Assessment of Screening Interventions and Programs

1
Assessment of Screening Interventions and
Programs
2
(No Transcript)
3
Well cover...

• The general conceptual model epidemiology for
  screening
• Problems in assessing evidence for effectiveness
  of screening
• An example Screening for AAA
• Hot topics in screening--
• breast cancer screening
• prostate cancer screening
• colorectal cancer screening

4
My copy shown here is pretty beat up... but still
an excellent reference for general epidemiology
of screening AS Morrison Screening in Chronic
Disease Oxford University Press, 1985.
5
Other references

• Report of US Preventive Services Task Force.
  Guide to Clinical Preventive Services. 2nd
  edition 1996.
• LB Russell. Is Prevention Better than Cure? The
  Brookings Institution Washington DC, 1986.
• LB Russell. Educated Guesses making policy
  about medical screening tests. University of
  California Press, 1994.

6
Screening(see Eddy, Vol. 1, 30, p 280)

• Two basic assumptions
• The disease is progressive
• Earlier treatment more effective than later
  treatment
• Often implied, but not always true
• cheaper to screen treat early, than to treat
  later stages and/or die from the disease

7
Things to look at when evaluating screening

• What is benefit of finding an early case? What
  is cost of missing an early case?
• is there good evidence of early treatment
  effectiveness?
• is there other tangible benefit of case finding?

8

• Characteristics of population to be screened
• Burden of disease in this population
• morbidity mortality
• prevalence incidence
• Is selective screening feasible?
• high risk group identifiable easily?
• What are the competing risks in this population?

9

• Characteristics of the screening test itself in
  the target population
• sensitivity
• specificity
• cost per screen
• Gold standard (confirmatory) test
• sensitivity, specificity
• costs
• side effects

10

• What are the costs of
• treatment for the disease?
• stage by stage
• a false negative?
• how are cases usually found? Does missing the
  case on screen mean it is missed forever?
• false positive?
• risks of confirmatory test
• psychologic risks / harms worry, etc.
• labeling

11
A general model for screening... first lets
look at a disease we may be familiar with
12
Natural history timeline for cervical carcinoma
dysplasia
Asymp-tomatic invasive cancer
Sympto-matic invasive cancer
Death from metastatic cancer
Ca in situ
(intervals not to scale -- will vary from patient
to patient)
13
General model for a progressive disease
Biologic onset
First detectable by screening test
Severe clinical illness (eg metastases)
Death from the disease
Usual time of diagnosis
Pre-sx interval
The benefit of screening is to gain lead time
important to have evidence this in fact produces
better outcomes!
Lead Time
Actually detected by screen
14
Causal Pathways(Battista Fletcher)

• Useful tool to map out relationship between
  screening and the clinical events that must occur
  for a given maneuver to influence a target
  condition
• illustrate with causal pathway for early
  detection of hypertension

15
Blood pressure measurement
Antihypetensive treatment
Hypertensive individuals identified
Blood pressure controlled
Occurence of stroke prevented
Asymptomatic individuals
The most direct -- and strongest -- evidence of
benefit would be RCT of program to measure blood
pressure in asymptomatic individuals, then
following them long term to measure reduction of
stroke incidence
16
evidence that control actually leads to desired
outcomes in these individuals
ability to control the intermediate condition
ability to detect the condition
Without such evidence, need to infer
effectiveness by combining evidence about the
separate links
17

• Recall how evidence is evaluated
• Vol. 1, 14 , p 109 ff --Mulrow et al.
  Integrating heterogeneous pieces of evidence in
  systematic reviews. Ann Intern Med 1997
  127989-995.

18
What can go wrong in assessing the evidence about
effectiveness of screening?

• Two major biases affect these data
• lead time bias
• length bias

19
Lead time bias
without screening
with screening
average survival time post diagnosis has increased
But actual time of death remains the same!!
20
Lead time bias

• We think early detection has increased survival
• in fact all it has done is increase the time the
  patient is aware of his disease!
• treatment could even hasten death and it might
  appear survival is longer post diagnosis!!
• A great deal of cancer literature is susceptible
  to lead time bias
• Cannot just look at survival time post diagnosis.

