Materials and Methods

To estimate the costs associated with use of chemoresponse assay results for treatment of recurrent ovarian cancer, we established 3 groups of patients: (1) a previously reported group of patients who underwent testing with the ChemoFx chemoresponse assay prior to receiving their physician’s choice of chemotherapy regimen (assay assisted); (2) a modeled group undergoing testing with the ChemoFx assay with all treatment based on assay results (assay adherent) ( Figure ); and (3) a modeled group with choice of chemotherapy regimen based on prevalent clinical practices during a similar time period to accrual of the assay-assisted group (empiric). The model was designed from a payor perspective; time horizon was 6 months and as such, costs were not discounted. This study is exempt from institutional review board approval due to the use of anonymous aggregated data.

Flow diagram

Flow diagram for creation of 3 cohorts.

ChemoFx; Precision Therapeutics, Pittsburgh, PA.

Havrilesky. Chemotherapy treatment costs for recurrent ovarian cancer. Am J Obstet Gynecol 2010.

Of the 256 previously described patients with ovarian cancer who underwent testing with the ChemoFx assay from 1997 through 2002, the assay-assisted group consisted of the 77 who had recurrent disease. Two patients from that study were excluded from current cost analysis because they were treated with regimens that are not in current clinical use for ovarian cancer, which made our claims-driven cost calculations unfeasible (one 5-fluorouracil and irinotecan, 1 ifosfamide, paclitaxel, and topotecan). The ChemoFx assay tested an average of 7.7 chemotherapy drugs per patient on tumor explants.

The empiric group was modeled in 2 steps. First, using available data on concurrent market shares for specific chemotherapy agents and combinations of agents, we estimated the proportion of patients who would receive each treatment for ovarian cancer (“Disease Dynamics: The Cancer Market,” UBS Warburg market share data report, November 2000). This time frame was chosen because it was similar to the time frame of treatment for the assay-assisted group. To assign 2007 costs to these 2 groups, we used outpatient claims filed from Jan. 1, 2006, through Dec. 31, 2007, for 15 predetermined chemotherapy regimens ( Table 1 ) for patients with ovarian cancer from commercial private health plans (>5 million individuals; database owned and maintained by The Lewin Group, Ingenix Consulting, Falls Church, VA; 2008) as follows. Patients with International Classification of Diseases, Ninth Revision, Clinical Modification codes for ovarian cancer-primary malignant (183.0), malignant neoplasm of ovary and uterine adnexa (183.8), ovarian carcinoma in situ (233.3), and ovarian neoplasm of uncertain behavior (236.2) were considered to have been diagnosed with ovarian cancer and their data were included. For each eligible patient, deidentified, Health Insurance Portability and Accountability Act-compliant patient identification numbers were assigned to allow extraction and analysis of all claim records for specific patients. The sequence of treatments for each patient identification number was reconstructed based on specific claim history. Cycles of specific regimens were identified on the basis of the combinations of drugs administered during a rolling 28-day observation period. Patients were assumed to have been treated with a single-agent regimen during any 28-day period if they received a drug on ≥2 dates that were at least 7 but not >28 days apart and received no other chemotherapy agent listed in Table 1 during that time. Patients were assumed to have been treated with a multiagent regimen if they received an anchor drug, identified in Table 1 as being used to define an encounter, on ≥2 dates between 7-28 days apart and received a secondary twice or more during the same 28-day period. Start and end dates for each course were established by defining the start as the date of the earliest claim for the single drug in the case of single-agent regimens or the anchor drug in a multiagent regimen and the end as either 1 day before the start of a new regimen or 28 days after the last claim for the drug or anchor drug, depending on which came first. After identifying start and end dates, all claims for health care services delivered in outpatient or ambulatory care settings, self-administered prescription drugs, and other support prescription drugs were totaled (drug cost table). These were filtered to eliminate claims associated with services unequivocally unrelated to either the direct treatment of ovarian cancer or supportive care. Examples of items eliminated include influenza vaccines, prescription drugs for treatment of hypercholesterolemia, second-generation nonsedating antihistamines, podiatry visits, cataract surgeries, and laboratory tests for specific pathogens such as West Nile virus.

For the empiric group, we calculated an aggregate cost by adding the cost for 6 cycles of each chemotherapy agent, determined as described above, weighted by its respective market share percentage in 2000 ( Table 1 ). The total treatment cost was the sum of the costs for hematopoiesis-stimulating agents (HSAs), support drugs, other medical costs, and total cost for chemotherapy agents.

For the assay-assisted group, we calculated an aggregate cost as the sum of the cost of 6 cycles of each chemotherapy regimen administered weighted by its percentage of the aggregate cohort, plus costs for HSAs, support drugs, and other medical administration. The cost of ChemoFx assay for 7.7 drugs (the average number of drugs tested per patient in the data set) was additionally incorporated into the total average cost for this group. Approximately 65% of the 75 patients were given a drug that was scored in their highest possible category and were therefore considered adherent to assay results.

For the assay-adherent group, we assumed that all treatment decisions would be made based on the results of the ChemoFx assay. The cost for a chemotherapy regimen was estimated for each patient by using the drug that was scored as the most responsive in the ChemoFx assay for each of the 75 patients described in the assay-assisted group. If the patient had multiple drugs scored as responsive, the mean cost across all responsive drugs was used to estimate the total cost for chemotherapy agents. Costs for HSAs, support drugs, and other medical administration as well as the cost of the ChemoFx assay for an average of 7.7 drugs were included.

To assess which variables had a significant impact on the cost model’s results, sensitivity analyses were conducted around 3 key variables: market share estimates, the number of chemotherapy cycles administered, and the number of chemotherapy drugs tested in the assay.