The impact of external reference pricing on pharmaceutical costs and market dynamics

Highlights

• •Growth in the cost of prescription drugs has generated interest in external reference pricing (ERP).
• •Use of ERP for price negotiation is widespread internationally.
• •We study countries that do and don’t employ ERP, focusing on high-cost drugs.
• •ERP associated with price reductions over time but also delayed product launches.
• •ERP approaches don’t appear to materially affect the initial price set for a drug.

Abstract

Growth in the cost of prescription drugs in the US has generated significant interest in the use of external reference pricing (ERP) to tie prices paid for drugs to those in other countries. We used data from the Pricentric ONE™ database, an international drug pricing database, to examine product launch timing, launch price, and price changes from January 2010 – October 2021 in both ERP and non-ERP settings, with a focus on 100 high-priced drugs of interest to Medicare and Medicaid. We found that ERP policies were associated with a 73% reduction in the likelihood of drug launch within 9 months of regulatory approval relative to non-ERP settings. In addition, while ERP was associated with statistically significant reductions in annual drug price changes, such policies did not impact launch price. In addition, no single ERP feature (e.g., number of countries referenced, ERP calculation) was materially associated with the outcomes of interest. We conclude that ERP policies do not appear to impact drug launch price and may delay access to new therapies, raising questions about the utility of such policies in the US and potential consequences abroad.

1. Introduction

Pharmaceutical cost growth has accelerated in the US in recent years, with per capita drug spending increasing from $783 to $1,025 from 2007 to 2017 [1]. The Centers for Medicare and Medicaid Services project that spending will further increase by 60%, reaching over $1,600 by 2027 [1]. To reign in drug costs, some US policymakers have proposed adoption of external reference pricing (ERP) [2].

ERP refers to the practice of using the price of a pharmaceutical product in one or several jurisdictions to derive a benchmark, or reference price, that informs setting or negotiating the price of the product in a given jurisdiction [3]. The stated goal of ERP is to contain costs and ensure that the maximum price paid for a drug is not excessive relative to its price in other countries [4]. ERP is widely used as a management tool in Europe, Asia, and other locations worldwide [5].

The US has considered using ERP as a mechanism to reduce drug prices, including recent legislation that proposed setting the maximum fair price for drugs that represent a substantial expense to Medicare and face limited competition at 120 percent of the average list price across six countries (Australia, Canada, France, Germany, Japan, and the United Kingdom) [2]. The Congressional Budget Office estimated that enacting this provision would reduce direct Medicare spending by about $450 billion over a 10-year period [6].

The adoption of ERP in the US would have both domestic and international implications. On one hand, ERP would more closely align drug prices in the US with those in comparable countries, resulting in substantial savings to the health care system. However, there are concerns about the sustainability of these savings, launch delays in other countries, and the potential that drug manufacturers may raise launch prices to counteract the effects of ERP [7], [8], [9].

To assess the impact of ERP as a means to inform potential US pricing policy, we sought to understand whether the time from market approval to product launch for selected drugs was significantly different in countries with ERP policies compared to those not using ERP. We also examined the effects of ERP on the price at product launch as well as the average annual change in price. As a secondary aim, we explored whether specific aspects of the ERP approach (e.g., number of countries in reference basket) were associated with these outcomes.

2. Study data and methods

2.1. Pricing and launch timing data

Information on product pricing and the timing of product launch was obtained from the Pricentric ONE™ database (Eversana, Chicago, IL). The database includes information from publicly-available sources in approximately 100 countries on drug prices (including manufacturer list price, pharmacy purchase price, and retail pharmacy price), regulatory market authorization and drug launch dates, and for countries with ERP policies, detailed descriptions of the rules and formulae applied. Data are updated within 72 hour of publication or posting on public sites. Pricing data are reported in both local currency and US dollars using exchange rates available during the reporting time period (e.g., data reported in 2014 dollars use the 2014 exchange rate).

