The economics of medical research: Public/private spillovers and the rate of return

Jonathan Grant, Jon Sussex, Yan Feng, Jorge Mestre-Ferrandiz(on behalf of Marco Hafner, Michele Pistollato and Peter Burridge)

Matttoprovideimage….

Outline of the presentation

Context and objectives

What we did

What we found

Policy implications

Estimating the economic returns research

Two types of economic returns

‘Spillover’ or GDP gain

Direct and indirect impact on the economy from medical research

Estimates are disease independent

Previously estimated GDP gains to be equivalent to ROI of 30% p.a. [range 26%-34%], based on review of the literature

Net health gain

Monetised health gains net of the health care costs of delivering them

Estimates are disease dependent: c9% for cardiovascular disease and c10% for cancer

Public/private spillovers

Four areas of further research highlighted:

Replication: repeat estimate of health gains for other clinical areas ü

Time lags: improve understanding and estimates for the time it takes for research to translate from bench to bedside ü

Attribution: improve understanding and estimates for attributing UK health gains to UK research

Spillovers: develop contemporary, UK specific, estimates for biomedical and health research

Outline of the presentation

Context and objectives

What we did

What we found

Policy implications

Main research question

What is the magnitude of the effect of government and charity biomedical and health research expenditure in the UK, separately and in total, on subsequent private pharmaceutical sector research and development (R&D) expenditure in the UK?

Jonathan Grant(Policy Institute, King’s College London)Peter Burridge(Department of Economics, University of York)Yan Feng(Office of Health Economics)Marco Hafner(RAND Europe)Jorge Mestre-Ferrandiz(Office of Health Economics)Michele Pistollato(formerly, Office of Health Economics)Jon Sussex(RAND Europe, formerly Office of Health Economics)

What we did

Reviewed theoretical, empirical and (UK) policy literature

Built 31-year time series of data for government, charitable and private sectors’ R&D spend

Modelled interaction between private R&D spend, government and charity research expenditure, plus control variables (global sales, dummy variables)

Data collection: Variables included

Data period: 1982 – 2012

Key variablesØ Government sector biomedical and health research expenditure in the UKØ Charitable sector biomedical and health research expenditure in the UK Ø Private pharmaceutical R&D expenditure in the UKØ Global pharmaceutical salesØ Dummy variables for 1993 and 1993 onwards

We combined the government sector expenditure and charitable sector expenditure as “public” expenditure for some analysis

Data collection: 10 disease areas

Time series broken down by 9 disease areas + “Other”:

BloodCancerCardiovascularCentral Nervous SystemGastroenterologyInfectionRespiratorySkinVision

Data collection – public medical/health research

GovernmentBuilt on 2008 What’s it worth? databaseReviewed annual reports and funders’ databases (e.g. MRC, NHS/DH, Research Councils)

CharityReviewed annual reports and funders’ databases (e.g. Wellcome Trust, Association of Medical Research Charities)AMRC allocated charities according to HRCS classification

Data collection – private industry R&D

Pharma R&D spend data available but not other medtech (pharma >95% of total UK biomedical industry R&D)

Measurement issue: there is no direct measure of private R&D expenditure by disease area

So we used a proxy: publications with authors giving UK industry addresses, by disease area

Literature suggests 3 year average lag from research activity to subsequent publication

We tried 0, 1, 2, 3, 4 and 5 year lags: most robust (“best”) model is when we assume a 4 year lag between R&D expenditure and subsequent publication

Biomedical and health care research expenditure 1982-2012

figures suggest that private R&D expenditure is subjectto more variation than public research expenditure.

