Hyderabad Sample Size 1 27jan03
Hyderabad Sample Size 1 27jan03
Hyderabad Sample Size 1 27jan03
Dhulkhed Professor and Head Department of Anaesthesiology KIMS (Medical College) KARAD Maharashtra
Is the study powerfull enough so that people can really believe the result or apply the the whole population
sample result to
Small Sample: Unable to detect clinically important results. Not very small and not very large
Clinical Significance:
In 1966 Caluson et al had shown 25% reduction in mortality due to proparanolol among MI patients. But that was not significant. 66 cases and 64 controls were studied.
Is 25% important?
Innocent
ri inal
Null True
Innocent
orrect
Type II error (or) Beta error
Judges decision
Null-False
Null -True
Researchers decision
Correct
Null-False
------------------------------------------------Type II (Beta) Error ------------------------------------------------p ZF ------------------------------------------------0.10 1.282 0.15 1.037 0.20 0.842 0.25 0.675 -------------------------------------------------
Power
TRUTH Ho D Type I Error E Not Reject H o E - level C I Type I Error S I E level Reject Ho O N HA
Power Type II 1 - Error F F Type II Power F Error
Power
Power is a measure of doing the right thing when Ha is true! Higher power is better (the closer the power is to 1.0 or 100%) We can calculate power for a given study if we specify a specific Ha
Types of Errors:
After any decision, questions remain to be answered.
2. What is the probability, that if a true difference of a stated magnitude existed, but the study would not have picked it up as statistically significant?.
Po er of the Study:
Po er: Complement of Beta error. The probability that if a true difference of stated magnitude existed then the study would have picked it up as statistically significant.
Serious Error:
Type I: Reject the null hypothesis, in fact O DIFFERENCE. (False rejection) BCG is effective, in fact No Difference between two arms
Type II: Accept the null hypothesis when it is false. BCG is not effective, in fact there is DIFFERENCE between two arms.
Estimation
Descriptive study
Proportion or Prevalence is the parameter of our study
Where
Z pq n= 2 d
Z = Standardized Normal deviate (Z value) p = Proportion or Prevalence of interest. q = 100 - p d = Clinically expected variation.
Example
Previous post operative shivering incidence rate is 30% Treatment and other measures instituted You ould like to find out the present incidence rate ith allo able error 10% or 20%
Example
For 10% error P=30% q=100-30=70%
Allowable error d=10% 0f 30%=3% n > 4pq/d=4x3070/3 933 For 20% error P=30% q=100-30=70% Allowable error d=20% 0f 30%=6% n > 4pq/d=4x3070/6 234
S d 2
Where Z= Standardized Normal deviate (Z value) S = Sample standard deviation d = Clinically expected variation
Example
In a Health Survey of schoolchildren it is found that the mean haemoglobin level of 55 boys is 10.2/100ml with a standard deviation of 2.1. Consider the precision as 0.8.
Mean = 10.2 Standard = 2.1 ZE= 1.96 for E at 0.05 d = 0.8
2 2
= 26
Testing Hypothesis
Formulae & Problems
Analytical study
Proportion is the parameter of our study
n=
Where
2 (Z + Z ) p q d
2
ZE = Z value for E level ZF = Z value for F level p = average percentage between two groups q = 100 - p d = Clinically meaningful difference between two groups
Example (Qualitative)
A postgraduate student in Anaesthesia wants to take up the research project to answer the research question. Is antiemetic metoclopramide effective in reducing the incidence of vomiting after ether anesthesia ? Previous studies indicated that 60% of patients vomit after recovery from ether anaesthesia Reduction to 20% will make it clinically useful What is the sample size for a power of 0.90 % and 95% two sided significance level ?
Sample size
Literature search indicating p1 i.e. proportion of people ho vomit =
60%
20%
p=
n=
p 1 + 2
p 2
60 + 20 2
= 40% ; q = 1 p = 60%
2 ( d2
)2 p q
2 ( 1.96 1.282 ) 2 60 40 = ( 40 ) 2
Total approx = 70
Group B patients with 25 g Whitaker needle Objective to detect a statistically significant less level of
headache incidence among Group B. This level should be 4 times less i.e. OR = 0.25 Calculate the required sample size.
