1) The document proposes using value of information (VOI) analysis to calculate the value of transferability (applying results from a multinational trial to a specific country) in terms of expected value of sample information (EVSI). 2) It applies this method to results from the Scandinavian Simvastatin Survival Study, defining the posterior distribution of country results given the multinational trial results. 3) The value of transferability is measured as the difference between the prior and posterior expected value of perfect information (EVPI) for each country, indicating whether applying the multinational results reduces uncertainty.

1. Value of Transferability and
Efficiency in HTA
Bernarda Zamora, Grace Marsden, Adrian Towse
Email: bzamora@ohe.org
1. BACKGROUND
• Existing literature on transferability advises that multinational
trials should report country-specific cost-effectiveness results,
yet typically these country specific results are only used to
assess the between-location variability.
• We use value of information (VOI) analysis (assuming a normal
distribution) to show that country specific results can be used to
calculate the value of transferability (transfer from the wide trial
to the country of interest) in terms of the Expected Value of
Sample Information (EVSI).
• Our contribution lies in the definition of the posterior distribution
of the country results given the wide trial results.
Acknowledgements
This presentation has benefitted from comments at the OHE Research
team meeting. Special thanks to Ed Wilson.
References
Briggs, Andrew, Mark Sculpher, and Karl Claxton, 2006. Decision modelling for health economic evaluation, Oxford University Press.
Burton, P. R., 1994. Helping doctors to draw appropriate inferences from the analysis of medical studies, Stat Med 13(17), pp. 1699-1713.
Claxton, K. and K. M. Thompson, 2001. A dynamic programming approach to the efficient design of clinical trials, J Health Econ 20(5), pp. 797-822.
Cook, J. R., M. Drummond, H. Glick, and J. F. Heyse, 2003. Assessing the appropriateness of combining economic data from multinational clinical trials,
Stat Med 22(12), pp. 1955-1976.
Gelman, Andrew, John B Carlin, Hal S Stern, and Donald B Rubin, 2014. Bayesian data analysis, Taylor & Francis.
Manca, A., P. C. Lambert, M. Sculpher, and N. Rice, 2007. Cost-effectiveness analysis using data from multinational trials: the use of bivariate hierarchical
modeling, Med Decis Making 27(4), pp. 471-490
Spiegelhalter, David J, Keith R Abrams, and Jonathan P Myles, 2004. Bayesian approaches to clinical trials and health-care evaluation, John Wiley & Sons.
Wilson, E. C., 2015. A practical guide to value of information analysis, Pharmacoeconomics 33(2), pp. 105-121.
2. AIMS
• To calculate the value of transferability (transfer from the wide
trial to the country of interest) in terms of the Expected Value of
Sample Information.
• To test this method on the results of the Scandinavian
Simvastatin Survival Study published by Cook et al. (2003).
3. METHODS
1. Posterior distribution of country n INMB: Bayesian Conjugate
Normal Analysis & Sequential use of Bayes theorem:
If the vector 𝑦 𝑛, 𝑦 𝑚 denotes the INMB of the individual country n,
and the wide trial m
𝑝 𝜃 𝑦 𝑛, 𝑦 𝑚 = 𝑁 𝜃𝑛
𝑦 𝑚
𝜎 𝑚
2
1 − 𝜌2 +
𝑦 𝑛
𝜎 𝑛
2
1
𝜎 𝑚
2
1 − 𝜌2 +
1
𝜎 𝑛
2
,
1
1
𝜎 𝑚
2
1 − 𝜌2 +
1
𝜎 𝑛
2
2. Value of Transferability in terms of Value of Information:
Prior EVPI: associated with the normal marginal distribution of
each country result.
Posterior EVPI: associated with the distribution obtained above
Value of Transferability = EVSI = Prior EVPI – Posterior EVPI
4.RESULTS
Figure 1: Distribution of Net Benefit
5. DISCUSSION
• It has been argued that in absence of significant between-
country differences in the parameters of interest, the wide trial
pooled results should be adopted by each country as the
“generalisable/transferable” results.
• Building on Bayesian statistical properties and VOI, it is possible
to further assess the appropriateness of the transferability of
wide trial results by valuing the benefits from reducing
uncertainty, or the cost from increasing uncertainty.
• Our method relies on conjugate normality. Although this is a
restrictive assumption, it allows us to fully define the posterior
distribution from published results on just CE means and
variances of countries and wide trial.
• Even though Cook et al.’s (2003) results imply that it would be
appropriate for each country to reimburse Simvastatin according
to the wide trial CE pooled results, we show that the
transferability of the wide trial decision is not appropriate for
Finland and Iceland.
• We measure the value of transferability in terms of monetary
benefits or costs.
.6.7.8.9
1
-1 -.5 0 .5 1
rho
Denmark
.6.7.8.9
1
-1 -.5 0 .5 1
rho
Sweden
.75
.8
.85
.9
.95
1
-1 -.5 0 .5 1
rho
Norway
0
.1.2.3.4
-1 -.5 0 .5 1
rho
Finland
0
.2.4.6
-1 -.5 0 .5 1
rho
Iceland
Implicit rho from bivariate normality
Posterior decision of
reimbursement:
Yes for all countries,
given the trial-wide
results
Denmark Finland Iceland Norway Sweden
Prior EVPI $43,555,214 $31,219,711 $4,900,946 $16,535,958 $49,568,440
ΔBj¦ ΔB.
Conditional mean INMB
$525 $525 $525 $525 $525
SE(ΔBj/ ΔB.) :
Conditional standard
error of mean INMB
$363.97 $372.68 $689.29 $0 $330.00
Posterior mean $547.40 $549.20 $478.96 $525 $526.14
Posterior standard error $357.88 $358.37 $675.57 $0 $316.22
Posterior absolute
normalised mean
1.53 1.53 0.71 - 1.66
Posterior probability of
wrong decision
6.31% 93.73% 23.92% 0% 4.81%
L(ΔB0,SE(ΔB)0) :unit
normal loss
0.027 1.444 0.141 0.000 0.020
Posterior EVPI $1,320,739 $86,112,293 $438,135 $0 $1,409,515
Value of Transferability:
Prior EVPI-Posterior
EVPI
$42,234,475 -$54,892,581 NA $16,535,958 $48,158,925
-10,000 -5,000 0 5,000 10,000
Finland Iceland
WideTrial Sweden
Denmark Norway
Figure 2: Correlation with Multinational results
6. CONCLUSION
Table 1: Posterior distribution of INMB and Value of Transferability