2.2. Selection of drugs and countries

We focused attention on drugs most likely to be subject to reference pricing policies in the US—i.e., those generating significant expenditures for public payers. We selected branded drugs without significant generic or biosimilar competition that were associated with annual spending per beneficiary greater than $10,000 on any one of the following: (a) the 2015 or 2019 Medicare Part B spending dashboards; [10] (b) the 2015 or 2019 Medicare Part D spending dashboards; [11] or (c) the 2015 or 2019 Medicaid spending dashboards [12]. Included drugs represented various disease areas, drug classes, routes of administration, age indications, orphan drug indications, and cancer indications. Additionally, we defined significant competition as drugs that have generics or biosimilars both approved and available on the market. In some cases, generic or biosimilar drugs have been approved by the FDA but are not yet marketed; this would not represent significant competition. We identified a total of 100 drugs meeting these criteria (see Appendix S1 for the full list of selected drugs).

Countries selected for the analysis were required to have stable pricing data reported in the Pricentric database from January 2010 up to the most current available update (October 25, 2021). The final list included 76 countries, of which 44 use ERP as the sole pricing mechanism, 12 use it as a supportive tool along with other mechanisms, 10 employ ERP for nominal informal benchmarking purposes only, and 10 do not use ERP at all (see Appendix S2 for the full list of countries).

For all analyses, the unit of observation was the drug-country dyad, given that policy-setting is done at the country level but drug price and launch timing varies by product.

2.3. Outcomes

There were three outcomes of interest. The timing of drug launch was calculated as the number of days from recorded market authorization (i.e., regulatory approval) to the date of product launch (i.e., the date a published price is first available) in each country. We note that this does not necessarily correlate with product availability, which may be delayed due to price negotiations, coverage determinations, or other reasons. To avoid issues of left-censoring, only products approved between January 1, 2010 (the beginning of our study timeframe) and the last available date of observation were included in this analysis. Dyads without a recorded launch date were censored as of the end of the study timeframe.

Drug prices used for the launch price and price change analyses were estimated using the “ex-factory”, or list price, recorded from officially published prices or estimated based on average and regulated pricing margins in each country (some countries report pharmacy purchase price or retail price only; Pricentric uses a formula based on regulated pricing margins in a given country to estimate list price in such cases). Given differences in package size and dosing form, the price per unit (e.g., mg, mL) was used to standardize comparisons across drugs. All prices were reported in US dollars.

Launch prices were adjusted for the effects of inflation using the overall US Consumer Price Index; [13] we focused on general inflation effects only, given the controls on price increases used by many countries in our dataset. However, because we intended to capture all price dynamics, inclusive of declines, stasis, and increases, we did not adjust prices for the effects of inflation in our analyses of price change. Price change was expressed as an annual average, by dividing the overall change in price for a given drug during the study timeframe by the number of years of observation.

Outcomes were assessed in two distinct datasets. The larger dataset included both ERP and non-ERP dyads, and focused on the impact of ERP on the outcomes of interest. The other dataset was limited to ERP settings only, and sought to identify the impact of various aspects of ERP policy on these outcomes.

2.4. Statistical analysis

Outcomes of interest were first compared between ERP and non-ERP dyads using descriptive statistics. Means and medians for launch timing, launch price, and annual price change were compared using two sample t tests and the Mann Whitney U Test (Wilcoxon Rank Sum Test).

For multivariable models, we assumed data non-independence given the fixed sets of drugs and countries in our analysis. All statistical models therefore included a random effect to account for clustering at the country level. Analyses predicting annual price change further standardized these outcomes by drug type, using methods described below. This standardization also helped account for non-independence and correlation within drug types.