Solution
P1 = 2% or 0.02
p1 OR = {1 p1( OR 1 )}
OR = = 0.25
p =
p1 + p 2 2
= 0.0125
q = 0.9875
E = 0.05
n= 2 ( +
(ZE = 1.65 )
)2 p q d2
(Z
+Z d
) S 2
2
Research Question
Does the duration of analgesia in laboring patients after intrathecal bupivacaine is more compared with ropivacaine ? Design a study to to answer the question
State Null Hypothesis There is no difference in the duration analgesia after intrathecal administration of either Bupivacaine or ropivacaine
Review the literature to know the clinically important difference or effect size Anesth Analg. 2004 Jan;98(1):235-9 A comparison of duration of analgesia of intrathecal 2.5 mg of bupivacaine, ropivacainein combined spinal epidural analgesia for patients in labor. Lim Y, Ocampo CE, Sia AT. Department of Anesthesia, KK Women's and Children's Hospital, 100 Bukit Timah Road, Singapore 229899, Republic of Singapore.
Review the literature The duration of labor analgesia by intrathecal (IT) local anesthetics using combined spinal-epidural technique in 40 nulliparous parturients with IT 2.5 mg bupivacaine (group B), ropivacaine (group R). The duration was the longest in group B P < 0.05 (mean +/- SE, 76.3 +/- 5.9 min vs 52.6 +/- 4.0 min ,). IMPLICATIONS: Intrathecal 2.5 mg bupivacaine significantly prolongs the duration of analgesia in laboring patients compared with ropivacaine
Does the duration of analgesia in laboring patients after intrathecal bupivacaine is more compared with ropivacaine ?
Design Double blind RCT Parameters Effect size: 25 minutes more with ropivacaine SD common for 2 groups = 24 minutes value =0.05 Z =1.65 value =0.10 Power=90% Z = 1.282
n =
(Z
+ Z d
) S
2
2
n =
( 1 . 96
1 . 28 ) 24 2 2 25
Issues:
Need to differentiate between Descriptive and Testing Hypothesis or analytical studies. Different formulae depending upon the type of study Incorporate drop out rates. Cluster designs should inflate the size based on Design effect.
Difference between two means is the parameter of our study Example 1 The lung function of two groups of men is to be compared. Using forced expiratory volume (FEV). Literature search: SD 0.5l Mean difference of clinical interest: 0.25l Two sided significance level = 0.05 ( Z2w = 1.96 ) Power = 80% ( ZF = 0.842) n > 2 [( Z w + Z F ) SD / D]2 = 2 [ ( 1.96 + 0.842 ) ( 0.5 )/ 0.25 ]2 = 62.8 63 men in each group
Example 2
Duration of ICU study (in hours):
o Dose ean D n i h Dose
5 (
= ) 5 = ( 4 5 ) *
n=
n=
Describing Variability
Example: n = 5 systolic blood pressures (mm Hg) X 1= 120 X 2= 80 X 3= 90 X 4= 110 X 5= 95
Describing Variability
Example: n = 5 systolic blood pressures (mm Hg) Recall, from earlier: X = 99 mm HG Now:
X i X = 120 110
2 2 2 2
i=1
= 1020
Describing Variability
Sample Variance
n
i= 1
X i X n 1
1020 = = 255 4
Notes on s
The bigger s is, the more variability there is s measures the spread about the mean s can equal 0 only if there is no spread
All n observations have the same value
The units of s is the same as the units of the data (for example, mm Hg)
Notes on s
.3
68%
.2
95%
.
99.7%
0 83 97 25 39 53 67
The 68-95-99.7 rule applied to the distribution of systolic blood pressure in men.
Example
Measure systolic blood pressure on random sample of 100 students Sample size n = 100 Sample mean X = 123.4 mm Hg Sample SD s = 14.0 mm Hg 14 = 1 . 4 mmHg SEM = 100
The director of community Health observes that rural women have a higher rate of miscarriage than urban women since farming is their primary occupation relying heavily on organophosphorous pesticide. He hypothesizes that rate of miscarriage is related to exposure to this pesticide. Test this at alpha = 0.01. Miscarriage-No miscarriage - Total Exposed - 30 - 70 - 100 Not exposed 10 90 - 100 - 40 - 160 - 200
Determine relative risk (RR) and odds ratio (OR) What are confidence intervals of OR. Determine the magnitude of the public health problem posed by such exposure. Calculate attributable risk, population attributable risk Attributable fraction, Total population attributable fraction
Quantifying risk
Relative Risk = Incidence of disease in exposed group Incidence in non exposed group = 30/100 = 3 10/100 Odds Ratio = Odds that exposed individual has disease Odds that non exposed has disease = 30 / 70 = 3.86 10/90
Example 1 From a pilot study it was reported that 28% post operative had Shivering. It was decided to have 95% C.I with 5% variability in the estimated 28%. How many patients are necessary to conduct the study.
P = 28% q = 72% ZE= 1.96 for E at 0.05 d = 5%
Example (Qualitative)
ta lit
ea i L ta l se se ( ( 6 ( ) ) )
liv e
ta l
in each group