STATISTICS

Sample Size Estimation in Clinical Trial

Tushar Vijay Sakpal

Abstract

Every clinical trial should be planned. This plan should include the objective of trial, primary and secondary end-
point, method of collecting data, sample to be included, sample size with scientific justification, method of handling
data, statistical methods and assumptions. This plan is termed as clinical trial protocol. One of the key aspects of this
protocol is sample size estimation. The aim of this article is to discuss how important sample size estimation is for a
clinical trial, and also to understand the effects of sample size over- estimation or under-estimation on outcome of a
trial. Also an attempt is made to understand importance of minimum sample to detect a clinically important difference.
This article is also an attempt to provide inputs on different parameters that impact sample size and basic rules for
these parameters with the help of some simple examples.

Key Words: Sample size, statistically significant, clinically significant, Type I error, Type II error, power,

Introduction 1) The above p-value indicates that a smaller sample size

was needed to prove statistical significance. This raises
The article will cover ethical issues as more subjects were exposed to test drug,
1) Why sample size estimation is important in a clinical trial (CT) which could have deserved getting NEW drug.
2) Inputs and their importance for sample size estimation in CT 2) As sample selected was very large even a small difference
3) Examples on sample size estimation between NEW and test drug will turn out statistically
4) Discuss some basic rules and guidelines significant even if that difference is not clinically
meaningful.
Importance of Sample Size in CT Hence sample size is an important factor in approval /
rejection of CT results irrespective of how clinically effective
Let us see an example. Dr. ABC had developed a drug
/ ineffective, the test drug may be.
NEW which was effective in reducing pain. The drug NEW is
What does it take to estimate a statistically appropriate
clinically better in terms of efficacy and safety as compared
sample size?
to test drug. Hence, the drug NEW truly deserves to reach
market so that patients suffering from pain can be at ease. Basic Requirement for Estimation of Sample
Dr. ABC decided to conduct a CT in order to get approval
from regulators. How will the sample size affect Dr. ABC’s
Size in CT
trial on patients suffering from pain? The estimation of sample size along with other study
Scenario 1 – Sample size under-estimation (Sample size related parameters depends on Type I error, Type II error and
selected was less than what was required) Power.
At the end of trial when data was analyzed, statistical
significance was not achieved (statistical jargon, p-value was Type I Error:
0.05).
Underestimation of sample size may result in drug turning Consider we are testing two brands of paracetamol to
out to be statistically non-significant even though clinical evaluate if Brand 1 is better in curing subject’s suffering from
significance exists. fever as compared to Brand 2. As both brands contain
Scenario 2 – Sample size over-estimation (Sample size paracetamol, it is expected that the effect of both brands is
selected was much more than what was required) similar. Let us try to built a statistical hypothesis around this
Dr ABC conducted CT with a large number of subjects. Null Hypothesis (H0): Brand 1 is equal to Brand 2
When the data was analyzed, strong statistical significance Alternate Hypothesis (H1): Brand 1 is better than Brand 2
was achieved as p-value turned out to be 0.000001. Let us try to evaluate error that can occur.
However Dr ABC has to consider the following issues: Error 1: Based on analysis it is concluded that Brand 1 is
better than Brand 2, basically we reject H0. Knowing that
Tushar Vijay Sakpal
Brand 1 is equal to Brand 2 (H0), we are making an error here
Principal Statistician, Biostatistics, PharmaNet Clinical by rejecting H0. This is called as Type I error. Statistically it is
Services Pvt. Ltd. defined as

