Control of chemotherapy-induced nausea and vomiting can be viewed as a quality of life issue and as an economic issue. The fact that emesis will have a negative impact on quality of life is intrinsically obvious and has been formally demonstrated by Coates et al , who determined that vomiting and nausea are the two toxicities of chemotherapy most feared by cancer patients. However, economic analysis of emesis (and therefore pharmacoeconomic analysis of antiemetics) is more difficult.
Viewpoint is a particularly important issue. From the viewpoint of the physician, cure of tumor and increase in survival may be the primary goals even at the price of severe but nonlethal toxicity. Even if the importance of supportive care is recognized, the physician may concentrate on potentially lethal toxicities such as neutropenic fever or renal insufficiency. From the viewpoint of the patient, however, prevention of nausea and vomiting may provide the greatest value for investment of the health care dollar.
Even accepting the importance of antiemesis, evaluation can be hampered by limitations in present pharmacoeconomic methodology. Standard definitions used for cost-effectiveness analysis and for calculation of incremental cost-effectiveness ratios often depend upon a denominator that measures increased survival . Thus, the value of supportive care measures, which increase quality but not duration of survival, may be severely underestimated.
There is no question that nausea and vomiting increase the cost of medical care. In 1993, O'Brien et al  evaluated the costs of nausea and vomiting both in terms of quality of life and in terms of direct and indirect costs required to respond to chemotherapy-induced emesis. This prospective study was performed in five Canadian hospitals prior to the introduction of 5-HT3 antagonists and used both serial administration of a quality of life instrument (Functional Living Index-Emesis) and a 5-day diary to estimate costs of medical resources. Using standard antiemetics, 41% of patients experienced emesis on day 1 and 78% experienced emesis at some time during the 5-day study period, leading to a significant decrease in functional status. Conversion of the patient diary into direct costs (nursing time, physician time, hospital admissions, salvage antiemetics) and indirect costs (out-of-pocket medical expenses and lost production) resulted in a mean cost per patient of approximately $200 Canadian.
The necessity of performing a formal pharmacoeconomic analysis of antiemetics has been driven both by a general concern about the cost of medical care and by the fact that antiemetics (particularly the 5-HT3 antagonists) have come to account for a significant portion of the pharmacy budget of the oncology hospital or ward (Figure 1). Several models have therefore been proposed for the analysis of such costs.
Buxton and O'Brien  created a probability tree model that took into account probability of significant emesis, probability of a significant antiemetic side effect, probability of the side effect being treated, and probability of resolution of the side effect, to produce 10 different possible outcomes (Figure 2). Outcome data from a large multicenter double-blind crossover study comparing ondansetron(Drug information on ondansetron) (Zofran) with metoclopramide(Drug information on metoclopramide)  were then inserted into the model to define the probability of each outcome.
Finite probabilities ranged from 67% for patients having no emesis and no significant side effects with ondansetron, to 1% for patients having significant emesis and refractory side effects from ondansetron. Cost estimates were obtained from standard price data for the British health care system. Direct comparison of costs suggested that the reduction of emesis resulted in a large enough saving to justify pricing ondansetron at 2.3 times the cost of metoclopramide, while calculation of cost effectiveness (cost per successful outcome) suggested that ondansetron could be priced at 5.1 times the cost of metoclopramide.
Jones et al  looked at a more general model of antiemetic management of patients receiving all forms of chemotherapy at a British cancer center and estimated the potential economic impact of administering a 5-HT3 antagonist-either ondansetron or granisetron(Drug information on granisetron) (Kytril)-as first- or second-line treatment to selected groups of patients at most risk of insufficient protection or unacceptable toxicity with standard antiemetic agents (Figure 3).
