It is very important to underline that the preliminary findings … do not address the question of whether any behaviour identified … could amount to a violation of EC competition law, not least as this would require the definition of the relevant product markets, which is rather complex in this sector.” (Preliminary Report in the Pharmaceutical Sector Inquiry, Para. 12)

Generic medicines are chemically equivalent to original medicines” (Press Releases to Case No. IV/M.5295, Teva / Barr, Commission Decision dated 19 December 2008 and Case No. IV/M.5253, Sanofi-Aventis / Zentiva, Commission Decision dated 4 February 2009)

Late last year, Teva Pharmaceuticals notified the European Commission (the "Commission”) of Teva’s proposed acquisition of Barr Pharmaceuticals. Both businesses produce and sell active pharmaceutical ingredients, proprietary-brand and generic drugs.  Teva/Barr drew attention because the Commission analysed some horizontal overlaps at molecule level, while, historically, the Commission’s approach has always been to use as its starting point wider drug classifications. 
Shortly after the Commission conditionally cleared this deal, the Commission published its Preliminary Report in the Pharmaceutical Sector Inquiry.1 The Preliminary Report obliged stakeholders to provide information on their products at molecule level. Since then, the Commission’s decision on the takeover of Czech generic drugs manufacturer Zentiva by Sanofi-Aventis has been published. Is the Commission changing the way that it defines pharmaceuticals markets? And if so, what does that change means for competitive analysis?

It is difficult to make generalisations about pharmaceutical products because of the numerous ways in which apparently similar drugs can differ from each other. Key differentiating factors include the part of the body targeted by the drug, the treatment(s) for which it is indicated, contraindications and side-effects, means of administration and monitoring requirements.  In addition, factors such as prescribing practices, price and availability are key. It is inevitable that any classification methodology for this sector needs to be used cautiously.

To understand how the Commission has defined pharmaceutical product markets to date, we must refer to two widely used and similar drug classification systems: (i) the Anatomical Therapeutic Chemical Classification System of the World Health Organisation (WHO); and (ii) the classification system used by the European Pharmaceutical Market Research Association (EphMRA).

Each system purports to classify drugs differently than the other. The WHO system classifies substances and is used for drug utilisation research and adverse drug reaction monitoring, whereas the EphMRA system classifies products and is used by IMS (Intercontinental Medical Statistics) to provide market statistics for the pharmaceutical industry. In practice, the systems are similar and are often confused.2

The EphMRA system is divided into four levels, with the first containing the broadest categories (the parts of the body treated by a drug) and the fourth, the narrowest (usually, but not always, by active pharmaceutical ingredient, or molecule). The second and third categories classify according to slightly more arbitrary therapeutic sub-dividers: broad therapeutic function and therapeutic/pharmacological indication, respectively. 

The WHO system works similarly to the EphMRA system, although it has five levels (with three therapeutic sub-dividers in the middle) and the categories with respect to particular areas are not always the same as the EphMRA system.  In both systems, the molecule level is the narrowest level.

Traditionally, the Commission has used the third level of the EphMRA system (referred to as “ATC3”) as a starting point for market definition. That is, it looks at the intended use of the drug in order to work out which products are substitutable with others.3  It should be stressed that the Commission regards this only as a starting point: it generally recognises that even if some drugs have the same intended use, factors such as contraindications may severely limit substitutability. This recognition is well-documented and goes back to the earliest pharmaceutical merger cases:

It is inevitable that any workable market definition in the pharmaceutical sector will involve a certain amount of arbitrariness, because in the final resort, substitutability among medicines may not only depend on the intrinsic characteristics of the drug itself, but also their intended use, taking into account the patient’s overall condition.4

The Commission is open to merging parties’ ideas of how products should be grouped together, but appears to test these in its market investigations. In Teva/Barr, the parties consistently argued for ATC3 market definitions, which the Commission used as its starting point but refused as a conclusion in six instances on the basis of its market investigation. In Sanofi-Aventis/Zentiva, ATC3 was always used as the starting point and, while the Commission recognised that certain markets were smaller, none were defined by molecule (J4A, tuberculosis products, was further subdivided, but not by molecule, and for B1B, Heparin, the Commission indicated that that part of the market accounted for by hospital customers might be looked at separately).