PICR April-June 2010 Vol 1 Issue 2 67

 Sample Size Estimation

Type I error = (Reject H0 / H0 is true). of a CT. The description of primary study end point should
Probability of Type I error is called as level of significance cover whether it is discrete or continuous or time-to-event.
and is denoted as . Sample size is estimated differently for each of these end
points. Sample size is adjusted if primary end point involves
Type II Error: multiple comparision.
Consider we are testing paracetamol against placebo to 4) Expected Response Test vs. Control
evaluate if paracetamol is better in curing subject’s suffering The information about expected response is usually
from fever as compared to placebo. (It is expected that the obtained from previous trials done on the test drug. If this
effect of paracetamol is better than placebo). Let us try to information is not available, it could be obtained from previous
built a statistical hypothesis around this published literature. The important information required is:
Null Hypothesis (H0): paracetamol is equal to placebo a) response expected with test drug (mean expected score
Alternate Hypothesis (H1): paracetamol is better than / proportion of subjects achieving success.)
placebo b) response expected with control drug (mean expected
Let us try to evaluate the error that can happen. score / proportion of subjects achieving success.)
Error 2: If analysis concludes that paracetamol is equal to c) variability (standard deviation.)
placebo, we accept H0. Knowing that paracetamol is better
than placebo (H1) we are making an error here by accepting 5) Clinical Important / Meaningful Difference / Margin
H0. This is called as Type II error. Statistically it is defined as This is one of most critical and one of most challenging
Type II error = (Accept H0 / H1 is true). parameters. The challenge here is to define a difference
Probability of type II error is denoted as . between test and reference which can be considered clinically
In above case if analysis concludes that paracetamol is meaningful. To put it in plain English, this is the difference
better than placebo, we reject H0, which would be correct which will make your family doctor to give you new medication
decision. Probability of such a decision taking place is called over and above the existing gold standard which he is
as “Power”. prescribing for last 10 – 20 years. This threshold figure is at
Power = Probability (Reject H0 / H1 is true) which is actually 1- . times not easily available and should be decided based on
We can tabulate type I and type II error as clinical judgment.
Decision Taken \ Actual Fact H0 is True H1 is True
Reject H0 Type I error No error 6) Level of Significance
Accept H0 No error Type II error This is typically assumed as 5% or lesser. 1, 2.Type I error is
inversely proportional to sample size.
Following are the basic considerations for estimating
sample size for a CT 7) Power
As per guideline.1, 2 power should not be less that 80%.
1) Study Design Type II error is directly proportional to sample size.
In CT, different statistical study designs are available to
achieve objectives. Typical designs that may be employed 8) Drop-out rate
are parallel group design, crossover design etc. The sample size estimation formula will provide number
For sample size estimation study design should be of evaluable subjects required for achieving desired statistical
explicitly defined in the objective of the trial. A cross over significance for a given hypothesis. However in practice we
design will have different approach and formula for sample may need to enroll more subjects to account for potential
size estimation as compared to parallel group design. dropouts.
If n is the sample size required as per formula and if d is
2) Hypothesis Testing the dropout rate then adjusted sample size N1 is obtained as
This is another critical parameter needed for sample size N1 = n/ (1-d).
estimation, which describes aim of a CT. The aim can be
equality, non-inferiority, superiority or equivalence. 9) Unequal Treatment Allocation
Equality and equivalence trials are two-sided trials where For some of clinical trials it is ethically desirable to have
as non-inferiority and superiority trials are one-sided trials. more subjects in one arm compared to the other arm. For
Superiority or non-inferiority trials can be conducted only example, for placebo-controlled trials with very ill subjects it
if there is prior information available about the test drug on a is unethical to assign equal subjects to each arm. In such
specific end point. cases sample size is adjusted as below.
If n is sample size required as per formula and n1 is the
3) Primary Study End Point sample size for test and n2 is sample size for placebo and
The sample size calculation depends on primary end point n1 / n2 = k, then
68 PICR April-June 2010 Vol 1 Issue 2
 Sample Size Estimation

