How do doctors know which drugs to give to which patients? Of course there are clinical trials giving the relative efficacy of each drug. With less than perfect adherence, however, clinical trials may not accurately predict a drug’s efficacy or the potential side effects.
A paper by Chintadunta, Jiang and Jin (2008) look at two types of physician learning about pharmaceuticals: learning across patients and learning within patients. Across patient learning occurs when the physician learns about the average quality of a drug. Physicians figure out which drugs are best suited for individual patients through within-patient learning.
The authors look at physician learning in the setting of Cox-2 Inhibitors:
“Between 1998 and 2001, the FDA approved three Cyclooxygenase-2 (Cox-2) Inhibitors: Celebrex (Dec. 1998), Vioxx (May. 1999), and Bextra (Nov. 2001). All of them were heavily advertised as safer alternatives to then existing pain killers. By September 2004, the class had more than 10 million patients, annual sales had reached $6 billion in 2003, and total advertising dollars spent in 2003 were as high as $400 million. After a clinical trial associated Vioxx with severe cardiovascular (CV) risks, Merck withdrew the blockbuster drug in September 2004. CV risks and enhanced concerns on skin irritation led to the withdrawal of Bextra in April 2005. As of today, Celebrex is the only Cox-2 Inhibitor remaining on the market, with warnings added in April 2005. ”
Data and Methods
The authors use data from four sources to identify physician learning:
- patient-level prescription and satisfaction data from the IPSOS patient diary database (IPSOS-PD),
- monthly advertising expenditures obtained from the New Product Spectra (NPS) database,
- the number of news articles covering Cox-2s derived from Lexis-Nexis for the period 1999 to 2005, and
- the number of academic articles covering Cox-2s from Medline from 1999 to 2005.
Patient level satisfaction is importance because the more satisfied a patient is with the prescription, the less likely the physician will decide to switch brands. Physician learns which brands are well-suited to which type of patients. Advertising, media coverage and academic articles all play a roll in across-patient learning.
The authors use a Bayesian model to identify learning. The authors assume each physician has a prior about the relative efficacy and safety of each of the Cox-2 Inhibitors. Patient satisfaction, advertising, media coverage, and academic articles all will update the posterior probabilities.
The utility of each of patient, p, from using each drug, j, at time, t, is:
- Upjt = β0j + βsE(satis)jt + βxjXpt + βzZjt +εpjt
The patient’s utility depends on a drug specific constant, the patient’s average satisfaction rating up to that date, patient specific factors, X, and drug specific information, Z.
The authors find that there is significant physician learning based on the evolution of patient satisfaction measures. Other types of information seem to have less of an effect on physician prescribing behavior. Physicians hold strong priors which leads to slow updating.
Interestingly, “News articles have a positive inﬂuence on prescriptions, no matter whether these titles sound negative or non-negative. This suggests that the major role of news articles is informing doctors/patients of the existence of Cox-2s, rather than revealing the quality of Cox-2s…In contrast, a medical article about Cox-2s has a signiﬁcant negative impact on prescription sales, even if its title and abstract are non-negative.”
Thus, the authors find that doctors learn more within-patient than across patient.
FDA updates have no impact on physician behavior. How can this be? By the time the FDA issues a warning on the risks of the potential risk of different drugs, there almost always have been academic articles published on these issues. This is not to say that the FDA updates are useless, only that they often come after the academic research has already taken place and the physicians have already updated their priors.
For policy makers, the conclusions of this study may be worrisome for proponents of evidenced based medicine. Currently, physician learning from new information is very slow. Secondly, physicians seem to focus on tailoring treatments to individual patients rather than updating whether or not the treatment is worthwhile in the first place.
- Pradeep Chintadunta, Renna Jiang, Ginger Z. Jin (2008) “Information, Learning, and Drug Diffusion: the Case of Cox-2 Inhibitors” NBER WP #14252.