Even without considering costs of response to emesis, this model suggested that the 28% of patients at highest risk of significant emesis or toxicity with standard antiemetics (patients over 30 years old who had failed standard antiemetic therapy and all patients under age 30, for both highly and moderately emetogenic chemotherapy) could receive these more expensive agents with only a 3.1% increase in the overall chemotherapy budget. If patients over 30 years of age receiving highly emetogenic chemotherapy for the first time also received a 5-HT3 antagonist antiemetic, the overall increase in the chemotherapy budget was still only 3.5%. Significant increases in total cost were not noted unless first-line treatment with a 5-HT3 antagonist was also offered to patients over age 30 receiving moderately emetogenic chemotherapy. This model was most sensitive to changes in the cost of the 5-HT3 antagonist antiemetic itself (as opposed to emetogenicity of chemotherapy, age of patients, or efficacy of standard antiemetics). Antiemetic cost, in turn, can be affected either by unit cost of the antiemetic agents or by amount (and appropriateness) of antiemetic dose.
The sensitivity to specific dose and usage recommendations cannot be overemphasized when such analyses are performed . Intravenous ondansetron has been recommended for highly emetogenic chemotherapy at doses of either 8 mg or 32 mg, while intravenous granisetron is variously recommended at doses of 10 mcg/kg in the United States or 40 mcg/kg in Europe. In addition, the value of any 5-HT3 antagonist in the setting of delayed nausea and vomiting has been questioned, thus raising the possibility of achieving significant economic savings through the elimination of chronic oral usage.
Several investigations have now appeared that review postmarketing usage of ondansetron in adult and pediatric populations [8-10]. Approximately 15% of orders are written for indications not currently approved by the FDA (usually due to emetogenicity of the chemotherapy or age of the patient). However even among those patients in whom the therapeutic indication was considered to be appropriate, approximately 40% received doses other than those formally approved. In most of these cases, this represented a dosage higher (and therefore more expensive) than recommended. Although restriction to the recommended indications and doses would result in a cost saving, such a strategy would limit the ability to take advantage of new indications and would limit the flexibility to respond to individual patient situations.
Direct comparisons have now been made of the cost and cost effectiveness of antiemetic protection with a 5-HT3 antagonist-based regimen, compared with that of a metoclopramide-based regimen. Direct costs for such an analysis can include antiemetic drug acquisition cost, antiemetic administration cost, materials for clean-up after emesis, salvage antiemetic acquisition cost, nursing staff time, medical staff time, and potential for increased hospitalization . Most such studies have included calculation not only of costs but also of costs prorated to the percentage of successfully treated patients.
Cunningham et al  studied 32 patients receiving highly emetogenic chemotherapy who were randomized to receive ondansetron or metoclopramide. Outcome was classified based on significant emesis (more than one vomiting episode) and on adverse events (Figure 4). Ondansetron produced 7 of 14 (50%) successful treatments (no more than one vomiting episode and no adverse events), compared with 4 of 18 (22%) successful treatments for metoclopramide. The ondansetron/metoclopramide ratio for drug costs was found to be 3.7:1. However, this ratio decreased to 2.3:1 when the costs of drug administration and management of emesis were included. When a cost-effectiveness ratio (cost per successfully treated patient) was calculated, the ratio further decreased to 1.03:1.
Ballatori et al  performed a retrospective cost analysis for 289 patients receiving cisplatin (Platinol) who were randomized to receive ondansetron/dexamethasone or metoclopramide/dexamethasone/diphenhydramine. Although the drug acquisition cost ratio (ondansetron combination/metoclopramide combination) was 5.23:1, the ratio decreased to 3.21:1 when costs of emesis management were included, and then to 2.43:1 when a cost-effectiveness ratio was calculated.
An additional study  compared 150 patients receiving cyclophosphamide(Drug information on cyclophosphamide)-based therapy who received either intravenous ondansetron/dexamethasone followed by oral ondansetron for 5 days or intravenous metoclopramide/dexamethasone followed by oral metoclopramide for 5 days. Outcome was classified according to the presence of vomiting, the presence of adverse events, and the intensity of treatment that was required for the adverse events (Figure 5). Although the ondansetron/metoclopramide ratio for drug acquisition cost was approximately 15:1, the ratio for costs of antiemetic management was 2.93:1 and the cost-effectiveness ratio was 1.15:1.
