Article Tools
Phase I and Clinical Pharmacology
Article Tools
Using Patients As Their Own Controls for Cost Evaluation of Phase I Clinical Trials
Abstract
Little is known about the cost of phase I trials in cancer patients compared with that of standard treatments, yet the former is often assumed to be greater than the latter. Our objective was to utilize a new approach, using patients as their own controls, to compare in a pilot study the costs of care for patients on phase I trials with those incurred for standard treatment.
We retrospectively assessed the direct medical costs (DMCs) of 59 patients participating in one of two phase I trials (TRIAL) in solid tumors conducted at Memorial Hospital (MH): (1) perillyl alcohol, and (2) flavopiridol with paclitaxel. Paired-control DMCs were those accrued by the same patient while receiving standard chemotherapy regimens just before (PRE; n = 41) or after (POST; n = 29) the trial at MH, averaged per day.
For the 41 PRE patients, the median and mean DMCs per day for the clinical trial versus standard treatment were (US $) $123 v $133 and $219 v $267, respectively. For the 29 POST patients, the median and mean DMCs for the clinical trial versus standard treatment were $157 v $152 and $226 v $226, respectively. Using a linear mixed model, there was no significant difference between TRIAL and standard treatment DMCs (P = .54).
Cancer care direct medical costs (DMCs) were estimated to be (US $) $35 billion in 1990,1 and there is concern about the effect of clinical trial enrollment on these costs. Since clinical trial participation is often assumed to cost more than standard treatments, many insurers have reservations about payment for subscriber participation in such studies. More than 3,000 patients were denied access to clinical trials in 1998, a number expected to rise.2 Despite recent changes in Medicare's policy toward reimbursement for clinical trials,3 and data suggesting that participation in cancer clinical trials only minimally increased, if at all, the costs compared with those associated with standard care,4-6 the issue remains one of great health policy importance.
Phase I studies, the most preliminary of clinical trials in humans, arguably are among those most vulnerable to restrictive reimbursement policies. Many patients participating in such studies have refractory cancer and have been heavily pretreated. Dose determination and definition of toxicity profiles are the primary end points, not efficacy. Anticipated response rates are low.7 Clinical correlative studies and other ancillary evaluations are frequently part of these protocols, so the assumption that participation in such studies will cost payers more than standard treatments would seem obvious.
Unfortunately, there are limited data available that address this question. Relevant previously published studies have not focused on phase I trials.4-6 Accordingly, the cost of participating in phase I trials compared with standard treatment, which may be different than for phase II and III trials, is not well described.
A randomized trial examining this issue would be the ideal, but is likely not feasible. Other studies comparing trial and standard treatment costs4-6 have used a matched case-control method to compare costs. However, matching case patients with controls is fraught with challenges, as all potential confounders are difficult, if not impossible, to address. In our deliberations as to how to best evaluate phase I trial costs, it became apparent that since many patients receive standard treatments before and/or after their participation in a phase I clinical trial, we could use patients as their own controls8 for cost-assessment comparisons to potentially obviate some of these confounding concerns. This methodology has been used in many clinical studies,9-15 and in a prior clinical economic study,16 but experience with it in the cancer therapeutic or phase I clinical trial settings is very limited. However, this approach has been endorsed by Karnon and Qizilbash17 to be used in future economic studies.
We report here a pilot study using this methodology, which examined these costs from a payer's perspective in patients participating in one of two phase I clinical trials at Memorial Sloan-Kettering Cancer Center (MSKCC).
All patients evaluated had been enrolled in one of two phase I trials, both of which were approved by the institutional review board at MSKCC: (1) perillyl alcohol,18 a monoterpene (trial A); or (2) flavopiridol, a protein kinase C inhibitor, and paclitaxel19 (trial B). Based on our knowledge of phase I clinical trials available at our institution and elsewhere, these studies were chosen as representative examples with regard to treatment intensity, potential toxicity, and content. Trial A investigated an oral agent; trial B studied the combination of a non–US Food and Drug Administration–approved agent (given intravenously over 24 hours) with a US Food and Drug Administration–approved drug. This convenience sample included all patients who had completed the respective clinical trial by the time of our analysis, and for whom cost data were available. Specifically, the sample included 18 of 22 patients entered onto trial A (cost data were unavailable on four patients: three who never received perillyl alcohol, or discontinued study shortly after starting perillyl alcohol; and one patient, for unknown reasons), and the first 41 of 48 patients entered onto trial B, which was still ongoing. Both trials were done at MSKCC with similar inclusion criteria. All patients were required to have a pathologically confirmed solid tumor; age ≥ 18 years; platelets ≥ 100,000 cells/mm3; and been off previous chemotherapy, radiotherapy, and immunotherapy for 4 weeks before study entry (6 weeks if they had received mitomycin or nitrosoureas). Pregnant women were excluded. Exclusion differences between the protocols are listed in Table 1. Informed consent was required.