For analyses comparing ERP to non-ERP dyads, candidate covariates included country attributes (e.g., income level, population size, region) as well as drug attributes (e.g., drug class, indicator for whether the drug has a biosimilar or generic version approved). For the ERP-only analyses, these covariates were supplemented by ERP-specific factors (e.g., whether ERP was the sole pricing strategy, used in a supportive role, or as an informal, “nominal” benchmarking tool); the drugs of focus for ERP (e.g., high-priced, all drugs, hospital only, etc.); the formula used (e.g., average, median, minimum); the number of countries in the reference “basket”; and the frequency of price revision. The ERP versus no ERP contrast and ERP-specific factors were forced into all models; the remainder were retained in final models if significant at p < 0.10 in bivariate analyses of variance. The full list of all covariate descriptions is available in the Appendix S3.

Time to product launch was assessed using Cox proportional hazards regression, with results expressed in terms of hazard ratios. An earlier launch date is generally considered beneficial to patient access; in contrast to traditional expressions of benefit, in which hazard ratios < 1.0 are considered positive (i.e., mediated by a delay in a “bad” outcome such as premature mortality), a hazard ratio < 1.0 in this case indicates a delay in launch that should be considered a negative outcome. Violations of the proportional hazards assumption were accounted for by the inclusion of a time variable to “band” the observation timeframe into distinctive periods of follow-up [14].

Launch price was evaluated using generalized linear models with a gamma distribution and a log-link function. Average adjusted marginal effect values were generated, allowing us to examine the effect of each covariate on launch price in dollar values.

Because changes in price included positive, negative, or zero values, standard non-parametric methods were not feasible for analysis. Data were therefore normalized for each drug based on the number of standard deviations from the mean across all countries. For example, if the price change for a drug was 1.5 SD above the mean price change across all countries, the dyad received a normalization factor of 1.5. Change in price was then analyzed using mixed-effects linear regression, with a random effect to account for clustering by country and the normalized outcomes [15].

Finally, we conducted several sensitivity analyses to check the robustness of our results. First, we evaluated whether inflation-adjusting prices affected our conclusions in analyses of price change. Second, in order to increase potential sample size for the time to launch analysis, we imputed market authorization dates when missing based on the median value across drug-country dyads where the data were available. Finally, because the US may be seen as an outlier in terms of launch timing, launch price, and price changes relative to other non-ERP settings, we repeated these analyses with US dyads excluded.

Candidate covariates were retained in final models using a threshold of p < 0.10, in keeping with standard approaches to model specification [16]. Final model results were considered statistically significant if p < 0.05; 95% confidence intervals were also reported. All analyses were programmed in Stata/SE, version 17.0.

3. Results

For the comparison of ERP and non-ERP settings, a total of 2,026 dyads were available for drugs with a regulatory approval date on or after January 1, 2010. All dyads were available for the launch price analysis; due to missing follow-up data, 1,976 (97.5%) dyads were used in the price change analysis. Finally, 1,273 dyads (62.8%) had available data for both market authorization date and launch or censoring date and were included in the launch timing analysis. Corresponding totals for the ERP-only analyses were 1,794, 1,749, and 1,160 respectively. Among the dyads included in the non-ERP launch price and price change analyses, US dyads accounted for 20.3% of the observations. For the non-ERP launch timing analysis, US dyads accounted for 31.6% of the observations.

Dyad characteristics are presented in Table 1, stratified by ERP status. ERP dyads were more often located in high-income settings (71.5% vs. 65.1% for non-ERP) and in Europe (59.6% vs. 31.0%). Among ERP dyads, two-thirds were focused on nominal (i.e., informal benchmarking) uses of ERP, and over half focused ERP application on all prescription drugs. Price revisions occurred most commonly every 0–3 or 7–12 months. ERP calculation was done most frequently using simple basket averages or minimums, but reference basket sizes were evenly distributed, although baskets with>30 countries were rare.

Table 1.

Characteristics of drug-country dyads according to use of external reference pricing.

