This document discusses key concepts in understanding evidence from epidemiological studies, including: - The difference between prevalence and incidence and how they are calculated - Common effect measures used to quantify associations including relative risk, risk difference, odds ratio, number needed to treat, and hazard ratio - The importance of confidence intervals and p-values in determining statistical significance and how they are interpreted - The difference between statistical significance and clinical significance when evaluating study results

1. สอนโดย อ.กมลวรรณ ตันติพิวัฒนสกุล
ได้รับการสนับสนุนจาก อ.มนทรัตม์ ถาวรเจริญทรัพย์

2.  Understanding the evidence
◦ Prevalence VS incidence
◦ RR, RD, OR, NNT, NNH, HR
◦ Significant testing (P-value, 95% CI)
◦ Clinical significant VS statistical significant

3.  Prevalence
◦Point prevalence
◦Period prevalence
 Incidence
◦Cumulative incidence
◦Incidence density (Incidence
rate)

4. 4
 Prevalence: Proportion of the population affected by the
disease at the time. (any cases!)
Prevalence per 1,000 =
No. of cases of a disease present in the
population at a specific time x 1,000
No. of persons in the population at the
specified time
Point prevalence: Prevalence of the disease at a certain
point in time. (Interview question: Do you currently have
asthma?)
Period prevalence: Prevalence of the disease at any point
during a certain time period. (Interview question: Have
you had asthma during the last 6 months?)

5. 5
 Incidence: The number of new cases of disease
that occur during a specified period of time in a
population at risk for developing the disease.
Incidence per 1,000 =
No. of NEW cases of a disease occurring in
the population at a specific time x 1,000
No. of persons who are at risk of developing
the disease during that period of time
Incidence rate per 1,000 =
No. of NEW cases of a disease occurring
in the population at a specific time x 1,000
Total person-time of observation

6.  On January 1, 2001:The total number of
HIV infected patients in City A was equal to
200. During January 2 through December
31, 2001, 100 more cases was developed in
the city. Assume that the total population in
City A is fixed at 20,000. (No death, no
migration)
◦ Point prevalence at January 1,2001= 200/20,000
◦ Period prevalence (during 1 year) = 300/20,000
◦ Incidence (during 1 year) = 100/19,800

7. Incidence density: (incidence rate)
Incidence density = # of new cases during the time period
total person-time of observation
•Assumption: everyone in the candidate population has
been followed for a specified time period.
•Person-time is accrued only while the candidate is
being followed.
•Accrual of person time stops when the person dies or
is lost to follow-up or diagnosed with disease of interest.

8. Person-time: is the denominator used
when people have been followed for
different lengths of time.
How to calculate?
 Count the number of time units
experienced by each person in the
study, then add them up!

9. Person-years (P-Y) = Equivalent number of
people who would have been at risk for one full
year.
Example:
1000 P-Y = 1000 people x 1 year
= 500 people x 2 years
= 100 people x 10 years
= (200 people x 2.5 years ) +
(500 people x 1 year)

11. Incidence rate = 100/ 194,400

12. Cumulative incidence = ?
Incidence rate = ?

13. Cumulative incidence = 40/350
Incidence rate = 40/1,560

14. Disease No disease
Exposed
(treatment)
A B A+B
Non
exposed
(control)
C D C+D
A+C B+D A+B+C+D

15.  Relative risk = Cumulative incidence in exposed gr.
Cumulative incidence in unexposed gr.
= [a /(a+b)] / [c / (c+d)]
 Relative rate = Incidence rate in exposed gr.
(Relative risk ratio) Incidence rate in unexposed gr
= a / person-time exposed.
c / person-time unexposed

16.  Use in cohort study, randomized clinical trial
 RR = Cumulative incidence of disease in exposed gr
Cumulative incidence of disease unexposed gr.
 RR = Incidence rate of disease in exposed gr
Incidence rate of disease in unexposed gr.
 Therefore,
RR = 1 means there is no association between
exposure and disease.
RR > 1 means exposure = risk factor
RR <1 means exposure = protective factor

17.  RR = Cumulative incidence of disease in exposed gr
Cumulative incidence of disease in unexposed gr.
Therefore,
RR = 2 means
 The exposed group is 2 times more likely to have
disease when compared to unexposed group.
RR = 0.4 means
 The exposed group is 60 % less likely to have disease
when compared to unexposed group.

18.  Absolute Risk difference
=  Cumulative incidence in exposed -
Cumulative incidence in unexposed 
=  Control event rate (CER) - Experimental
event rate (EER) 
Risk difference describes the excess risk of
disease in those exposed compared with those
who were unexposed.

19. Absolute risk difference
Situation 1: Treatment  bad event (CER > EER)
Absolute risk reduction (ARR)  CER - EER
Relative risk reduction (RRR)  CER - EER / CER =1-RR
Situation 2: Treatment  bad event (CER < EER)
Absolute risk increase (ARI)  CER - EER
Relative risk increase (RRI)  CER - EER / CER

20. Absolute risk difference
Situation 3: Treatment  good event (CER < EER)
Absolute benefit increase (ABI)  CER - EER
Relative benefit increase (RBI)  CER - EER / CER
Situation 4: Treatment  good event (CER > EER)
Absolute benefit reduction
(ABR)
 CER - EER
Relative benefit reduction (RBR)  CER - EER / CER
Example: Sildenafil for male erectile dysfunction. Percentage of men
experiencing at least 1 intercourse success during treatment (83% VS 45%)
Relative Risk= 1.8; 95% CI (1.17 – 1.9), Absolute benefit increase = 38%
Arch Intern Med 2002, 162: 1359-1360

21. A clinical trial compares the effect of a new oral
anti-diabetic drug and placebo on the
incidence of stroke.
 incidence of stroke is 4% with the new oral
anti-diabetic drug and 6% with placebo.
Stroke:
Absolute risk reduction (ARR) with new oral
hypoglycemic = 2%
Relative risk reduction (RRR) = 2% / 6% or
= 1- (4%/6%) = 33.33%

22. A clinical trial compares the effect of a new oral
anti- diabetic drug and placebo on the incidence
of hypoglycemic .
 Incidence of hypoglycemic is 8% with the new oral
anti-diabetic drug and 5 % with placebo.
hypoglycemic:
Absolute risk increase (ARI) with new oral
hypoglycemic = 3%
Relative risk increase (RRI) = 3% / 5% = 60%

23.  Stroke incidence in drug A group = 0.00030
 Stroke incidence in placebo group = 0.00020
Risk difference (Absolute risk increase)
= 0.00030 - 0.00020 = 0.00010
Interpretation: Those who received drug A
increased the risk of stroke by 0.00010 =
10 per 100,000 person
How about Relative risk?

24.  Stroke incidence in drug A group = 0.03
 Stroke incidence in placebo group = 0.02
Risk difference (Absolute risk increase)
= 0.030 - 0.020 = 0.01
Interpretation: Those who received drug A
increased the risk of stroke by 0.01 = 10
per 100 person
How about RR?

25.  In 1970s oral contraceptives were
found to increase risk of MI 2.5 to 5-
fold.
 This statistic sound very alarming until
one considers that this is an absolute
risk of 3.5 death per 100,000 users per
year!
Gehlback S.H. Interpreting the medical literature 3rd ed. 1993,
McGraw Hill.

26. The number needed to treat (NNT) is the estimate
represents the number of patients one would need to
be treated over a specific time to prevent one clinical
event.
NNT = 1/ Absolute risk reduction
= 1/ (CER-EER)
Example: NNT = 15.17 means to prevent 1person from
developing y disease, 15.17 people would need to get x
treatment.
The ideal NNT is 1, where everyone improves with
treatment and no one improves with control.
The higher the NNT, the less effective is the treatment.

27.  The number needed to harm (NNH) is the
number of patients who would need to be
treated over a specific period to cause harm in
one patient that would not otherwise have been
harmed
 NNH is defined as the inverse of the absolute
risk increase
 The lower the number needed to harm, the
worse the medicine

28.  The NNT of aspirin to prevent 1 vascular event
is about 25.
 The NNH inducing 1 cerebral bleeding is about
1000.
 The NNH to provoke 1 severe extracerebral
bleeding about 100-200.
Antiaggregation:aspirin; The Umsch. 2003; 60(1):15-8.

29. Suppose, for 5 years of follow up, stroke occurs
in 60 % of control group and 40% in
experiment group.
NNT = 1 / (60%-40%)
= 5
Suppose, for 5 years of follow up, 37% of control
group has a chance of developing ADR ,
compared to 64% in the experiment group.
NNH = 1 / (64%-37%)
= 4

30.  By 33 month, the disability occurred in
◦ 50% of patients with multiple sclerosis
randomized to control group (placebo)
◦ 39% of patients assigned to receive
interferon.
 Please calculate the relative risk, relative risk
reduction, number needed to treat (NNT)

31.  RRR = (50% - 39%) / 50 %
= 22 %
= 1-RR = 1- (0.39/0.5) =0.22
Interpretation: Interferon decreased the
risk of disability by 22 %

32. NNT = 1 / (50% -39%)
= 1/11%
= 9
Interpretation: We would need to treat 9
people with interferon for 33 months in
order to prevent 1 additional person from
disability due to MS .

33.  Concept of NNT always refers to a
comparison group, a particular treatment
outcome, and a defined period of treatment.
 The NNT is the number of patients that you
will need to treat with drug A to achieve an
improvement in outcome compared with
drug B for a treatment period of C weeks or
year.

34.  Very small NNT( that is, one that approaches 1)
means that a favorable outcomes occurs in nearly
every patient who receives the treatment.
 It is inappropriate to compare NNTs across disease
conditions, particularly when the outcomes of
interest differ.
◦ NNT of 30 for preventing deep venous
thrombosis ≠ NNT of 30 for preventing the
disabling stroke.
◦ If we have NNT for different interventions for the
same condition (and severity) with the same
outcome and duration, then, and only then, is it
appropriate to directly compare NNTs.

35.  Odds of an event = the number of event
the number of non event
 Odd ratio = odds of exposure in cases = A/C = AD/BC
odds of exposure in control B/D
= (a/c) / (b/d)
= ad/bc
Disease No disease
Exposed
(treatment)
A B A+B
Non
exposed
(control)
C D C+D
A+C B+D A+B+C+D

36.  OR = AD / BC
 Interpretation: Same as RR
= (a/c) / (b/d)
= ad/bc
Disease No disease
Exposed
(treatment)
A B A+B
Non
exposed
(control)
C D C+D
A+C B+D A+B+C+D

37. 1. The controls are representative of the target population
2. The cases are representative of all cases.
3. The frequency of the disease in the population is rare. (<15%)
RR = (A/A+B) / (C/C+D) If rare disease, A<<<<<B then A+B =
B, C<<<<D then C+D=D. Therefore, in rare disease, Relative
Risk = Odd Ratio
= (a/c) / (b/d)
= ad/bc
Disease No disease
Exposed
(treatment)
A B A+B
Non
exposed
(control)
C D C+D
A+C B+D A+B+C+D

38.  Survival analysis: Comparison of time-to-
event among different groups.
 Kaplan-Meier survival curve: A life table
curve showing the percent of people free of
a specific event at time following
randomization.
 Log-rank test: A statistics used to
compared 2 survival curves.

40. Median ratio for placebo VS 500 mg
Valaciclovir = 5.9 / 4.0 = 1.5

42.  Hazard ratio (is an estimator of RR), is an estimate of
the ratio of the hazard in the treated VS the control
group.
 Hazard = Instantaneous end point probability at time t
survival probability at time t
 Interpretation: Same as RR

44.  Significant test
 P-value
 Confidence interval (CI)
 Clinical VS statistical significant

45.  Commonly use a cutoff point of 0.05 to determine if the
Ho should or should not be rejected. This cutoff point is
known as alpha level or significant level.
 The significance level is the chance of rejecting the null
hypothesis when it is true. (Incorrectly conclude that there
is a difference when there is none- false positive.)
 P-value= probability of obtaining the observed result and
more extreme result by chance alone, given that the null
hypothesis were true.
 Usually if P value < 0.05(level of significant) we will reject
Null hypothesis and conclude that there is a significant
difference.
 P-Value is calculated, Level of significant is set !

46.  Confidence interval is used to express the
degree of confidence in an estimate (such as
odd ratio, relative risk)
 Confidence Interval (CI) gives range within which
that “true value” probably lies.
 95% CI - if we repeated the experiment with similar
populations an infinite number of times, the results
would fall within the CI 95% of the time. 95% certain
that the “true value” will fall within the 95% CI
range.

47.  For Odd ratio and Relative Risk, if
95% CI contains 1 means there is no
significant difference.
 For risk difference and mean
difference if 95% CI contains 0 means
there is no significant difference.

