X hits on this document





23 / 39

and construct the clinical trials that will define the attributes years before the products are approved and marketed. It is quite plausible, therefore, that at any given point in time there could more differentiation on certain attributes than others.

Two of the side effect attributes have negative coefficients and three have positive coefficients in Table 2. Cancer patients often take drugs that ameliorate the impact of certain side effects, such as pain, nausea, and vomiting. In addition to the explanation above, this may be another reason why physicians seem to assign a positive value to the abdominal pain, nausea, and vomiting side effects. If a physician prescribes anti-pain and antiemetic drugs in conjunction with the anti-cancer drugs, she may downgrade the importance of these side effects when choosing a regimen. A third possible explanation for the positive coefficients on three of the side effect attributes is that physicians may believe that the efficacy of the newer drugs are better than the measures reported in phase 3 clinical trials. This could occur, for example, if physicians use the drugs differently in practice than as they were used in the trials due to learning about patient-drug matching. Because the newer drugs are more toxic and generally have greater side effects, the physician beliefs would be captured as positive coefficients on the side effect measures.

The very large positive coefficient on the second-line indicator variable con- firms that these therapies are priced much higher than first-line therapies, con- trolling for the attributes of each. This implies there is a much higher willingness to pay for an extra life year for the very sick relative to the less sick. Finally, reg- imens containing the capecitabine tablet are priced slightly lower than regimens that use 5-FU/LV instead. This does not necessarily imply that physicians do not value the convenience associated with the tablet but also less dosage of other drugs. As indicated in Appendix 1, the dosages of certain branded drugs (e.g., irinotecan) are lower when capecitabine is used instead of 5-FU/LV.

We exponentiate the coefficients on the quarter indicator variables (the first


Document info
Document views117
Page views117
Page last viewedWed Jan 18 08:07:38 UTC 2017