21
Length bias
Survival due to screening and treatment may be
over rated because screening will tend to
discover more slow-growing disease.
22
Suppose there are two subtypes of the disease
Type 1 fast progression
Type 2 slow progression
23
Length of time in pre-clinical phase longer in
Type 2 than in Type 1
Type 1
Type 2
24
Periodic screening will tend to detect more of
Type 2, as these have longer exposure in the
critical interval for screening.
Type 1
Type 2
25
But look!! Type 2 individuals have a longer
survival time from time of diagnosis than do Type
1.
Type 1
Type 2
26

• Without screening, suppose type 1 and type 2 were
  equal fractions of the population
• average survival time is 5050 mixture of the
  short and long survival times.
• With screening, the screen-detected population
  has a higher fraction of type 2 (slow)
  individuals
• mix will be proportional to ratio of the two
  intervals
• suppose it is 7030 in favor of long interval
• average survival time will be longer in screen
  detected individuals!

27
Length bias

• Even if the treatment tended to be harmful and
  shorten life, because more longer interval
  individuals tend to be detected by screening, the
  screening program will appear to be effective!!

28

• Does not apply to all screening
• Assumes both subtypes of the disease is equally
  detectable in the pre-clinical phase (prostate
  cancer? breast cancer?)
• But if, say, more malignant disease is more
  detectable for physiologic reasons (bladder
  cancer, where more malignant cells may exfoliate
  faster?) then screening selectively finds more
  aggressive disease.

29

• Prostate cancer --
• clearly there is a range of more and less
  aggressive subtypes (see Albertsen et al, JAMA
  1998, for long-term survival in untreated
  prostate cancer by Gleason score).
• Breast cancer --
• Polun Chang dissertation 20-30 of
  screen-detected breast cancers have limited
  malignant potential

30
Length bias particularly worrisome

• Black Welch (Vol. 2 13, p 678)Advances in
  detection often go hand in hand with broadening
  of definition of disease (e.g., DCIS and breast
  cancer) thus we are technologically increasing
  the detection interval, but also increasing the
  defined interval from biologic onset without
  being sure that all that is detected is truly
  disease.
• Welch Black (Vol. 1, 32 (k), p 351) autopsy
  studies show huge reservoir of DCIS so perhaps
  most women have very slow breast cancer that
  will never bother them if undetected!

31

• A prospective RCT of screening with follow up to
  the critical endpoint may avoid length bias
• look at overall survival rates in both groups,
  which start out with same mix of disease types.
• But you cant look at survival in just
  screen-detected cases and compare to
  non-randomized population control.
• Probably affects disease-specific survival
  analyses too.Very tough problem for
  epidemiology and statistics!!

32
Elements of Screening Protocol Design

• Many, many factors go into the design of a
  screening intervention.
• Not all combinations of these will be evaluated
  empirically even though they affect the
  effectiveness of the program.

33
Some of the control knobs for a screening
program

• Which screening test is used
• Which follow up protocol is used for
  screen-positives
• Target population
• risk factors for selective screening
• setting in which they are screened (incidental,
  systematic, prompted, etc.)
• age start screening
• age end screening
• Screening interval

34
Each of these will affect the effectiveness and
the costs of screening

• Eddy (vol. 1, 30, p. 280) demonstrates the
  variation in costs and effectiveness of screening
  for cervical cancer as function of
• screening interval (1 yr, 3 yr, 5 yr)
• age to begin (20, 25, 30, 35)
• assumptions about natural history
• increasing sensitivity and/or specificity (and
  costs!) of the screening exam
• You should read this paper!

35

• When these parameters are changed, the
  cost-effectiveness changes generally in the
  direction you expect.
• But often the rate of change is different from
  expectations.
• Point is You have to do the calculations to be
  able to figure out the effects.

36
How are these calculations done?

• Often a computer simulation of a cohort of
  targeted individuals.
• Model the natural history of the disease
• Model the screening intervention
• Model the treatment
• Do all of this simulated year-by-year.
• Example Screening for Abdominal Aortic Aneurysm
  -- AAA

37
I just happen to have a spreadsheet for one
example...
Before showing the spreadsheet, lets look a bit
at the problem and the method...
38
Should we screen for AAA?

• AAA is clearly a progressive disease
• aneurysm starts small, then gradually expands
• Effective treatment
• elective surgical repair of AAA relatively safe
  and effective.
• Known risk groups
• age, sex, smoking status
• Good test available
• abdominal ultrasound is both relatively sensitive
  and specific