Clinical Epidemiology: Thyroid disease and test results

1
Clinical EpidemiologyThyroid disease and test
results

• Wiley D. Jenkins, PhD, MPH
• Research Assistant Professor
• Southern Illinois University School of Medicine
• Department of Family and Community Medicine

2
Who I am

• My name is Wiley D. Jenkins and I am currently
  Research Assistant Professor at the SIU-SOM
  Department of Family and Community Medicine.
  Prior to this I spent 13 years in the state
  health department laboratory.
• I received my MPH-Epidemiology from Tulane
  University in 2002. This was followed by my PhD
  in Health policy from the University of Illinois
  at Chicago in 2007.
• Much of my research and work experience has
  concerned laboratory testing, STDs and the
  quality of laboratory data.

3
Learning objectives

• To understand the concepts of test sensitivity,
  specificity, positive predictive value and
  negative predictive value.
• To understand how these factors effect the
  utility of individual tests when diagnosing a
  condition.
• To understand how these factors are manipulated
  by targeting screening tests to specific
  populations.

4
Performance objectives

• To be able to calculate the sensitivity,
  specificity, positive predictive value and
  negative predictive value for a given test.
• To be able to determine if a tests result is
  useful given its calculated values.
• To be able to show how screening guidelines
  should be adjusted to increase positive and
  negative predictive values to maximize result
  usefulness.

5
There is always uncertainty

• Our common language incorporates uncertainty.
• Usually implies error bars
• Physics tells us that in an infinite universe,
  anything is possible. Some things are just more
  or less likely.
• Heisenberg uncertainty principle
• statement that locating a particle in a small
  region of space makes the momentum of the
  particle uncertain and conversely, that
  measuring the momentum of a particle precisely
  makes the position uncertain
• As a matter of practicality, some things are
  essentially 100 or always something.
  HOWEVER, its important to know when this is not
  the case, and that is not always obvious.

6
Quick review of terms

• Sensitivity the ability of a test to correctly
  identify those who have a condition
• Specificity the ability of a test to correctly
  identify those who do not have a condition
• Positive predictive value the number of
  individuals who have a condition from all those
  who test positive
• Negative predictive value - the number of
  individuals who do not have a condition from all
  those who test negative

7
The 2 x 2 table

• Youll use this a lot later in life

8
Sensitivity

• 90 sensitivity implies that of all those who
  have the disease, 10 will not be identified by
  the test. If prevalence is 20 of the population

9
Specificity

• 75 specificity implies that of all those who do
  not have the disease, 25 will not be identified
  by the test. If prevalence is 20 of the
  population

10
Positive/negative predictive value

• We complete the remaining marginals and find
• PPV for our example test is 180/380 47
• NPV is 600/620 97.
• What do we draw from this about the usefulness of
  the test?

11
Time for a clinical example

• 27-year-old woman
• 10 lb weight loss in past two months, not trying
• Some difficulty sleeping
• Never had anything like this before
• No signs/symptoms of depression
• Meds Oral contraceptive pills
• 1-cm, firm, smooth nodule in right lobe of
  thyroid
• BMI 20
• Skin slightly dry
• Remainder of physical examination normal
• What do you think?
• What should we do?

12
Lab tests and results
13
What next?

• Order more tests?
• Schedule for surgery?
• Prescribe medication, therapy, hamburgers?
• 1st, lets see what the tests are really telling
  us.

14
Thyroid stimulating hormone

• Our patient has a (low) normal TSH
• Sensitivity 92
• Specificity 94
• Are these good values?
• Assume prevalence for thyroid disease of 4 in
  large populations
• Calculate PPV and NPV for TSH
• Do we care more about the PPV or NPV for this
  scenario?

15
TSH 2 x 2 table

• Complete the table and calculate the PPV and NPV
  assuming sens 92, spec 94 and prevalence
  4

16
TSH 2 x 2 table - completed

• We find
• PPV 37/95 31
• NPV 902/905 100
• Which do we care about and what conclusions do we
  draw?

17
Free T4

• Our patient has an elevated Free T4
• Sensitivity 82
• Specificity 94
• Assume prevalence for thyroid disease of 4 in
  large populations
• Calculate PPV and NPV for Free T4
• Do we care more about the PPV or NPV for this
  scenario?

18
Free T4 table

• Complete the table and calculate the PPV and NPV
  assuming sens 82, spec 94 and prevalence
  4

19
Free T4 table - completed

• We find
• PPV 33/91 36
• NPV 902/909 99
• Which do we care about and what conclusions do we
  draw?

