − Sss − 2RSst + R Stt − g  Sss − 
2

 Stt  d t  Stt 
RL = …(23a)
1− g
 Sst  t  Sst2 
R −g + − + − −
2
 Sss 2RS st R S tt g  ss
S 
 Stt  dt  Stt 
RU = …(23b)
1− g
t 2Stt
where g = … (24)
d t2
If both preparations are independent, i.e., Sst = 0 , the fiducial limits will be
t
R− Sss + R 2Stt − gSss
dt
RL = …(25a)
1− g
t
R+ Sss + R 2Stt − gSss
dt
RU = …(25b)
1− g
If the value of g is neglected, equation (23) becomes

t
RL = R − Sss − 2RSst + R 2Stt …(26a)
dt
t
RU = R + Sss − 2RSst + R 2Stt …(26b)
dt

The result obtained in equation (26) will be same as given in equation (11)
which is considered as a particular case if Sst = 0 , i.e., when both doses are
independent.

102 In this case, the approximate formula for the variance of R is given as:
 1 Introduction to Bioassay
Var(R) = Sss − 2RSst + R 2Stt  … (27)
dt2 

Note that Fieller’s theorem is not applicable in all situations. Complication

arises when more than one error mean square occurs. Similar formula based on
the Behrens-Fisher distribution instead of t-distribution can be used for data
involving two independent mean squares. Here, we are not going to discuss
Behrens-Fisher distribution as it is beyond the scope of this unit.
If two stimuli of unequal potency are applied at equal rates (e.g., two drugs
infused at equal speeds), the subjects receiving the less potent stimulus will
have longer average times under treatment than those receiving more potent
stimulus. If there is any time-lag in the responses or effects on the subjects or
any cumulative effect of simply additive nature, the comparison of two drugs
will be biased. This difficulty might be largely overcome by applying
equipotent doses of each stimulus per unit of time. But presupposes knowledge
of relative potency to say how much (even a quite small) deviation from the
ideal dose will affect the validity of the potency may be impossible.
In many situations, there may be a time-lag between the administration of the
doses and the appearance of the responses. In that case, the direct assays will
not provide the appropriate results. We may explore other alternative assays
like indirect assays in such situations.
Let us solve an example to get a better idea about computation of the fiducial
limits of the relative potency using Fieller’s theorem.
Example 3: Let us consider the data given in Example 2 and calculate the 95%
fiducial limits of the relative potency using Fieller’s theorem.
Solution: From the solution of Example 2, we have

ds = 1.80222 and ds = 1.60667

We obtain the values of Sss and Stt using equation (8) as:

S2 0.09797
Sss = Stt = = = 0.01089
9 9
and Sst = 0 (Since both preparations are independent)

Since α = 0.05 and α / 2 = 0.025 , the tabulated t value at 5% level of

significance with (9+9−2=16) degrees of freedom will be t16,0.025 = 2.12 .

We calculate the value of g using equation (24) as:

( 2.12 )  0.01089 = 0.01895

2
t 2S
g = 2tt =
(1.60667 )
2
dt

From equation (25), 95% lower and upper fiducial limits of ρ can be computed
as:
2.12
1.12172 − 0.01089 + 1.121722  0.01089 − 0.01895  0.01089
RL = 1.60667
1 − 0.01895
103
Bioassay 2.12
1.12172 −  0.15613
= 1.60667 = 0.93340
1 − 0.01895
2.12
1.12172 + 0.01089 + 1.121722  0.01089 − 0.01895  0.01089
RU = 1.60667
1 − 0.01895

2.12
1.12172 +  0.15613
= 1.60667 = 1.35337
1 − 0.01895
Hence, 95% fiducial limits of ρ are (0.93340, 1.35337)
We can say that 1 mg of Preparation B may be asserted to obtain a potency
lying between approximately 0.93 mg and 1.35 mg of Preparation A.
Note that the computed value of g = 0.1895 is small enough to be practically
negligible, so the theorem gives the limits which are almost same as the
approximate values computed in Example 2.
Before ending this unit, you may like to solve the following exercises for
practice.