A preliminary analysis using granisetron has been performed by Kirchner et al . Twenty-three patients receiving 5-day fractionated chemotherapy were given either granisetron or metoclopramide/dexamethasone. Looking only at costs for acquisition and administration of antiemetics, the granisetron/metoclopramide ratio was 1.07:1. However, a majority of the patients receiving metoclopramide were withdrawn for adverse events or lack of efficacy, whereas no granisetron recipients were withdrawn.
Economic analysis of health care costs requires that comparable units be used. Although cost per effectively treated patient may provide an estimate of the relative efficacy of various antiemetic interventions, such analyses will be at a disadvantage when compared to incremental analyses based on increased patient survival. However, this problem may be addressed through the use of cost-utility analysis. Standard cost-effectiveness incremental analysis addresses the cost per increased year of survival, but cost-utility analysis addresses the cost per increased quality-adjusted year of survival. [Quality-adjusted life years (QALYs) = utility score × years of survival]. An intervention that increases quality but not duration of survival (such as an effective supportive care measure) will still increase the utility score and number of QALYs and therefore provide a finite and comparable denominator for cost-utility analysis. The major methodologic problem for this strategy then becomes definition of the utility score and the time at risk.
Zbrozek et al  recently attempted to perform a cost-utility analysis comparing ondansetron and metoclopramide for the prevention of cisplatin(Drug information on cisplatin)-induced vomiting. The database for efficacy and toxicity was derived from a metaanalysis of published clinical trials in which chemotherapy-naïve patients received cisplatin with either ondansetron or metoclopramide. Labor and material estimates from the author's institution and published drug prices were used to construct cost estimates. The metaanalysis suggested that patients receiving ondansetron had a mean of 2.03 vomiting episodes, compared to 2.69 vomiting episodes with metoclopramide, a decrease of 25%. It was assumed that this decrease in emesis accounted for only a 0.05 unit difference in utility score and that the period of risk for change in utility was just the one day of chemotherapy. The prorated incremental annual improvement was therefore only 0.00014 QALY (0.05 utility unit × 1 day × 1 year/365 days). However, both of the assumptions used in this calculation may be questioned.
We recently attempted to obtain a preliminary estimate of a utility score reflecting antiemetic protection . A series of 31 patients receiving cyclical chemotherapy in our clinic completed a pair of 100 mm visual analog scales requesting a global ranking of quality of life during the previous cycle of chemotherapy with or without antiemetic protection (all other toxicities of chemotherapy remaining unchanged). The median visual analog scale score for quality of life with "no vomiting" was 88 mm vs 20 mm with "vomiting" (see Table 1). Conversion from a visual analog scale score range (0-100 mm) to a utility score range (0-1) suggests an incremental improvement of 0.68 units.
If patients valued a 25% reduction in vomiting (the benefit of ondansetron compared with metoclopramide) one quarter as much as they valued complete elimination of vomiting, then this benefit would improve their utility by 0.17 unit (0.68 unit × 25%). It is reasonable to assume that quality of life during chemotherapy is affected by the presence and anticipation of vomiting for more than one day. Our study evaluated quality of life during an entire cycle of chemotherapy, and a full course of chemotherapy might be expected to last for 6 months. The appropriate calculation for prorated incremental improvement would therefore yield a value of 0.085 QALY (0.17 utility unit × 0.5 year).
This difference in incremental improvement is not trivial. Zbrozek et al  noted an additional cost of $56 for ondansetron for a 70-kg patient and suggested that the incremental cost for use of ondansetron was unacceptably high with a value of $407,667/QALY ($56/0.00014 QALY). Six cycles of chemotherapy would result in an additional drug acquisition cost of $336 (6 × $56). However, using the revised estimate of incremental improvement in quality of life, this back of the envelope calculation suggests that the incremental cost for the use of ondansetron in this study population is only $3,953/QALY ($336/0.085 QALY), an amount that would be considered extraordinarily inexpensive for this level of gain and would suggest significant value (and therefore justification for investment of resources) for aggressive antiemetic intervention.
These examples have suggested both the potential and the pitfalls inherent in the use of pharmacoeconomic analysis principles in the area of supportive care. Further pharmacoeconomic analyses of supportive care measures should attempt to use accurate standard definitions so that an appropriate priority for such measures in the setting of limited or controlled health care resources can be determined.