n1 = (0.5) x n x (1 + k) in Active drug = 4.5,

n2 = (0.5) x n x (1 + (1/k)) μ1-μ2 = clinically significant difference = 0.5
Suppose n = 100 and ratio between test (n1) and placebo ó = standard deviation = 1.195
(n2) is 2:1 (k=2) then Z /2: This depends on level of significance, for 5% this is 1.96
n1 = (0.5) x (100) x (1+2) = 150 Z : This depends on power, for 80% this is 0.84
n2 = (0.5) x (100) x (1+[1/2]) = 75 Based on the above formula, the sample size required per
As a statistician one should have clear and clinically group is 90. Hence total sample size required is 180.
meaningful inputs on the above points prior to working on Considering a drop-out rate of 10% total sample size required
sample size estimation. is 200 (100 in each arm).
The description sample size in the protocol will be: A
Calculation of Sample Size sample size of 180 subjects, 90 in each arm, is sufficient to
detect a clinically important difference of 0.5 between groups
Example 1: Comparing two proportions
in reducing pain assuming a standard deviation of 1.195 using
A placebo-controlled randomized trial proposes to assess
a two-tailed t-test of difference between means with 80%
the effectiveness of Drug A in curing infants suffering from
power and a 5% level of significance. Considering a dropout
sepsis. A previous study showed that proportion of subjects
rate of 10% the sample size required is 200 (100 per group).
cured by Drug A is 50% and a clinically important difference
of 16% as compared to placebo is acceptable. Basic Rules for Estimating Sample Size in CT
Level of siginificance = 5%, Power = 80%, Type of test =
two-sided Some basic rules for on sample size estimations are:
Formula of calculating sample size is 1) Level of significance – It is typically taken as 5%. The
n = [(Z /2 + Z ) 2 x {(p1 (1-p1) + (p2 (1-p2))}] / (p1 – p2)2 sample size increases as level of significance decreases.
where 2) Power – It should be >= 80%. Sample size increases as
n = sample size required in each group, power increases. Higher the power, lower the chance of
p1 = proportion of subject cured by Drug A = 0.50, missing a real effect.
p2 = proportion of subject cured by Placebo = 0.34, 3) Clinically meaningful difference - To detect a smaller difference,
p1-p2 = clinically significant difference = 0.16 one needs a sample of large size and vice a versa.
Z /2: This depends on level of significance, for 5% this is 1.96 4) Sample size required to demonstrate equivalence is highest
Z : This depends on power, for 80% this is 0.84 and to demonstrate equality is lowest.
Based on above formula the sample size required per group The sample size estimation is challenging for complex
is 146. Hence total sample size required is 292. designs such as non-inferiority or, time to event end points.
The description sample size in the protocol will be: Also, the sample size estimation needs adjustment in
A sample size of 292 infants, 142 in each arm, is sufficient accommodating a) unplanned interim analysis b) planned
to detect a clinically important difference of 16% between interim analysis and c) adjustment for covariates.
groups in curing sepsis using a two-tailed z-test of In conclusion, the sample size is one of the critical steps
proportions between two groups with 80% power and a 5% in planning a CT and any negligence in its estimation may
level of significance. This 16% difference represents a 50% lead to rejection of an efficacious drug and an ineffective
cure rate using drug A and 34% cure rate using placebo. drug may get approval. As calculation of sample size depends
Example 2: Comparing two means on statistical concepts, it is desirable to consult an experienced
An active-controlled randomized trial proposes to assess statistician in estimation of this vital study parameter.
the effectiveness of Drug A in reducing pain. A previous
study showed that Drug A can reduce pain score by 5 points Acknowledgement:
from baseline to week 24 with a standard deviation (ó) of
I would like to thank Dr. Suresh Bowalekar, Managing
1.195. A clinically important difference of 0.5 as compared to
Director PharmaNet Clinical Services Pvt. Ltd., for his
active drug is considered to be acceptable. Consider a drop-
guidance and inputs on writing this article.
out rate of 10%.
Level of significance = 5%, Power = 80%, Type of test = References:
two-sided
Formula of calculating sample size is 1 Shein-Shung C., Jun S. and Hansheng W. Sample Size
n = [(Z /2 + Z ) 2 x {2(ó) 2}]/ (μ1 – μ2)2 Calculation in Clinical Trial. New York, Marcel Dekker Inc.,
where 2003, Chapter 1, page 11, section 1.2.3.
n = sample size required in each group, 2 ICH Topic E 9, Statistical Principal for Clinical Trials Step
μ1 = mean change in pain score from baseline to week 24 4, Consensus guideline, 05Feb1998 Note for Guidance
in Drug A = 5, On Statistical Principles For Clinical Trial. Available via
μ2 = mean change in pain score from baseline to week 24 http://www.ich.org.

PICR April-June 2010 Vol 1 Issue 2 69