E7) Fieller’s theorem provides the fiducial limits of the

i) Strength of test preparation
ii) Strength of standard preparation
iii) Relative potency
iv) An estimate of relative potency.
E8) For the data given in E6, determine the 95% fiducial limits of the
relative potency of Preparation B relative to Preparation A using
Fieller’s theorem and also compare the result with the result obtained in
E6.
We end this unit by giving a summary of its contents. In the next unit, we shall
discuss about the indirect bioassays.

5.7 SUMMARY
1) In a bioassay, a stimulus is applied on a subject and a response is
observed. This process is repeated until enough observations have been
made with the desired level of precision.
2) The role of statistician is to validate and provide statistical support by
means of various statistical tools of experimental design as well as
statistical analysis.
3) In direct assays, a specified response can be measured directly for the
doses of standard and test preparations.

4) If d s and d t are mean doses for the standard and test preparations, then
the relative potency (R) of test preparation corresponding to standard
preparations is defined as:
104
 Introduction to Bioassay
ds
R=
dt

5) The variance of R is computed as:

S2  1 R2 
Var ( R ) = K + L 
dt 2  

6) The simple fiducial limits of relative potency are defined as:

R L = R − Var ( R )

R U = R + Var ( R )

7) The (1 − α ) % fiducial limits based on the t-distribution are determined as:

R L = R − t (K + L −2),α/2 Var(R)

R U = R + t (K + L −2),α/2 Var(R)

where t (K + L−2),α/2 is a two-tailed t-variate based on ( K + L − 2 ) degrees

of freedom and α % level of significance.

8) The (1– )100% lower and upper fiducial limits of the relative potency
using Fieller’s theorem are obtained as:

 Sst  t  S2 
R −g  Sss − 2RSst + R 2Stt − g  Sss − st 
 Stt  dt  Stt 
(R L , R U ) =
1− g

t 2Stt
where g = , Sss = Var(ds ) , Stt = Var(dt ) , Sst = Cov(ds , dt ) and
d t2
t t (K + L−2), α/2 .

5.8 SOLUTIONS / ANSWERS

E1) Refer Sec. 5.2.
E2) i) Option (d) is the correct option because we know that the
application of Statistics includes randomisation, as well as
formulate experimental design and analysis of data.
i) Option (c) is the correct option because the bioassay can be
utilised to standardise the drugs as well as estimate potency of the
test drug.
E3) Refer Sec. 5.3.
E4) Refer Sec. 5.3.
E5) Option (b) is correct because R > 1 means smaller dose of the test
preparation will be able to produce same response as the larger dose of
the standard preparation. 105
Bioassay E6) i) To determine the relative potencies and its fiducial limits, we construct
the following table:
Preparation Preparation B Preparation 2
A ( d Ai )
d 2Ai 2
( d Bi ) C ( d Ci )
d Bi dCi

18.90 16.20 30.60 306.1800 495.7200 9369.1080

Doses 23.40 11.70 32.22 273.7800 376.9740 8821.1916
14.40 12.15 34.65 174.9600 420.9975 6062.3640
17.10 10.35 33.30 176.9850 344.6550 5893.6005
19.80 13.50 31.50 267.3000 425.2500 8419.9500
Total 93.60 63.90 162.27 5981.0400 10369.0530 970543.3608

From the above table, we obtain the average doses of Preparations A,

B and C as:
93.60 63.90 162.27
dA = = 18.72 , dB = = 12.78 and dC = = 32.454
5 5 5
We compute the relative potency of Preparation B with respect to
Preparation A as:
dA 18.72
RB = = = 1.46479
dB 12.78
The relative potency of Preparation C with respect to Preparation A is
given as:
dA 18.72
RC = = = 0.57682
dC 32.454
We can conclude that 1 mg of Preparation B is equivalent to
approximately 1.47 ml and 1 mg of Preparation C is alike to 0.58 ml
of Preparation A.
ii) We now determine the variances of the given values of all preparations.
For Preparation A
1  5 2 
VA =  
5 − 1  i =1
d Ai − 5 d A2 


