Dr. Ron Bouchard is an Associate Professor in the Faculties of Law and of Medicine at the University of Manitoba, a CIHR New Investigator and an IP Osgoode Research Affiliate.
In Part 1, we learned that it is both possible and valuable to import empirical scientific methods typically used in the hard sciences to the study of law. In fact, in our analysis of patent law and policy we can move beyond patent valuation to assess how and indeed whether a given piece of law or policy is working in conjunction with its so-called original policy intent. This includes the assessment of innovation within the context of the patent bargain and whether governments that have accepted pharmaceutical linkage laws are being rewarded in their twin policy goals of producing more new and innovative drugs and facilitating timely generic entry. Using the new tools of empirical legal research, we hope to assess whether, as Senator Hatch put it at the time the U.S. legislation came into force, the public is in fact “receiving the best of both worlds – cheaper drugs today and better drugs tomorrow.”
We can address this possibility using the innovation index discussed in Part 1 in combination with 3-D spatiotemporal models such as those used in the medical sciences. Over the last few decades, these models have been used increasingly for studying protein, DNA, RNA, and other structure-function relationships, including using x-ray and other crystallography techniques. Consistent with their use in medicine, 3-D legal models can be used to construct data for both descriptive (structural) and prescriptive (functional) law-making and law-reform purposes.
For example, in our Northwestern study, we developed a 2-D model of identifying patents in relation to “new and innovative” drugs and “follow-on” drugs that tracked the functional and temporal evolution of drug forms and associated patents over time. The example below is for the combination of Salmeterol and Fluticasone into one of several available forms of Advair®. We referred to this technique as a “patent tree” method and used it specifically to identify legally-related drug forms, associated patents, and patent types.
Fig. 1. Example of Convergent Patent Tree Analysis for Fourth Generation Product Advair Diskus.®
Patents were identified using the specific and general search strings described in our Berkeley study. In addition to quantifying patents per drug, the patent tree method allows assessment of how specific drugs evolve into related drug forms or (in this case) drug products representing combinations of known drugs. In addition, the patent tree analysis allows for identification of relevant patent types based on the classification nomenclature described in the Northwestern study. Finally, the patent tree analysis provides data relating to drug development, but also on the type of patents selected by pharmaceutical companies for listing on the patent register in order to prevent generic entry.
This method can be extended, as shown below, to identify “product clusters.” The patent tree method was expanded to include patents listed on the patent register under linkage law and a diagonally increasing axis of cumulative spatiotemporal growth. The model represents a constellation of legally and functionally related new and follow-on drug forms and regulatory approvals, patents associated with these drug forms, the fraction of total patents listed on the patent register in order to slow down generic entry under linkage laws, and how each of the data classes relate to one another over time.
Fig. 2. Product Cluster-Based Model of Drug Development. Product clusters begin at some point in time with the first new and innovative drug (big orange circle; NCE) and associated originating patent (small purple circle). With time, and vetting by the market and regulators, further follow-on drug approvals (big blue circle) and patents (small green circle) are granted within the cluster, and an increasing number of these patents are listed on the patent register (small red circle). Listed patents can be used increasingly over time to prohibit generic entry not only on the originating new and innovative drug, but also on all drugs in the cluster that are deemed under law to be relevant to the originating drug.
We are now in a position to take our 2-D product cluster model above, first reported in 2011, and combine it with the innovation index depicted in Part 1 of this series, reproduced below for convenience. The result is the spatiotemporal product cluster model shown below in Fig. 4.
Fig. 3. Innovation Index Data for Total Approval Cohort. Bar graphs showing the number of total approvals expressed as a function of the level of innovation (LOI) before (a) and after (b) of generic approval data. (c) shows brand approvals expressed as a function of LOI. Solid line is a fit of the data to a single exponential function. (d) shows cumulative normalized brand approvals expressed as a function of LOI. Solid line is fit using a sigmoidal function.
The combination of the drug nomenclature, product cluster and innovation index described in Fig. 4 yields a potentially new way of looking at the impact of regulatory and market incentives on drug development by multinational firms. As shown by the data in the Boston study, this clearly includes both brand-name firms and generic firms, as both are pursuing cluster-based models of drug development. The resulting analytical model focuses on drug development driven by purposeful policy, and cumulative vetting of serial products by regulators and the market.
Fig. 4. Combining Innovation Index and Product Cluster Models to Study Portfolio-Based Drug Development and Hedging. Product clusters are hypothesized to begin life at the most innovative end of the spectrum, with few patents and a small or negligible number of listed patents. Over time, and with increased vetting by regulators and the market, the cluster expands to include more products, patents and listed patents but, as a whole, becomes less and less innovative. The desired end point (the “home run”) is a substantial but low-level cluster with numerous products, patents and listed patents, and the widest scope of market exclusivity and cumulative patent protection. Prior to this point, clusters are “at bat”, as they reach a critical state prior to moving into an expanded spatiotemporal state or merely “on deck” as firms await critical regulator and market vetting.