Not checking a market definition, however large or small, by referring to other stakeholders  risks producing a perverse result.  ATC3 is clearly not always appropriate. For example, in Teva/Barr, the Commission identified that there was limited substitutability between drugs belonging to ATC3 category L01C (vinca alkaloids and other products, a subcategory of Category L - antineoplastic & immunomodulating agents5).  Using this category would group Paclitaxel, an oncology treatment, together with other drugs with dissimilar indications, including irinotecan and topotecan, which are based on different molecules. Using the ATC4 level (taxanes, or drugs derived from the Yew tree family) would still not be sufficient: this would place Paclitaxel with Doclitaxel. While use of these two drugs overlaps for certain breast cancers and non-small-cell lung cancer, for other types of cancer only one or the other is appropriate. The Commission concluded that drugs based on the same Paclitaxel molecule would be in the same product market. Paclitaxel has a wide range of uses: it treats certain types of lung, ovarian, breast, head and neck cancer, Kaposi’s sarcoma and the prevention of restenosis (the recurrent narrowing of coronary stents).

In the same decision, the Commission also defined the market for Tamoxifen by molecule after consulting hospitals and recognising that there was only partial overlap between different molecules within the relevant ATC3 category.  Tamoxifen is used mainly in the treatment of post-menopausal breast cancer, but is also used to treat infertility in women with anovulatory disorders, gynecomastia (abnormally large mammary glands in men), bipolar disorder and angiogenosis (the development of new blood vessels).

The range of distinct uses for both Paclitaxel and Tamoxifen hints at the conflict between workable market definitions and accurate ones.  For each, the Commission used definition-by-molecule, but did it go far enough?  Accuracy would dictate that Tamoxifen’s use in treating post-menopausal breast cancer should place it in the same market as other products, based on other molecules, that also target post-menopausal breast cancer in the same or similar ways and are considered alternatives by prescribers (prima facie, Anastrozole and Raloxifene belong to the same group as Tamoxifen).  That is, Tamoxifen’s function in the treatment of breast cancer should be looked at separately from its other functions in order to shape an accurate market definition for this particular use.

Nonetheless, the practical problem of this approach, even if the Commission’s market investigation has included sufficiently helpful medical input to arrive at such accuracy, is the fact that the prescribing data do not record the purpose for which a drug was prescribed.  Molecule level is the most detailed level at which IMS sales data can be collated.  In this case, can the substantive assessment take these factors into account instead?

In early ECMR decisions such as Procordia/Erbamont, Rhone Poulenc/Cooper and  Roche/Syntex,6 the Commission defined markets according to ATC3 and went on to consider substitutability in detail in its substantive assessment, after it had noted the resulting market shares.  A potential overlap concern was identified and eliminated because the products in question turned out not to be substitutable.  The trouble with adopting a revised version of this approach for multi-function drugs being defined at molecule level is that the Commission would not know how much weight to attach to the distinct uses of the drug.  Therefore the current approach may be described as the best workable.

The Commission will still use a wider market definition if there are grounds for doing so.  In Teva/Barr, it used ATC3 where drugs based on separate molecules had the same application and patients could be switched between them.7  In Sanofi-Aventis/Zentiva, ATC3 was used for a number of markets (although the Commission’s reasons for doing so are not set out clearly).8  In other decisions it has discussed the merits of ATC2.9  But it should be stressed that there are a number of potential phenomena that can affect a drug’s market definition, the solutions for which may require more work than going further up or down the ATC scale (e.g., picking and grouping together certain ATC4 categories within an ATC3 group, but not others).

In merger cases, the Commission’s investigation is focussed on overlaps.  In the context of a horizontal overlap, use of a narrower market definition is not necessarily disadvantageous for the parties concerned because, in theory, there is always the possibility that an overlap will disappear.  However, narrower market definitions are more controversial in the context of behavioural cases, in particular those involving single firm dominance. One example is the Commission’s product market definition in COMP/A37.507/AstraZeneca, which is one of the points currently on appeal to the Court of First Instance.10  Here, the Commission’s definition failed to include a category of drug that has a similar mechanism and is used to treat the same medical conditions as the one for which AstraZeneca was alleged to be dominant.  The Commission would not have been able to conclude that AstraZeneca was dominant had the additional category formed part of the relevant market.