1
= 5981.0400 − 5  (18.72) 2  = 11.09700
5 −1

For Preparation B
1  5 2 
VB =  
5 − 1  i =1
d Bi − 5 dB2 

1
= 10369.0530 − 5  (12.78) 2  = 4.92075
5 −1
Similarly, for Preparation C
1  5 2  1
VC =  
5 − 1  i =1
d Ci − 5 dC2  =
 5 − 1
970543.3608 − 5(32.454) 2 

106 = 2.48508
 The pooled variance of Preparations A and B is calculated as: Introduction to Bioassay

(5 − 1) 11.09700 + ( 5 − 1)  4.92075
S2AB = = 8.00888
5+5−2
The pooled variance of Preparations A and C is obtained as:
(5 − 1) 11.09700 + ( 5 − 1)  2.48508
S2AC = = 6.79104
5+5−2
From equation (9), we compute the variances of the estimated relative
potencies of Preparations B and C as:

8.00888  1 (1.46479 ) 
2

Var(R B ) =  +  = 0.03085
(12.78) 2  5 5 

8.00888  1 ( 0.57682 )2 
Var(R C ) =  +  = 0.00172
(32.454) 2  5 5 
We determine the fiducial limits of relative potencies of Preparations B
and C as:
R BL = 1.46479 − 0.03085 = 1.28915

and R BU = 1.46479 + 0.03085 = 1.64043

R CL = 0.57682 − 0.00172 = 0.53536 and

and R CU = 0.57682 + 0.00172 = 0.61827

Since α = 0.05 and α / 2 = 0.025 , the tabulated t value at 5% level of
significance with (5+5−2=8) degrees of freedom will be t 8,0.025 = 2.31 .

95% fiducial limits of relative potencies based on t-variate can be

calculated as:
For Preparation B
R BL = 1.46479 − 2.31 0.03085 = 1.05906

R BU = 1.46479 + 2.31 0.03085 = 1.87052

For Preparation C
R CL = 0.57682 − 2.31 0.00172 = 0.48105

R CU = 0.57682 + 2.31 0.00172 = 0.67258

E7) Option (c) is correct because we determine fiducial limits for the
relative potency ( ρ ) using Filler’s theorem.
E8) From the solution of E6, we have
ds = dA = 18.72 and dt = dB = 12.78
From equation (8), we obtain
S2AB 8.00888
Sss = Stt = = = 1.60178
5 5 107
Bioassay
and Sst = 0 (Since both preparations are independent)
Since α = 0.05 and α / 2 = 0.025 , the tabulated t value at 5% level of
significance with 8 degrees of freedom will be t 8,0.025 = 2.31 .

The value of g is determined using equation (24) as:

( 2.31)  1.60178 = 0.00020

2
t 2Stt
g= 2 =
(12.78)
2
dt

95% lower and upper fiducial limits of relative potency of Preparation B

using equation (25) can be computed as:

2.32 1.60178 + (1.46479) 1.60178 − 0.00020

2
1.46479 −
12.78 1.60178
R BL =
1 − 0.00020
3.31
1.46479 −  0.13931
= 12.78 = 1.05929
1 − 0.00020

2.12 0.01089 + 1.121722  0.01089 − 0.01895

1.12172 +
1.60667 0.01089
R BU =
1 − 0.01895
2.12
1.12172 +  0.15613
= 1.60667 = 1.87088
1 − 0.01895
Hence, 95% fiducial limits of relative potency of Preparation B are
(1.05929, 1.87088).
It is to be noted the fiducial limits computed here using Fieller’s theorem
are the same as obtained based on the t-distribution in E6.

108