Described in detail in a forthcoming book, which summarizes our research over the last four years, drug clusters denoted ‘on deck’, ‘at bat’, and ‘home run’ represent a theoretical mock-up of how drug clusters grow in time from a spatiotemporal perspective. In this model, product-patent clusters begin their life as single-drug products or small groupings at the most innovative end of the index and, with increased vetting of products in the cluster over time by regulators, the market grows in scope to encompass an increasing number of products and patents. As this occurs, the cluster may be anticipated to ‘swing up and to the left’ of the innovation index, moving from a high level of innovation with a low number of patents and listed patents to first a moderate and then a much lower level of innovation but with greater spatiotemporal characteristics. The model shown here is for 2,087 drug approvals over an eight-year study period; similar results have been obtained using patents and chemical components.
An important observation with regard to product-patent drug clusters is that as a given cluster grows spatiotemporally over time, it grows not only in scope but also in the scale of the interrelatedness of its functional components over time.
As noted in 2001 by Kingston and later by Polk & Parchomovsky and, notably, the EC Pharmaceutical Sector Inquiry, the strength of patent portfolios and related product clusters from an intellectual property law perspective is “greater than the sum of its parts”. An easily identified example of the value-added dimension of clusters is the strong profitability for follow-on Me Too drugs that nevertheless have low levels of innovation. This “more is different” element of product clusters, originally described in 1972 by PW Anderson, is characteristic of complex systems, including complex legal systems such as those described by JB Ruhl and many others in the mid-1990s. As noted in Part 1, we have referred to the complex multidirectional interrelationships and interdependencies between drug development, drug regulation and intellectual property law in our previous McGill and Berkeley studies as a regulated Therapeutic Product Lifecycle or rTPL.
Of interest, our data show that the profit of a given molecule is strongly related to the number of patents, regulatory approvals, the number of patents listed on the register, and the range of drugs and regulatory approvals that are legally related but separated by only very minimal changes to existing uses and chemistry. This is true even for drugs thought be innovative such as those with First in Class and New Active Substance (New Chemical Entities), owing to regulatory loopholes.
At least somewhat surprisingly, in light of global innovation policy over the last 50 years, the greater the number and scope of these metrics, the lower is the calculated level of innovation of a basket of drugs in a product cluster. As market and regulator vetting increases with time, one sees generally (1) more patents, regulatory approvals, fractional patent listing, patent classifications per marketed drug, (2) a greater follow-on-to-new drug ratio in the cohorts studied, and (3) greater profitability for less innovative drugs.
Indeed, drug clusters driven by line extension, or follow-on, drugs are proving to be very profitable. For example, we found that the vast majority of approval, patenting and chemical development activity associated with brand pharmaceutical products is directed to the development of Me Too drugs, in particular follow-on Me Too drugs. Of the top 25 most profitable drugs in 2006, 48% (12) were line extension Me Too drugs. The combined sales of these drugs were US $45.7 billion dollars. Follow-on First in Class drugs represented 28% of the top 25 selling drugs, and 7 of the top 15 selling drugs. Profit on this group of drugs was US $39.7 billion dollars in 2006. Combined, follow-on Me Too and First in Class drugs accounted for 19 of 25 of the most profitable drugs, with total sales of US $85.5 billion in a single year.
From a science of law perspective, a major advantage of the rTPL and product cluster models is that there is, in fact, considerable empirical evidence available for study. This includes the various types of new and follow-on drugs, patents, patent classifications, listed patents, related litigation, as well as the relation of these metrics to one another over time. This wide array of empirically observable metrics and the observation that they change over time sets up the possibility that, akin to protein folding and X-ray crystallography models, the data can be expressed in 3-D spatiotemporal form.
Indeed, the goal of our empirical work over the last four years involving new and follow-on drugs, patent trees, patent types, WHO Anatomical Therapeutic Classification (ATC) data, litigation data, the innovation index, and product cluster model is to convert the cumulative data into 3-D formats used in the medical sciences. For example, the protein-RNA model presented below underscores the utility of 3‑D “rotational” models to both identify and quantify the complex structural and functional characteristics in a given network of biological components, here those between an RNA strand and protein components in the context of Multiple Sclerosis.
Fig. 5. Medical Sciences Template for Rotational 3-D Spatiotemporal Models of Cluster-Based Drug Development. From: Joint Evolutionary Tree Method for Study of MS.
As discussed previously, rotational 3-D drug product-drug patent cluster models would be particularly useful to policy-makers and law-makers in order to enable visual and numerical quantification of the impact of intellectual property law on drug development, generic entry, and access to essential medications in the same manner that one might look at a car from behind (highlighting the ‘gas tank,’ or original drug product and associated patent tandems) as well as from the side (from the rear to the front of the vehicle, underscoring how and when approvals, patents, and listed patents increase over time with market and regulator vetting).
In this manner, extrapolating the empirical techniques conventionally used in the hard sciences to the study of law, including patent law and innovation policy, offers an important opportunity to not only quantify the effect of a given piece of law or policy, but also to help determine the vires of such laws after they have been put in motion and to guide law reform efforts in light of objective arm’s length evidence.
It is hoped this series of articles has shed some light on the utility of traditional scientific methods for quantitative and qualitative assessment of patent value, and on whether laws made decades ago to enhance innovation in the pharmaceutical sector and to facilitate timely generic entry are producing intended effects, unintended effects, or some combination of both.
In any event, it will be interesting to see whether, as in other fields such as medicine and engineering that are accustomed to taking an “evidence-based” approach to problem identification and problem solving, we in the legal field may also include empirical evidence in our expanding toolkit of legal assessment and interpretation methods.