UK R&D expenditures by disease areas (public and private)Figures 3 and 4 illustrate the public (government andcharity) and private R&D expenditure figures by diseasearea between 1982 and 2008 in the logarithmic form inwhich they feed into the econometric models. Note that

the underlying data is available in Additional file 6 (in-cluding public expenditure figures broken down by gov-ernment and charity).Similar to the aggregated data series reported in the

previous section, at the disease area level we observe, ineach case, an overall upward trend in public researchand private R&D spending and with more variation,expressed as upward and downward movements in the

2.8

3.2

3.6

4.0

4.4

82 84 86 88 90 92 94 96 98 00 02 04 06 08

BLOOD

4.0

4.4

4.8

5.2

5.6

82 84 86 88 90 92 94 96 98 00 02 04 06 08

CNS

4.8

5.2

5.6

6.0

6.4

6.8

82 84 86 88 90 92 94 96 98 00 02 04 06 08

CANCER

3.2

3.6

4.0

4.4

4.8

5.2

82 84 86 88 90 92 94 96 98 00 02 04 06 08

Cardiology

2.8

3.2

3.6

4.0

4.4

82 84 86 88 90 92 94 96 98 00 02 04 06 08

Gastroenterology

3.0

3.5

4.0

4.5

5.0

5.5

82 84 86 88 90 92 94 96 98 00 02 04 06 08

Infectious

2.4

2.6

2.8

3.0

3.2

3.4

3.6

82 84 86 88 90 92 94 96 98 00 02 04 06 08

Skin

2.6

2.8

3.0

3.2

3.4

3.6

3.8

82 84 86 88 90 92 94 96 98 00 02 04 06 08

VISION

2.8

3.2

3.6

4.0

4.4

82 84 86 88 90 92 94 96 98 00 02 04 06 08

Respiratory

6.8

7.0

7.2

7.4

7.6

82 84 86 88 90 92 94 96 98 00 02 04 06 08

Other

LNPUBLIC

Fig. 3 Public (government and charity) research and development (log) expenditure by disease area, 1982–2008 (£m, 2012 constant prices)

Fig. 2 Total UK research and development expenditure (government, charity and private), 1982–2012 (£m, 2012 constant prices)

Sussex et al. BMC Medicine (2016) 14:32 Page 11 of 23

Distribution of ln(public research)

figures suggest that private R&D expenditure is subjectto more variation than public research expenditure.

UK R&D expenditures by disease areas (public and private)Figures 3 and 4 illustrate the public (government andcharity) and private R&D expenditure figures by diseasearea between 1982 and 2008 in the logarithmic form inwhich they feed into the econometric models. Note that

the underlying data is available in Additional file 6 (in-cluding public expenditure figures broken down by gov-ernment and charity).Similar to the aggregated data series reported in the

previous section, at the disease area level we observe, ineach case, an overall upward trend in public researchand private R&D spending and with more variation,expressed as upward and downward movements in the

2.8

3.2

3.6

4.0

4.4

82 84 86 88 90 92 94 96 98 00 02 04 06 08

BLOOD

4.0

4.4

4.8

5.2

5.6

82 84 86 88 90 92 94 96 98 00 02 04 06 08

CNS

4.8

5.2

5.6

6.0

6.4

6.8

82 84 86 88 90 92 94 96 98 00 02 04 06 08

CANCER

3.2

3.6

4.0

4.4

4.8

5.2

82 84 86 88 90 92 94 96 98 00 02 04 06 08

Cardiology

2.8

3.2

3.6

4.0

4.4

82 84 86 88 90 92 94 96 98 00 02 04 06 08

Gastroenterology

3.0

3.5

4.0

4.5

5.0

5.5

82 84 86 88 90 92 94 96 98 00 02 04 06 08

Infectious

2.4

2.6

2.8

3.0

3.2

3.4

3.6

82 84 86 88 90 92 94 96 98 00 02 04 06 08

Skin

2.6

2.8

3.0

3.2

3.4

3.6

3.8

82 84 86 88 90 92 94 96 98 00 02 04 06 08

VISION

2.8

3.2

3.6

4.0

4.4

82 84 86 88 90 92 94 96 98 00 02 04 06 08

Respiratory

6.8

7.0

7.2

7.4

7.6

82 84 86 88 90 92 94 96 98 00 02 04 06 08

Other

LNPUBLIC

Fig. 3 Public (government and charity) research and development (log) expenditure by disease area, 1982–2008 (£m, 2012 constant prices)