Variable	ERP N (%) (N = 1,794)	No-ERP N (%) (N = 232)
Income Level
High-Income	1,283 (71.5)	151 (65.1)
Lower-Middle Income	74 (4.1)	20 (8.6)
Upper-Middle Income	437 (24.4)	61 (26.3)
Health System Financing
Fully Public	1,122 (62.5)	177 (76.3)
Public-Private	672 (37.5)	55 (23.7)
Country Population
<3,000,000	304 (16.9)	4 (1.7)
3,000,000–9,999,999	594 (33.1)	9 (3.9)
10,000,000–45,000,000	446 (24.9)	61 (26.3)
>45,000,000	450 (25.1)	158 (68.1)
WHO Region
Africa	35 (1.9)	0 (0)
Americas	132 (7.4)	142 (61.2)
Eastern Mediterranean	348 (19.4)	0 (0)
Europe	1,069 (59.6)	72 (31.0)
South-East Asia	70 (3.9)	18 (7.8)
Western Pacific	140 (7.8)	0 (0)
ERP Usage
Sole	260 (14.5)	–
Nominal	1,205 (67.2)	–
Supportive	329 (18.3)	–
No ERP	–	232 (100)
ERP Focus
All prescription drugs	972 (54.2)	–
Branded only	624 (34.8)	–
High priced	75 (4.2)	–
Hospital only	64 (3.6)	–
Other	59 (3.2)	–
No ERP	–	232 (100)
Price Revision
0–3	479 (26.7)	–
4–6	194 (10.8)	–
7–12	540 (30.1)	–
13–24	102 (5.7)	–
25–36	114 (6.3)	–
>36	365 (20.4)	–
No ERP	–	232 (100)
ERP Formula
Average	722 (40.2)	–
Average lowest of 2	58 (3.2)	–
Average lowest of 3	139 (7.8)	–
Median	132 (7.4)	–
Minimum	743 (41.4)	–
No ERP	–	232 (100)
Primary Reference Basket
1–5	458 (25.5)	–
6–10	494 (27.6)	–
11–20	388 (21.6)	–
21–30	305 (17.0)	–
>30	120 (6.7)	–
Not Reported	29 (1.6)	–
No ERP	–	232 (100)
Drug Class
Alimentary Tract And Metabolism	34 (1.9)	4 (1.7)
Anti-infective For Systemic Use	280 (15.6)	43 (18.5)
Antineoplastic And Immunomodulating Agents	898 (50.1)	112 (48.2)
Blood And Blood Forming Organs	76 (4.2)	12 (5.2)
Multiple	257 (14.3)	35 (15.1)
Nervous System	104 (5.8)	6 (2.7)
Respiratory System	120 (6.7)	12 (5.2)
Systemic Hormonal Preparations, Excl. Sex Hormones And Insulins	25 (1.4)	8 (3.4)
Age Indication
Yes	745 (41.5)	91 (39.2)
No	1,049 (58.5)	141 (60.8)
Cancer Indication
Yes	692 (38.6)	93 (40.1)
No	1,102 (61.4)	139 (59.9)
Orphan Drug Indication
Yes	840 (46.8)	115 (49.6)
No	954 (53.2)	117 (50.4)
Biosimilar or Generic Approved
Yes	350 (19.5)	50 (21.6)
No	1,444 (80.5)	182 (78.4)
Route of Administration
Self-Administration Only	842 (46.9)	105 (45.3)
Physician-Administration Only	868 (48.4)	113 (48.7)
Both Self- and Physician-Administered	84 (4.7)	14 (6.0)

Abbreviation: ERP, External Reference Pricing.

Source: Authors’ analysis of data from the Pricentric ONE™ database (Eversana, Chicago, IL).

3.1. Comparisons of ERP and non-ERP settings

3.1.1. Launch timing

In descriptive analyses, ERP was associated with a statistically significantly longer time to product launch (mean: 625 days vs. 262 for no ERP, p < 0.001; median: 399 days vs. 70 for no ERP, rank sum p < 0.001) (Table 2).

Table 2.

Descriptive analyses of impact of external reference pricing on launch timing, launch price, and price changes.

