Chronic diseases
1. Chronic diseases have long and variable
preclinical phases.
2. The preclinical phase is that portion of
the disease natural history during which
the disease is potentially detectable, but
unrecognized.
Chronic diseases
3. One can conceive of complex functional
relationships (sensitivity function)
between time in preclinical phase and
the sensitivity of a screening test for
disease.
*
Chronic diseases
4. Lead time is the interval between the
time of disease detection through
screening and the time of disease
recognition in the absence of screening.
The lead time produced by a screening
program for a given individual depends
on the time of screening, in relation to
the preclinical phase, and on the
sensitivity function of the screening test.
Chronic diseases
5. The relative success of treatment for
screen detected disease depends on the
earliness of detection (or the amount of
lead time produced by screening).
*
**
Chronic diseases
6. Some chronic diseases have relatively
short preclinical phases (on average),
without much variability between
diseased individuals. Some chronic
diseases have very long preclinical
phases (on average), with sizable
variability between diseased individuals.
Chronic diseases
7. The distribution of lead times produced
by a screening program depends on
durations of the preclinical phase, on the
periodicity (or frequency) of screening,
and on the sensitivity function of the
screening test.
Chronic diseases
8. Pseudodisease is a condition known only as a
result of screening, a condition that would have
remained unrecognized if not for screening.
Pseudodisease is an instance of screen-
detected disease with indefinitely long lead
time. Pseudodisease includes screen-
detected cases which would have regressed
(even in the absence of treatment), screen-
detected cases which would not have pro-
gressed (even in the absence of treatment),
and screen-detected cases with very long lead
times, relative to remaining life expectancy.
Chronic diseases
9. The success of secondary prevention
depends on progressive disease existing
in a preclinical phase, a screening test
capable of detecting disease in the
preclinical phase, and the production of
lead times (coupled with treatments)
sufficient to alter long term outcomes
from disease.
Difficulties associated with
inferences based on prognosis
1. Prognosis refers to outcomes among
persons known to have disease, whether
screen-detected or symptom-detected.
Case-fatality is a measure of prognosis.
Difficulties associated with
inferences based on prognosis
2. As a natural consequence of the earliness of detection,
one would expect a successful screening program to
improve the prognosis of screen-detected cases,
relative to symptom-detected cases. One would also
expect better prognosis among all cases, which come
to attention among participants in a screening program,
relative to all cases, which come to attention among
non-participants of the program. Improved prognosis
(reduced case-fatality) is a direct consequence of lead
times produced through the act of screening for
disease.
Difficulties associated with
inferences based on prognosis
3. However, two important biases influence
prognosis among screen-detected cases,
even in the absence of any benefit from
treatment.
Difficulties associated with
inferences based on prognosis
4. Lead time bias -- Screen-detected cases
experience lead time. Among screen-
detected cases, lead time contributes to
the duration at risk for poor outcome.
Thus, even if screening does not change
the time of death, screening will increase
the proportion of screen-detected cases
surviving beyond defined time intervals.
Difficulties associated with
inferences based on prognosis
5. Length biased sampling -- Intermittent
screening preferentially detects cases
with a long preclinical phase; that is,
cases of less rapidly progressive
disease. These persons would be
expected to experience a relatively
favorable outcome, even if allowed to
progress to symptoms.
Difficulties associated with
inferences based on prognosis
6. The effects of lead time bias and length
biased sampling are difficult to separate
from the effects of treatment. Therefore,
prospective evaluations of the efficacy of
screening must compare outcomes
among all persons exposed to screening
with outcomes among all persons not
exposed to screening.
Prospective evaluation
Outcome
Poor Good Total
Exposure status Screened group a b N1
Unscreened group c d N2
N
Difficulties associated with
inferences based on prognosis
7. Retrospective evaluations of the efficacy
of screening must compare cases who
experienced a preventable adverse
disease outcome and controls who are
at-risk for the adverse disease outcome.
Remote history of screening during the
preclinical phase constitutes the relevant
exposure.
Retrospective evaluation
Screening history
Present Absent Total
Disease status Poor outcome a b N1
Good outcome c d N2
N
Effects of periodic screening
1. Upon initiation of a new screening program,
the prevalence of preclinical disease in the
population depends on the incidence of
preclinical disease and on the average
duration of preclinical disease.
2. The first round of screening preferentially
removes prevalent cases of disease with
relatively little time remaining in their preclinical
phase.
Effects of periodic screening
3. Subsequent rounds of screening have the potential of
detecting new individuals (incident cases) entering their
preclinical phase since the last round of screening.
4. Depending on the periodicity of screening, in relation to
the average duration of the preclinical phase, cases
detected during subsequent rounds of screening
include a high proportion of incident cases with
relatively long lead times. Relative to cases detected
at the first screening encounter, cases detected during
later encounters may experience more benefit (on
average) from early treatment of disease.
Estimating sensitivity in the
absence of a gold standard
1. Sensitivity is the proportion detected by a
screening test among all cases in the
preclinical phase. (Sensitivity is the area
under the sensitivity function, weighted
according to the distribution of times remaining
in the preclinical phase).
2. The sensitivity of a screening test will depend
on any factor which changes the distribution of
times remaining in the preclinical phase (such
as, age of screened group, previous screening
history, self-referral status, level of awareness
of disease in the population).
Estimating sensitivity in the
absence of a gold standard
3. The calculation of sensitivity requires an ability
to distinguish false negatives from true
negatives. In practice, the application of the
gold standard may require invasive or
expensive diagnostic testing. It may be
inappropriate or unacceptable to perform such
testing among persons with negative screening
test results. Moreover, even the so-called
"gold standard" may fail among cases early in
the preclinical phase. But, screening programs
attempt to detect disease as early as possible
in the preclinical phase.
Estimating sensitivity in the
absence of a gold standard
4. Indirect calculation of sensitivity (detection
method) = ratio between screen-detected
cases and screen-detected plus interval cases.
This calculation may overestimate sensitivity.
5. Indirect calculation of sensitivity (incidence
method) = 1 - ratio of interval disease
incidence rate in a screened group and the
disease incidence rate in a comparison group.
May underestimate sensitivity.