20
So

• We have
• A symptomatic woman on OCPs with a thyroid nodule
• A normal TSH
• An elevated Total T4
• An elevated Free T4
• What next?
• Scintigraphy?
• Fine Needle Aspiration Biopsy?
• Excisional Biopsy?

21
Fine needle aspiration biopsy

• Indeterminate result
• 15-20 false positive rate (assume 20 for
  calculations to follow)
• 3 false negative rate
• If we assume a 4 prevalence of thyroid cancer,
  calculate the sensitivity and specificity of the
  biopsy.
• Calculate the positive and negative predictive
  value.

22
The FNAB 2 x 2 table

• What do we know?
• Prevalence 4
• False positive rate 20
• False negative rate 3

23
The FNAB 2 x 2 table

• False positives FP rate x all negatives 0.20
  x 960 192
• False negatives FN rate x all positives .03 x
  40 1

24
The FNAB 2 x 2 table - completed

• We find
• PPV 39/231 17
• NPV 768/769 100
• Which do we care about and what conclusions do we
  draw?

25
Clinical course

• The patient was referred to a surgeon for
  excisional biopsy.
• Nodule was removed, was a benign colloid goiter,
  no malignancy and no evidence of Hashimotos or
  other disease.

26
Lab results
27
How do laboratory tests contribute to medical
errors?

• Are not always right
• May result in unnecessary further testing
• May result in unnecessary surgery
• With attendant complications
• If we assume that tests are correct 95 of the
  time, what is the likelihood that, in a battery
  of 20 tests, one will be a false result?
• So, for every Chem 20 you order (or other battery
  of 20 tests), 1 will be either a FALSE POSITIVE
  or a FALSE NEGATIVE.
• Need to know how to work with sensitivity and
  specificity in order to know what to believe.

28
Time for a population example

• Why, because we like you! (M I C)
• Seriously though, population-level studies are
  translated into clinical guidelines.
• In 2006, the number of reported cases of
  Chlamydia trachomatis (Ct) in the US exceeded
  1,000,000 for the 1st time.
• The great majority of cases (70 in women) are
  entirely asymptomatic.
• Upwards of 40 of untreated Ct progress to PID
  followed by chronic pelvic pain, ectopic
  pregnancy and infertility.
• How do we address this?

29
Chlamydia trachomatis screening

• Diagnostic companies have spent considerable
  money developing rapid and accurate tests for the
  detection of Ct.
• Current tests offer
• 95 sensitivity
• 98 specificity
• So, do we just test everyone? Lets see.
  (150,000,000 women) x (10/test) need for
  other alternative.
• Who has Ct?
• 0.35 all Americans
• 0.52 women
• 0.17 men
• 1.76 Black women
• 0.24 White women
• 2.9 women aged 15-19
• 2.8 women aged 20-24

30
The Ct 2 x 2 table - completed

• For the general population (0.35) we find
• PPV 33/233 14
• NPV 9765/9767 100

31
The Ct 2 x 2 table - completed

• For all women (0.52) we find
• PPV 49/248 20
• NPV 9749/9752 100

32
The Ct 2 x 2 table - completed

• For all women aged 16-24 (2.9) we find
• PPV 276/470 59
• NPV 9516/9530 100

33
Utility of targeted testing

• By purposefully targeting our testing to at-risk
  populations, we increase the PPV of the test and
  better allocate resources.
• General population
• Prevalence 0.35 PPV 14
• All women
• Prevalence 0.52 PPV 20
• Women aged 16-24
• Prevalence 2.9 PPV 59
• Females admitted into juvenile detention
  centers??
• Prevalence 12-20 PPV gt90!
• Other risk factors important.
• This works for clinical guidelines for screening,
  such as mammography, prostate exams, cholesterol

34
Take away items

• Not a good practice to order tests just because
  we can or for fishing expeditions.
• Costs can quickly become quite significant (e.g.
  compare HC expenditure for US versus other
  industrialized countries and resultant health
  outcomes).
• Utility of the results is directly impacted by
  the population/person to which they are given.
• Multiple tests increase the likelihood of a
  correct diagnosis.
• E.g. Ct in 16-24, PPV 59
• Additional test on just these positives (e.g. 59
  prevalence) with same sens/spec results in PPV of
  99!
• In the absence (always) of the ultimate test,
  use multiple results to arrive at the best
  conclusion.

35
Questions or comments?? Contact info Wiley D.
Jenkins, PhD, MPH wjenkins_at_siumed.edu 217-545-8717