	ERP	No ERP	p-value
Launch Timing (days)
Mean	625.15	262.35	<0.001
Median	399.00	70.00	<0.001
Launch Price ($/unit)
Mean	1,285.65	1,249.86	0.781
Median	465.07	259.33	0.094
Annual Price Change ($/unit)
Mean	−25.61	111.31	<0.001
Median	−0.03	0.003	<0.001

Note: statistical significance testing using two sample t test for means and Mann Whitney U Test (Wilcoxon Rank Sum Test) for medians.

Source: Authors’ analysis of data from the Pricentric ONE™ database (Eversana, Chicago, IL).

In multivariable analyses, a time variable (0–250 days, 251–1,000 days, and > 1,000 days) was introduced to account for a violation of the proportional hazards assumption; full results are presented in Appendix S4. In the first 250 days, ERP was associated with a 73% reduction in the likelihood of product launch relative to no ERP (HR: 0.27, 95% CI: 0.12, 0.60). As an illustration, the cancer therapy pertuzumab (Perjeta®, Genentech USA, Inc.) had an average time to launch of 1,007 days in ERP settings compared to 423 days in non-ERP settings (Fig. 1), but this varied by setting and region. No statistically significant differences in launch timing were observed in the 251–1,000 and > 1,000 day windows. Other factors associated with later launch timing within the first 250 days included population size between 10 and 45 million versus > 45 million (HR: 0.27, 95% CI: 0.15, 0.51), as well as nervous system (HR: 0.10, 95% CI: 0.03, 0.41) and respiratory system (HR: 0.59, 95% CI: 0.40, 0.87) drug class. Factors significantly associated with earlier launch timing during this window included higher country income, public–private health financing, and European location.

Fig. 1.

Launch timing of pertuzumab by country and external reference pricing status Source: Authors’ analysis of data from the Pricentric ONE™ database (Eversana, Chicago, IL).

3.1.2. Launch price

Average launch price did not differ between ERP and non-ERP dyads, both in descriptive (mean: $1,286 vs. $1,250 per unit for no ERP, p = 0.781; median: $465 vs. $259 per unit for no ERP, rank sum p = 0.094) and multivariable (between-group difference, -$175, p = 0.42) analyses (see Appendix S5). Other factors significantly associated with a lower launch price included lower country income, alimentary tract and metabolism, anti-infectives, nervous system, and systemic hormonal preparation drug classes, and drugs with a competing approved generic or biosimilar product ($907 lower). Factors significantly associated with a higher launch price included antineoplastic and immunomodulating agents, blood and blood forming organs, and respiratory system drug classes.

3.1.3. Annual price change

In descriptive analysis, prices declined annually on average in ERP dyads, whereas they increased in non-ERP dyads (mean annual price change: -$26 per unit vs. +$111 for no ERP, p < 0.001; median annual price change: -$0.03 per unit vs. 0.003 for no ERP, rank sum p < 0.001). In mixed-effects regression analysis (displayed in Appendix S6), ERP dyads experienced average annual price changes 0.65 standard deviations lower than the mean for their drug type in comparison to non-ERP dyads (p = 0.003), representing monetary changes ranging from -$0.03 to -$2,410 per unit. For example, the ocular medication aflibercept (Eylea®, Regeneron Pharmaceuticals, Inc., US) had annual price declines of $23 per unit on average in ERP dyads, whereas prices increased by $30 per unit in non-ERP dyads (Fig. 2), although this varied by setting and region. Additional factors statistically significantly associated with a lower than average annual price change included European location and dyads with a competing approved generic or biosimilar product.

Fig. 2.

Price change of aflibercept by country and external reference pricing status Source: Authors’ analysis of data from the Pricentric ONE™ database (Eversana, Chicago, IL).