Fig. 2 Total UK research and development expenditure (government, charity and private), 1982–2012 (£m, 2012 constant prices)

Sussex et al. BMC Medicine (2016) 14:32 Page 11 of 23

Distributions of ln(private R&D)

time series, in the expenditure data for the private sectorthan for the public sector.Looking at specific disease areas in more depth, we

further observe, for instance, that for ‘Blood’ publicR&D expenditure is for almost all years over the obser-vation period smaller than private expenditure (except inyear 1983, where there is a fall in private expenditure,followed by a strong upward trend). In ‘CNS’ private ex-penditure there is an upward trend since 1982 (withshort interruptions in 1984 and 1993), which flattensand somewhat reverses with the start of year 1997. Incontrast, the trend for public ‘CNS’ expenditure is some-what flat between 1982 and 1990 and subsequently risessteeply until around 2006. For all years, private R&D ex-penditure exceeds public R&D in the disease area ‘CNS’.The disease area ‘Cancer’ represents one of the largestareas of biomedical research funding, both for the publicand the private sector, and for the majority of years overthe observation period, public expenditure related tocancer research exceed private expenditure. The publiccancer research series further follows a relatively steadyupward trend since 1982, with two smaller interruptionsin 1990 and 2004. The private expenditure series for‘Cancer’ shows more variation over time with two largerinterruptions in 1990 and 1995. In ‘Cardiovascular’, theprivate expenditure series follows no clear trend up until1996, followed by a small upward trend thereafter butwith rather strong variation between 1982 and 2008,with the lowest level of private funding in the area rea-lised in 1992. In contrast, the ‘Cardiovascular’ public

R&D expenditure series follows a slightly increasingtrend between 1982, interrupted in 1996 and followedby a steady rise until 2008. For the majority of the yearsin our observation period, private expenditure exceedspublic expenditure in cardiovascular research. ‘Gastro-enterology’ is an interesting area insofar as public ex-penditure in the field follows a downward trend untilabout 1992, increases somewhat thereafter and increas-ing strongly since 1999. Furthermore, the private ex-penditure series for ‘Gastroenterology’ shows an upwardtrend overall with some variation over time (for instance,1986 and between 1993 and 1996). Similar to the ‘CNS’expenditure series, private expenditure on gastroenter-ology research exceeds public expenditure for all years.For the disease area ‘Infectious Diseases’, the public andprivate R&D expenditure series differ with regard to thetrend pattern. Whereas the public expenditure seriesshows a steady increasing trend since 1982 (with smallinterruptions in 1984, 1990 and 2005), the private ex-penditure series shows a lot more variation and a fairlyflat trend after 1994. However, for the majority of theyears, private R&D still exceeds public research expend-iture in the ‘Infectious Diseases’ area. The two smallestdisease areas we look at in more depth are ‘Skin’ and ‘Vi-sion’, which both follow similar trends over time. Publicexpenditure in the two areas is relatively flat until about1993 and subsequently takes off. In both areas, publicexpenditure levels are very similar over the observationperiod. This is different in the private expenditure series,where private spending in ‘Skin’ generally exceeds

2.8

3.2

3.6

4.0

4.4

4.8

5.2

82 84 86 88 90 92 94 96 98 00 02 04 06 08

BLOOD

4.5

5.0

5.5

6.0

6.5

7.0

82 84 86 88 90 92 94 96 98 00 02 04 06 08

CNS

3

4

5

6

7

82 84 86 88 90 92 94 96 98 00 02 04 06 08

CANCER

3.5

4.0

4.5

5.0

5.5

82 84 86 88 90 92 94 96 98 00 02 04 06 08

Cardiology

3.0

3.5

4.0

4.5

5.0

5.5

82 84 86 88 90 92 94 96 98 00 02 04 06 08

Gastroenterology

3.2

3.6

4.0

4.4

4.8

5.2

5.6

82 84 86 88 90 92 94 96 98 00 02 04 06 08

Infectious

2.5

3.0

3.5

4.0

4.5

5.0

82 84 86 88 90 92 94 96 98 00 02 04 06 08

Skin

1

2

3

4

5

82 84 86 88 90 92 94 96 98 00 02 04 06 08

VISION

2

3

4

5

6

7

82 84 86 88 90 92 94 96 98 00 02 04 06 08

Respiratory

6.0

6.4

6.8

7.2

7.6

8.0

82 84 86 88 90 92 94 96 98 00 02 04 06 08

Other

LNPRIVATE

Fig. 4 Private research and development (log) expenditure by disease area, 1982–2008 (£m, 2012 constant prices)

Sussex et al. BMC Medicine (2016) 14:32 Page 12 of 23

Distributions of ln(global pharma sales)

funding in ‘Vision’, but both private series follow a some-what upward trend with interruptions. Although privateexpenditure exceeds public expenditure for all years in‘Skin’, in ‘Vision’, for the majority of years, public spendexceeds private spend. In the disease area ‘Respiratory’we observe a relative flat trend in public research, withvariation, until around 1997, when a general upwardtrend in public expenditure in the area kicks in. For theprivate expenditure series in the respiratory area we ob-serve a relatively strong downturn from 1982 to 1985,followed by an upward trend thereafter. What is more,for the majority of the years, in ‘Respiratory’ private ex-penditure exceeds public expenditure (except 1984 and1985).

Global pharmaceutical salesFigure 5 illustrates global pharmaceutical sales (in £m, in2012 price terms) in logarithmic form by disease areafrom 1982. Overall, the expenditure patterns in the fig-ure reveal an upward trend in global medicine sales inall the disease areas. However, looking at specific diseaseareas we observe some variation. For instance in ‘Blood’,there is a decrease in sales from 1988 to 1989, followedby a steady upward trend thereafter. The ‘Cancer’medicines global sales series somewhat interrupts in1993, with a decrease in sales between 1993 and 1994,but is followed by a steady increase thereafter. Interest-ingly, the global pharmaceutical sales series shows a par-ticularly strong rise in sales starting in 1999 in mostdisease areas.

Econometric modellingOverall, our results suggest that there is a statisticallysignificant complementary relationship between publicbiomedical and health research expenditure and privatepharmaceutical R&D expenditure. A 1 % increase inpublic sector expenditure is associated in the best-fitmodel with a 0.81 % increase in private sector expend-iture. The sensitivity analysis, with dummy variables for1993 and subsequently, produces a similar and statisti-cally significant result but with a slightly smaller positiveelasticity of 0.68.

VariablesThe ADF unit root test results are reported in Table Ain Additional file 7. The null hypothesis for the ADFunit root test is that the individual series is a unit rootprocess. The results suggest that the five variables arenon-stationary in levels or logs, but stationary in firstdifferences of levels or logs; we treat the contrary resultfor log(private) as an artefact of the pooling of individualseries outcomes and discount it.

Determining the best-fit modelResults from the 12 tested models are reported in TablesB–D in Additional file 7. Each table reports performanceof four specifications, one for each deterministic trendspecification. In this preliminary search the time-lag be-tween funding and publication is treated as zero.For each specification of the model we report six sta-