48. RR 95% CI
All cause mortality 0.83 0.73 to 0.95
Fatal and non-fatal
CVD
0.71 0.61 to 0.79
Revascularisation
rate
0.66 0.53 to 0.83
Cochrane Database Syst Rev 2011; 19(1):CD004816
Statins for the primary prevention of cardiovascular disease

49. N Eng J Med 2006;354:1706-1717

50. 50N Engl J Med 2009;360:225-35

51.  Statistical significance measures how likely that
any apparent differences in outcome between
treatment and control groups are real and not
due to chance. p Values and confidence intervals
(CI) are the most commonly used measures of
statistical significance.
 Statistical significant does not imply medical or
clinical significant and does not mean that bias
or confounding have been ruled out.( It is
entirely possible to have a statistical significant
association that is invalid.
 Clinical significance measures how large the
differences in treatment effects are in clinical
practice.

52. Penciclovir cream for the treatment of herpes simplex labialis. A
randomized, multicenter, double-blind, placebo-controlled trial. Topical
Penciclovir Collaborative Study Group.
JAMA. 1997 May 7;277(17):1374-9
 OBJECTIVE:To compare the safety and efficacy of topical 1%
penciclovir cream with vehicle control cream (placebo) for the
treatment of a recurrent episode of herpes simplex labialis (cold
sores) in immunocompetent patients.
 Results: Healing of classical lesions (vesicles, ulcers, and/or
crusts) was 0.7 day faster for penciclovir-treated patients
compared with those who received vehicle control cream (median,
4.8 days vs 5.5 days; hazard ratio [HR], 1.33; 95% confidence
interval [CI], 1.18-1.49; P<.001). Pain (median, 3.5 days vs 4.1
days; HR, 1.22; 95% CI, 1.09-1.36; P<.001) …

53. Three ground rules in analysis of experimental study:
 Participants used in treatment comparison should be
counted in the treatment group to which they are
assigned.
 The denominator for a treatment should be all
participants assigned to that treatment.
 All events counted in the comparison of primary
interest.
Benefits of intention to treat analysis:Maintains the
protection of randomization(prevention of bias), since
analysis based on actual assignment.

54.  All individuals who are randomly allocated to a treatment are
analyzed, regardless of whether they complete or even receive the
treatment.
enroll eligible and willing patients
Random assignment
Treatment 1 Treatment 2
Completed Did not complete Completed Did not complete
Treatment 1 treatment 1 treatment 2 treatment 2
Group 1 Group 2 Group 3 Group 4

55.  ITT prevents bias caused by loss of
participants, which may disrupt the baseline
equivalence by random assignment and
may reflect nonadherence to the protocol.
 Clinical effectiveness may be overestimated
if an ITT is not done.
 For superiority trials, the intent- to- treat
analysis (ITT) is considered the primary
analysis.
 For noninferiority, both intent-to-treat
analysis and per-protocol-analyses should
be performed.

56. R
Surgery
500
Drug
500
1,000
Eligible
patients
1 year 5 year
10 deaths occurred
before surgery
10 deaths occurred
after surgery
10 deaths 10 deaths
Analysis
PP:
RR =10/490
20/500
= 0.51
ITT:
RR =20/500
20/500
=1
All patients received drug at the 1st day of study

57. Compliance Clofibrate Placebo
Number of
patients
Mortality (%) Number of
patients
Mortality (%)
Poor (< 80%) 357 24.6% 882 28.2%
Good (> 80%) 708 15.0% 1813 15.1%
Total group 1065 18.2% 2695 19.4%
57
•Clofibrate (good compliance) VS placebo (total group): 15.1% VS 19.4%
•ITT: No significant different was found (18.2% VS 19.4%)
Note: Mortality risk is different between poor compliance and good
compliance even in placebo group ! ( Those who comply with the
medication are basically different in many factors. That is why ITT
should be done- to maintain the benefit of randomization)
Adapted from N Eng J Med 1980; 303: 1038-41

58. ◦ Multivariate analysis of prognostic factors to predict
the most likely outcomes in those loss to follow up.
◦ Imputation of outcomes by carrying the last known
outcome status forward,
◦ Best-case and worst-case scenario
However, if there is significant loss to follow-up,
statements that investigators conducted an ITT
generally provide reassurance!