3.2. Analyses of ERP features

3.2.1. Launch timing

In analyses of ERP dyads only, factors significantly associated with later launch timing within the first 250 days (Appendix S7) included use of ERP on all prescription drugs versus a subset (HR: 0.37, 95% CI: 0.20, 0.68), less frequent price revision (i.e., every 13–24 or > 36 months vs. 0–3 months), an ERP formula that uses the minimum reference price in the reference basket versus the average price (HR: 0.19, 95% CI: 0.10, 0.36), and dyads with reference baskets of 21 to 30 countries versus 0 to 5 (HR: 0.35, 95% CI: 0.16, 0.74).

3.2.2. Launch price

Relatively few features of ERP policy were associated with lower launch prices (Appendix S8). An ERP focus on high-priced drugs only was associated with a launch price over $1,100 per unit lower than other foci (p = 0.04). Additionally, policies referencing the average of the 3 lowest-price countries was associated with a launch price that was $428 lower than those using the overall average, while countries utilizing the median were associated with a launch price $923 higher in comparison to average price (p = 0.013 and 0.045, respectively).

3.2.3. Annual price change

No feature of ERP policy was significantly associated with either a negative or positive average annual price change (Appendix S9). The only factors with a trend toward a lower average annual price change were European or South-East Asian location (0.73 and 1.1 standard deviation below the mean, p = 0.064 and 0.068, respectively).

3.3. Sensitivity analyses

Results of our sensitivity analyses generally mirrored our primary findings. Adjusting prices for inflation produced similar results for annual price change (ERP dyads saw price changes 0.59 standard deviations below the mean for their drug type vs. 0.65 in primary analyses). ERP features that were statistically significant in primary analyses remained so when adjusting these prices for inflation (Appendix S10-S11).

Imputing market authorization dates when missing increased our sample size from 1,273 dyads to 1,993(Appendix S12-S13). Results were similar after imputation. ERP settings were 53% less likely to see a product launch in the first 250 days relative to non-ERP settings (HR: 0.47, 95% CI: 0.24, 0.92), versus 73% in primary analyses. In the ERP-only analysis, factors significantly associated with later launch timing within the first 250 days were similar to those in primary analyses, although more formulae for price calculation were significant (i.e., average of the 2 lowest-price countries, average of the 3 lowest-price countries, median, and minimum).

Finally, exclusion of US dyads from the analysis produced results consistent with primary analyses (Appendix S14-S16). For example, in the first 250 days, ERP was associated with a 72% reduction in the likelihood of product launch relative to no ERP when the US was excluded (vs. 73% in primary analyses). As with primary analyses, no significant difference in launch price was observed between ERP and non-ERP settings when the US was excluded. Annual price changes in ERP settings were 0.58 standard deviations lower than the mean for a given drug type versus non-ERP dyads (p = 0.012), similar to the 0.65 seen in primary analyses.

4. Discussion

Using a commercially-available international database of prescription drug prices and launch timing, our analysis of a diverse set of high-cost medications and countries suggests that countries employing external reference pricing practices see greater reductions in list price annually relative to those who rely on other pricing approaches. However, ERP was also associated with significant delays in product launch after regulatory approval; ERP settings were nearly 70% less likely to have a new drug launched within the first 9 months of market authorization than non-ERP settings. Additionally, despite the impact on price change, our analysis finds that ERP does not appear to have significant effects on launch price, suggesting that its utility may relate more to price management than negotiation.

Our findings are consistent with those of other comprehensive studies in the field. For example, Lee et al. conducted a systematic review of 255 studies of drug expenditure controls, 29 of which involved “industry regulation” (primarily through reference pricing) [17]. The authors found that, while some cost savings over time were realized, these primarily came from shifts in utilization to reference products and increasing patient cost-sharing, and that ERP policies contributed to delays in access in referenced countries. A combined systematic review and expert survey of ERP activity in 21 countries identified a short-term effect of ERP on price trends, but some evidence of launch delays and/or product withdrawals in referenced countries [18].

