Title: Surrogate outcomes in liver diseases: Does making a test better make a patient better
1Surrogate outcomes in liver diseases Does
making a test better make a patient better?
- Ronald L. Koretz, M.D.
- Emeritus Professor of Clinical Medicine
- David Geffen-UCLA School of Medicine
2Outcomes of trials
- Effect on mortality1
- Effect on morbidity1
- Estimate of resource utilization needed2
- Effect on tests3
- 1 Primary concern of patient
- 2 Primary concern of payer
- 3 Only concern of patient/payer if it reflects
other outcomes
3Proving mortality effect in liver disease
- When is death an issue in liver disease?
- Fulminant hepatitis
- Decompensated cirrhosis
- Time periods
- Fulminant hepatitis days/weeks
- Decompensated cirrhosis months/years after
cirrhosis develops - Progression to cirrhosis - decades
4Proving morbidity effect in liver disease
- Time course parallels mortality considerations
5Summary of CHBG Reviews(March 2009)
- Surgical techniques1 23
- Treatment acute/acute on chronic 28
- Preventing complications cirrhosis 7
- Preventing cirrhosis 32
- Other topics2 8
- 1 Includes endoscopic/transplant issues
- 2 Immunization/biliary/screening
6How to do a study that requires decades
- Be very patient and have understanding promotion
committee/funding source - Multicenter, multiple investigators
- Find outcome that requires less time
7Outcome requiring less time
- Disease Outcome
- PBC Biochemical markers
- NASH Enzyme levels
- Hepatic fat content
- Hepatitis B/C Serologic markers
- Hemochromatosis Body iron content
8Surrogate outcome1
- Definition Predicts a clinical outcome
- Advantages
- Occurs earlier
- Easier to measure
- Disadvantage Needs validation
- 1 Typically a test
9Association versus causation
- I wish that they would stop turning on that
Fasten Seat Belt sign. Every time that they
do, the ride gets bumpy.
10Criteria for good surrogate1
- Strong and consistent correlation between
surrogate and clinical outcome - Randomized trials of intervention(s) show
consistent improvement in both surrogate and
clinical outcomes2 - 1 Guyatt et al, Assessing a surrogate outcome
- 2 If no difference in both, cannot tell
11Validation
- Easy to show that a surrogate factor associated
with clinical outcome1 - Harder to show that changing the surrogate factor
changes the clinical outcome2 - 1 Association can be demonstrated with
observational study - 2 Requires randomized trial with both outcomes
measured
12When a surrogate got it wrong
- Encainide/Flecainide/Moricizine reduced
ventricular arrhythmias released by FDA on that
basis - CAST1 study RCT comparing anti-arrhythmia
therapy to placebo after myocardial infarction - 1 Cardiac Arrhythmia Suppression Trial
13CAST outcomes
- Parameter Anti-arrhythmic Placebo
- Number pts 1380 1371
- 1-year mortality 10 5
- Arrest/arrhythmic 7 4
- death
- p 0.0006
- p 0.003
14Another surrogate that wasnt
- Parenteral/enteral nutrition to treat or prevent
morbidity associated with malnutrition
15Studley, 1936
- Patients with gt20 body weight loss had poorer
postoperative survival1 - 1 Association subsequently demonstrated many
times in surgical and non-surgical disorders
16AssumptionMalnutrition causes problems
- Consequent assumption
- Improving nutritional status will improve
outcome - Consequent action
- Focus on nutrition markers
17Necessary condition for accepting a surrogate
end-point as a substitute for a clinical one
- Clinical and surrogate end-points must be
parallel (concordant)
18Concordance vs Discordance
- Clin EP
- Rx gt Cont Rx Cont Rx lt Cont
- Rx gt Cont C PD CD
- Nutr better Nutr better
- Nutr
- Rx Cont PD C PD
- Clin better Nutr better
- EP
- Rx lt Cont CD PD C
- Clin better Clin better
- Clin Clinical Nutr Nutritional EP
End-point - Rx Treatment group Cont Control group
- gt Better than lt Worse than
- C Concordance PD Partial Discordance CD
Complete discordance
19Clinical End-points
- Mortality
- Infectious complications
- Total complications (including infectious)
- Duration of hospitalization
20Nutritional End-points
- Protein/calorie intake
- Body weight
- Nitrogen balance
- Anthropometrics (TSF, MAC, MAMC)
- Visceral proteins (albumin, prealb, transferrin)
- Skin testing
- Total lymphocyte count
- TSF Triceps skinfold thickness MAC Midarm
circumference MAMC Midarm muscle
circumference prealb prealbumin
21Summary of concordance1
- Rate of concordance lt 50 in 41/48 comparisons2
- Rates of complete discordance
- gt 25 in 43/48 comparisons
- gt 50 in 13/48 comparisons
- 1 Data from 99 RCTs
- 2 Comparing particular nutritional/clinical
outcome
22Concordance in trials of liver disease
- Same observation discordance was rule
23Nutritional outcome as surrogate
- Improving nutritional end-points does not
guarantee that clinical end-points will improve
24Do surrogates ever work?
- HIV/AIDS viral titer, CD4 count
- 5 FU therapy for colon cancer 3 year
disease-free survival1 - Severe hypercholesterolemia serum cholesterol
levels - 1 Compared to 5-year total survival
25Surrogates in liver disease
- Bilirubin in PBC
- HCV-RNA in hepatitis C
26Surrogates in primary biliary cirrhosis
- Alkaline phosphatase (enzyme not clinical)
- Bilirubin (function test possibly clinical)
27PBC Bilirubin vs mortality1
- Outcome Effect 95 CI
- Mortality 1.17 (RR) 0.75, 1.82
- Bilirubin -10 (WMD) -16, -5
- 1 Yan Gong, DDW, 2006 6 trials ursodeoxycholic
acid - 2 micromoles/l
28SVR and Hepatitis C
- Objective of treatment Prevent
mortality/morbidity - SVR is surrogate outcome that does not directly
measure mortality or morbidity
29Natural history of hepatitis C1
- Vast majority of patients (at least 80-90) who
are infected by hepatitis C will never develop
end stage liver disease - 1 From prospective cohort studies and
epidemiologic data
30What does 50 SVR rate imply?
- Prognostic factors for SVR1
- Little or no fibrosis on biopsy
- Female
- Normal weight
- Cannot assume 50 reduction in subsequent
incidence of adverse events - 1 Factors suggesting lower likelihood to progress
31All Viral Hepatitis Patients
Treatment Responders 50
Cirrhosis 10-20
32SVR vs Cure
- Case reports of responders who develop
cancer/decompensation
33Need RCT to prove efficacy
- Problem
- Take decades to see decompensation
- Need large number in trial
- Compromise study those with severe fibrosis
34Interferon in cirrhosisHALT-C, NEJM, 2008
- IFN1 Control
- Number patients 517 533
- Death 36 (7) 26 (5)
- Decompensated 72 (14) 69 (13)
- Number HCCs 15 (3) 15 (3)
- Surrogates2 Better
- P lt 0.05
- 1 Long-term use
- 2 ALT, histologic inflammation, SVRs
35Discordance in HALT-C
- HALT-C design did not test routine use of
antiviral therapy - Nonetheless, see discordance between surrogate
and clinical outcomes - We have to consider possibility that SVR is
inadequate measure of treatment efficacy
36In order for a difference to be a difference, it
must make a difference
- Courtesy of John Rombeau, M.D.