At the time of writing, it is still too early to judge what the outcome of the Commission’s Sector Inquiry will be. The Commission has not yet addressed market definition, recognising that this will be challenging (see opening quotation). However, the Commission’s request for data from stakeholders at molecule level has led to a concern that molecule level market definitions would be used as a basis for the identification of anti-competitive behaviour.  This concern seems misguided, given the Commission’s continued use of ATC3 as a starting point in Teva/Barr and Sanofi-Aventis/Zentiva.

The use of molecule level as the basis for the information collated for the Sector Inquiry may have something to do with the fact that the Commission’s work stream here is part-focused on generic medicines.  To paraphrase the Commission, generic medicines are chemical equivalents of originator products. One of the Commission’s concerns is the alleged delayed entry of generic medicines onto the markets within the EU. Collating data at molecule level, rather than at ATC3 level, is convenient in this regard because it easily identifies those instances where originator and generic medicines overlap.

Nonetheless, the Commission’s continued awareness of this sector’s product heterogeneity and consideration of the factors that contribute to drug substitutability are necessary to arrive at market definitions that are as accurate as possible. It seems that ATC3 is still the starting point in pharma-related ECMR decisions, and that there are grounds for defining certain drugs markets by molecule. However, even if molecule level divisions have replaced ATC3 as the starting point for the sector inquiry, it would be surprising if the Commission did not continue to qualify and test its starting point by reference to these factors.


1. Sector inquiries are investigation carried out by the Commission into an entire sector or industry in which the markets do not appear to be working as well as they could.
2. The Commission sometimes confuses one with the other: e.g. Case No. IV/M.72, Sanofi / Sterling Drug, Commission Decision dated 10 June 1991, Para. 12.
3. Examples include Case No. IV/M.3544, Bayer Healthcare / Roche, Commission Decision dated 19 November 2004 and Case No. IV/M.3354, Sanofi-Synthelabo / Aventis, Commission Decision dated 26 April 2004.
4. Case No, IV/M.72, Sanofi / Sterling Drugs, Commission Decision dated 10 June 1991, Para. 12.  Similar comments may also be found in e.g. Case No. IV/M.323, Procordia / Erbamont, Commission Decision dated 29 April 1993, Para. 8, and Case No. IV/M.737, Ciba-Geigy / Sandoz, Commission Decision of 17 July 1996: “The interchangeability of products depends in principle not on their physical, technical or chemical properties but on their functional substitutability as viewed by those supervising their consumption”.
5. Antineoplastic agents treat neoplasia (abnormal cell growths), which can develop into cancer.  Immunomodulating agents are used for their effect on the immune system.  This broad category contains all drugs that are generally indicated for the prevention of cancer.  Vinca alkaloids are chemical compounds that contain basic nitrogen atoms.  These inhibit cancer growth by preventing cell division.
6. Case No. IV/M.323, Procordia/Erbamont, Commission Decision dated 29 April 1993; Case No. IV/M.426, Rhone Poulenc/Cooper, Commission Decision dated 18 April 1994and Case No. IV/M.457,  Roche/Syntex, Commission Decision dated 20 June 1994.
7. Category A11B – multivitamins without minerals (although it could be argued that the risks of taking one product rather than another in this category are significantly less than for drugs treating specific diseases).
8. E.g. A5B, hepatic protectors and lipotropics, for which the discussion is limited to why a further division between over-the-counter and prescription drugs is not appropriate; A7A, intestinal anti-infective antidiarrhoeals - here the Commission only references an earlier case in which the market investigation indicated that ATC3 was appropriate, without giving reasons.
9. Case No. IV/M.1403, Astra/Zeneca, Commission Decision dated 26 February 1999 (see e.g. discussion of markets for drugs treating hypertension, Para. 11 onwards).
10. At the time of writing, the Court is expected to deliver its judgment in Spring 2009.