Danzon and Epstein examined the impact of ERP policies on drug prices across 12 therapeutic classes in 15 countries from 1992 to 2003 [19]. As in our analysis, the authors find significant evidence that reference pricing by high-price countries creates incentives to delay launch in lower-price countries until higher prices have been established. Our study adds to this evidence through updated data, a broader set of countries and drugs, and by quantifying the magnitude of these delays.

However, our analysis also suggests that teasing out the specifics of ERP and their impact on these important outcomes is exceedingly difficult. Our own assessment of ERP components produced mixed and inconsistent results. For example, few aspects of ERP policy were statistically significantly associated with launch price or price change. And, while the association of some ERP features with launch delays makes intuitive sense (e.g., large reference basket size, use of minimum to set reference price), other associations (e.g., infrequent price revisions) do not. These issues are not uncommon in quasi-experimental assessments of policy interventions, particularly in cases like this where a clear counterfactual (i.e., outcomes of interest in the absence of ERP policy) is not observable and the impact of unobserved factors is of potential concern [20]. We did use non-ERP settings as a control group, but these represented different health systems, financing mechanisms, and societal norms.

We note several limitations of this analysis that warrant consideration. First, while we accounted for the impact of measurable factors on our outcomes of interest, we could not exclude the potential impact of unmeasured confounding. For example, drug pricing dynamics could also be affected by other country-specific policies outside of the reference pricing sphere, such as trade agreements or industry-wide “clawback” provisions. Similarly, launch timing could be affected by company strategic priorities and resourcing as well as regulatory timelines.

Second, we used actual or estimated prices that were based on publicly-available data. Manufacturers often gain favorable access in a given country by negotiating confidential price discounts [21]. While such activity may not directly affect reference price calculations, which are typically based on public prices, there may be impacts on market entry or exit that we cannot account for using the available data.

Our analyses of launch timing focused on time from market authorization to product launch in a given setting. However, further delays in adoption and reimbursement after official launch can also occur, due to protracted negotiations with payers and health systems, requirements for completion of a HTA process, or other factors. Our launch timing estimates should therefore be considered somewhat conservative.

Finally, we focused attention on drugs that result in high expenditures for US public payers. While the sample was large, there may be other drugs of interest to commercial payers and/or other countries that were not evaluated, and the effects of their inclusion on the magnitude or direction of our findings is unknown.

What should US policymakers take from this study? First, while the precise impact of ERP in the US is unknown, there is likely to be a short-term reduction in prices for established drugs, in no small part because of a significant divide in prices between the US and other high-income countries [22]. However, as demonstrated by our study, drug launches may be delayed in other countries to preserve US pricing. Other effects are possible given the threat to the industry’s most lucrative market, including drug exits from the lowest-price markets for established medicines and increases in US launch prices [4], [7].

Of note, many of the indices proposed in recent legislation or proposed administrative rules include countries that formally seek to align drug prices with the value they bring, through health technology assessment, cost-effectiveness analysis, or both [7], [23], [24], [25], [26]. It is possible that some version of ERP could be used as a supportive mechanism to set ceiling prices or to modulate price changes over time. In the long term, however, the US should consider developing its own approach to value-based pricing as a sustainable policy that could avoid price gaming and manipulation, both here and abroad.

CRediT authorship contribution statement

Dominic Voehler: Conceptualization, Data curation, Formal analysis, Methodology, Visualization, Writing – original draft. Benjamin C. Koethe: Formal analysis, Methodology, Investigation, Software, Writing – review & editing. Patricia G. Synnott: Conceptualization, Project administration, Validation, Writing – review & editing. Daniel A. Ollendorf: Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Supervision, Validation, Writing – review & editing.

Declaration of Competing Interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Mr. Koethe declares no known competing interests. Mr. Voehler, Ms. Synnott, and Dr. Ollendorf are employed by a research institute that receives funding from life sciences companies, government agencies, and academic institutions to support and maintain multiple databases. Dr. Ollendorf also provides consulting services to pharmaceutical and device manufacturers, outside the scope of this work.

Appendix A. Supplementary data

The following are the Supplementary data to this article: