Screening

UMO
Do you agree?
 early diagnosis” of disease is beneficial, then …
• screening is bound to be effective
• screening for lung cancer using sputum cytology or
chest X-rays can reduce mortality from the disease
 screening can involve simple clinical
examinations
• such as assessment of blood pressure, or a clinical
breast examination
Screening: defined
 “the presumptive identification
• of unrecognised disease or defect *
• by the application of tests, examinations or other
procedures
• that can be applied rapidly.”
 A screening test is not intended to be diagnostic
• Rather a positive finding will have to be confirmed by
special diagnostic procedures
• United States Commission on Chronic Illness (1957)

 *subtly different from being asymptomatic

General Principles Governing the
Introduction of Screening
 General Principles of Screening
1. the disease should be an important health problem;
2. the disease should have a detectable preclinical phase;
3. the natural history of the lesions identified by screening
should be known;
4. there should be an effective treatment for such lesions;
and
5. the screening test should be acceptable and safe
• Wilson and Junger (1968)
• Miller (1978,1996) Cuckle and Wald (1984)
1. The disease should be an important
health problem
 disease prevalence should be high
 the disease should be the cause of
substantial mortality and/or morbidity
 However, the life expectancy of a screened population may be changed little even if
the program is successful
 There may be certain circumstances when the major benefit from screening follows,
not from reduction in mortality, but from reduction of morbidity consequent upon the
diagnosis of the disease in a more treatable phase in its natural history
2. The disease should have a detectable
preclinical phase (DPCP)
4. The disease should be treatable
 there should be a recognized treatment
for lesions identified following screening
• there should be evidence of the effectiveness
of treatment of lesions discovered as a result
of screening in reducing
• disease incidence and|or mortality and
• the level of improvement expected should be
stated
5. The Acceptability of the Test
 A screening test, therefore, has to be
acceptable to the population in its widest sense
• simple, easily administered as far as possible
• involve procedures that are not unacceptable
• not have unpleasant or potentially hazardous
implications
• e.g. of acceptability
• cervical cytology screening: generally high; but low
among high risk groups
• Breast cancer screening: varies – 74% to 90%
• screening for colorectal cancer (occult blood)– low, to
75%; ? scopy ??
Validity of a Screening Test
Validity of a Screening Test
 Sensitivity is defined as the ability of a test to detect
all those with the disease in the screened population
expressed as
• the proportion of those with the disease in whom a
screening test gives a positive result

 Specificity is defined as the ability of a test to

correctly identify those free of the disease in the
screened population
expressed as
• the proportion of people free of the disease in whom the
screening test gives a negative result
ROC curve
 If the test result is expressed in a quantitative
form
• so that the boundaries between what is defined as
positive and negative can be varied at will
• it is possible to plot a receiver operating characteristic
(ROC) curve (Swets 1979)
 What is plotted is
• the sensitivity in the vertical axis
• 1-specificity (the false positive rate) in the horizontal
axis
Receiver Operator Characteristic (ROC) Curve
ROC curve to determine best cutoff point for scc by means of meanrlu

 Plot true positive rate 100

(sensitivity) against false 90
50 10
100
positive rate (1-specificity) for 80
several choice of positively s
70
1000 (mean rlu)

criterion e
n 60
10000
 choose closest to top left s
i 50
corner to maximized the t
40
25000
50000
i
discriminative ability of the test v 30
i
t 20
y
10

0
0 20 40 60 80 100
1- specificity
Receiver Operator Characteristic (ROC) Curve

ROC curve to determine best cutoff point for Wilsom Risk sum
 The area under the curve scoring to detect difficulty of endotracheal intubation
100
represent overall 1
0
90
accuracy of the test
80
 useful to compare two test70
2
sensitivity
60

503
40

30

205

10
0
0 20 40 60 80 100
1- specificity
It is important
 The validity of the screening test to be
used should have been evaluated and
their expected values stated,
 There should be an acceptable program
of quality control to ensure that the
stated levels of sensitivity and specificity
are attained and maintained.
2*2 table
D+ D-

Test + 180

Test - 20

Prevalence 2% 200 9,800 10,000

Sen = 180/200=90%
 2*2 table
D+ D-

Test + 180 490

Test - 20 9310

Prevalence 2% 200 9800 10000

Spec = 9310/9800=95%
 Sensitivity & Specificity
Sensitivity: ability of a test to
detect a disease when it is
present [a / a+c]
If the test is not sensitive- it fail
to detect disease in some of
the diseased subjects, and is
called False negative error
rate [c/a+c] = 1-sen
 Sensitivity & Specificity
Specificity: ability of a test
to indicate non-disease when it
is absent [d/b+d]
If the test is not specific- it falsely
indicate the present of disease
in some of the non-diseased
subjects, and is called False
positive error rate [b/b+d] = 1-
sp
 Rule Out & Rule in
 A screening test, which is used to rule out a
diagnosis should have a high degree of
sensitivity >> SnNout

 A confirmatory test which is used to rule in

a diagnosis should have a high degree of
specificity>> SpPin
predictive values
 Predictive Values

 If a patient’s result is positive, what is the

probability that he or she has the disease being
tested?
 How good is the test at identifying people with
the disease?
 …. people without the disease?
 2*2 table

D+ D-
PPV= 180/670
Test + 180 490 670 = 27%

Test - 20 9310

Prevalence 2% 200 9800 10000

 2*2 table

D+ D-

Test + 180 490

Test - 20 9310 9330

NPV= 9310/9330
= 99.8%
Prevalence 2% 200 9800 1000
 Predictive Values
 Positive predictive
value (PPV): proportion
of subjects who had
positive test results
had the disease
 [a / (a+b)]
 Predictive Values
 Negative predictive
value (NPV):
proportion of subjects
who had negative test
results were free of the
disease [d/(c+d)]
Revising estimates of Predictive values

D+ D-
PVP= 1800/2200
T+ 1800 400 2200 = 82%
T- 200 7600 7800 PVN= 7600/7800
2000 8000 10000 = 97%

Prevalence 20%
relationship of disease prevalence and predicitve values

General Rule:
low prevalence > low PVP; high PVN
high prevalence > high PVP; low PVN
use of multiple tests
 sequential (two-stage) testing
• there is a lost in net sensitivity
• there is a gain in net specificity
 simultaneous testing
• gain net sensitivity
• lost net specificity
Reliability (Repeatability) of Tests
Results replicated if repeated
 Intrasubject variation

 Intraobserver variation

 Interobserver variation
Measuring agreement

Observed agreement = a+d =Ao

Max. possible agreement = a+b+c+d = N
Overall percent agreement = (a+d) / (a+b+c+d)
 Measuring agreement

cell a agreement expected by chance = (a+b)*(a+c) / (a+b+c+d)

cell d agreement expected by chance = (c+d)*(b+d) / (a+b+c+d)
Ac: Total agreement expected by chance = cell a agreement expected ..
+ cell d agreement expected ...
(Ao - Ac) / (N- Ac) = kappa
 Measuring agreement
Kappa statistics

Kappa is a ratio:
 the numerator is the observed improvement
over chance agreement (A0-Ac)
 the denominator is the maximum possible
improvement over chance agreement (N-Ac)

Kappa: agreement beyond chance agreement

 Measuring agreement

Observed agreement = Ao = 30+60=90

Max. possible agreement = N =100
Overall percent agreement = 90/100=90%
Cell a agreement expected by chance = (37)*(33) / (100)=12.2
Cell d agreement expected by chance = (67)*(63) / (100)=42.2
Ac: Total agreement expected by chance = 12.2+42.2=54.4 ...
Kappa = (Ao - Ac) / (N- Ac) = (90-54.4)/(100-54.4) = 0.78=78%
 Interpreting Kappa statistics
The kappa statistics measurement of agreement is scaled to be:
0 when the amount of agreement is what would be
expected to be observed by chance, and,
1 when there is perfect agreement

Size of  agreement
0.81 to 1.00 Almost perfect
0.61 to 0.80 Substantial
0.41 to 0.60 Fair
0.21 to 0.40 Slight
0.00 to 0.20 Poor

* Suggested by Landis and Koch ( 1977a, 165)

Four biases associated with screening
 Lead time: the period by which screening
advances the diagnosis of the disease
 Length bias: less rapidly progressive cases
are likely to be detected by screening
 Selection bias: volunteers for screening are
likely to have a better outcome from their
disease than those who decline screening
 Overdiagnosis bias: lesions identified by
screening which would not have progressed to
clinical disease in the absence of screening
lead time bias
 the problem of an illusion of better
survival only because of earlier detection
is called the lead time bias
The Ethics of Screening
 burden of proof for efficacy of the procedures
and the necessity to avoid harm
• screening will not benefit to everyone who is screened
• harm from a screening test is not only related to the risk
of being false positive or negative: may also incur
psychological consequences
• proscription of mammography in women under the age of 50 in the
Breast Cancer Detection Demonstration Projects in the US (Beahrs et
al.1979)
The Ethics of Screening
 equity issue
• obligations for appropriate care in the
community than towards individuals
• problem for developing countries by diverting
resources intended for routine health care into
screening
• resources diverted to a screening project, which might
be regarded as prestigious, especially if involving high
technology, could lower the resources available for
other more pressing but also more mundane health
problems
The Ethics of Screening
 implementation of informed consent
• information about risks and benefits of tests and
treatments
• difficult to provide
• sometimes lack evidence
summary for screening
 definition
 general principles
 validity of a screening test
 predictive values
 relationship bt prevalence and predictive
values
 use of multiple testing
 bias
 ethics and (economics)
Task: at library
 Definition: See JM Last: p 165
 types of screening: mass, multiple,
prescriptive
 characteristics: accuracy, yields,
precision, reproducibility, validity
 others: case finding; DPCP
HW
 Definitions
 types of screening: mass, multiple,
targeted, opportunistic, prescriptive
 SN: Bias - from Gordis ++
 Issues of use of multiple test in hospitals

 Screening =/= Screaming !