tistics: the cointegration rank, the statistics from the

8.0

8.5

9.0

9.5

10.0

10.5

82 84 86 88 90 92 94 96 98 00 02 04 06 08

BLOOD

8.5

9.0

9.5

10.0

10.5

11.0

82 84 86 88 90 92 94 96 98 00 02 04 06 08

CNS

7

8

9

10

11

82 84 86 88 90 92 94 96 98 00 02 04 06 08

CANCER

9.2

9.6

10.0

10.4

10.8

11.2

82 84 86 88 90 92 94 96 98 00 02 04 06 08

Cardiology

9.2

9.6

10.0

10.4

10.8

82 84 86 88 90 92 94 96 98 00 02 04 06 08

Gastroenterology

9.2

9.6

10.0

10.4

10.8

11.2

82 84 86 88 90 92 94 96 98 00 02 04 06 08

Infectious

8.2

8.4

8.6

8.8

9.0

9.2

9.4

82 84 86 88 90 92 94 96 98 00 02 04 06 08

Skin

7.0

7.5

8.0

8.5

9.0

9.5

82 84 86 88 90 92 94 96 98 00 02 04 06 08

VISION

8.5

9.0

9.5

10.0

10.5

82 84 86 88 90 92 94 96 98 00 02 04 06 08

Respiratory

9.5

10.0

10.5

11.0

11.5

12.0

82 84 86 88 90 92 94 96 98 00 02 04 06 08

Other

LNSALE

Fig. 5 Global pharmaceutical (log) sales by disease area 1982–2008 (£m, 2012 constant prices)

Sussex et al. BMC Medicine (2016) 14:32 Page 13 of 23

Modelling aims

Is there a long run equilibrium relationship between the level of:Ø Government and charity biomedical and health research expenditure in the UK; andØ Private sector pharmaceutical R&D expenditure in the UK?

Are these complements or substitutes in the long run?

What is the long-run elasticity of private spend with respect to public spend?

Modelling methods

Vector Error Correction Model (VECM) Various specifications

˗ Check the number of long run equilibrium conditions ˗ Experiment with different specifications of the deterministic trend˗ Experiment with different numbers of lags˗ Check for the absence of residual autocorrelation up to 6 lags

Comparing the performance of the specifications ˗ AIC, Schwarz Criterion and log likelihood statistics ˗ Number of insignificant coefficients in the Error Correction model

Trend model selection and determination of cointegrating rank follows the parsimony principle advanced by Pantula

Heterogeneity between different disease areas?˗ We tried various approaches to handling heterogeneity explicitly˗ The results were unstable and hard to interpret, so we have not differentiated between

diseases areas

Outline of the presentation

Context and objectives

What we did

What we found

Policy implications

What we found (1)

Long term equilibrium relationship between UK ‘public’ biomedical and health research expenditure, private sector R&D expenditure in the UK and global sales

Complementary relationship between ‘public’ and private R&D expenditure, that is statistically significant at the 5% level in all specifications

1% increase in ‘public’ expenditure on R&D will eventually lead to between 0.38% and 1.12% increase in private expenditure on R&D. Best model suggests an elasticity of 0.81

What we found (2)

We analysed the impulse response functions based on the best model

44% of the response occurs after one year, but the full response takes many years

Outline of the presentation

Context and objectives

What we did

What we found

Policy implications

So what?

Crowding in, not crowding out. We have confirmed that government and charity medical research spend stimulates additional private pharmaceutical industry R&D in the UK

Given that in 2012 government+charity and private research spend were respectively £3.43bn and £4.21bn, the elasticity of 0.81 means that a £1 increase in government+charity research spend produces an eventual £0.99 increase in private pharmaceutical industry R&D in the UK

Implications for the rate of return to public medical research

Substituting into the “What’s it worth?” RoR calculation, other parameters unchanged, implies economic RoR in UK to UK public (govt+charity) medical research is in the range 15%-18% (vs. 30%, range 26%-34%, for original estimates)

Lower UK value than US could be due to smaller size and greater openness of UK economy to the rest of the world

Added to estimates of the net monetary benefit of health gains arising from cardiovascular research (9%) and cancer research (10%), total return between 24% and 28%

UK-specific estimate still shows that investment in medical research gives a very good rate of return

What next ….

Use current dataset to:Ø Look at relationship between government

and charityØ Look at spillovers from private to public

Update social rate of return to R&D figures (old, international and non-biomedical/health)

Qualitatively understand how spillovers actually manifest themselves and what are the effective policy levers