In trial A, perillyl alcohol was administered four times daily by mouth for 28 consecutive days in the ambulatory setting, followed by disease reassessment. Planned dose escalation levels of perillyl alcohol ranged from 4,800 mg/m2/d to 14,800 mg/m2/d in divided doses.
In trial B, on day 1 of the cycle, patients were given paclitaxel at a dose of 135 mg/m2 or 175 mg/m2 as a 3-hour infusion. On day 2, patients received flavopiridol as a 24-hour infusion in the outpatient setting. For this purpose, MediPort, Broviac, or peripherally inserted central catheter access was required. Planned dose escalation levels of flavopiridol ranged from 10 to 94 mg/m2/d. Each cycle was repeated every 21 days, with disease reassessment after two cycles.
All patients underwent a physical examination and toxicity evaluation weekly for the first 4 months of the trial, then every 2 weeks thereafter. For the first 6 weeks, patients had weekly laboratory tests (CBC, electrolytes, blood urea nitrogen, creatinine, liver function tests), and then had these tests every other week. When disease was measurable, tumor assessments were recorded by the appropriate method every 4 weeks.
Each week during the first two cycles, then only at the initiation of each cycle, the following evaluations were performed: physical examination and toxicity evaluation by a physician and routine laboratory studies (CBC, electrolytes, blood urea nitrogen, creatinine, liver function tests). A measurement of the tumor indicator lesion by the appropriate method was to be performed after every two cycles of therapy.
Other research procedures (eg, pharmacology studies) were not charged to the patient in either study.
Charges for all inpatient and outpatient medical care at MSKCC were collected from the billing department at the hospital. The charges were then converted to costs using individual departmental cost-to-charges ratios (RCCs), since charges are not a good proxy of economic costs.20 A list of the departmental RCCs for MSKCC in 1998 is presented in Table 2. The total costs included blood work, ECGs, cardiac and pulmonary tests, drugs/intravenous fluids/clinical nutrition, laboratory tests, medical supplies, radiographic studies, operating room expenses, radiation therapy, radioisotopes, room and board, and physical and occupational therapy. Physician billing charges at MSKCC were included and consisted of the amounts billed, but did not reflect insurance allowances or the amount collected. Charges were used in this situation because they may be a better proxy of transaction costs than Medicare reimbursement based on the resource-based relative value scale.21 All costs were discounted to 1998 US dollars.
Indirect and direct nonmedical costs were omitted, as they represent the costs to the patient and society, not the payer, and are more challenging to quantitate, particularly when done retrospectively.22 Also, we felt this narrower focus would not significantly undermine our objective of evaluating in a pilot study the utility of using patients as their own controls to compare costs of care for patients on phase I trials to those incurred for standard treatments. Only periods when the treatment took place at MSKCC were included. For each individual, the total cost during the time period was averaged per day. This was done to correct for the different lengths of time throughout each treatment period.
As this was a preliminary retrospective study, the costs of resources outside of MSKCC could not be comprehensively collected and analyzed, especially during the period when patients were receiving standard therapy. Partial data on radiographic studies and laboratory tests done outside MSKCC during the clinical trials are reported to present the potential impact of this omission.
Control treatment periods were those that occurred just before (PRE) or after (POST) the protocol period while the patient was receiving standard treatment (defined as US Food and Drug Administration–approved chemotherapy drugs while not participating in an active protocol). DMCs were collected from the time of treatment initiation until 3 weeks after the last cycle. A period of 3 weeks was chosen since many treatment regimens have 3-week cycles; therefore, the costs of follow-up studies and treatment of acute toxicity would likely be included, but not the costs of subsequent, different therapy. DMCs were collected and calculated as described above.
DMCs were included from the time of treatment initiation until the patient was removed from the trial (TRIAL). This end point for the TRIAL time period was chosen instead of the last day of treatment so that the costs of the final evaluation would be included in the analysis. Only costs charged to the patient or insurance company were gathered. Research costs (eg, experimental medication, pharmacokinetic studies, research study assistants) that were covered by other sources of funding supporting the research (eg, grants) were excluded, as they are not accounted for in a cost analysis from a payer's perspective.
Costs averaged per day were separated by periods collected (TRIAL, PRE, and POST) and underwent log transformation due to the skewness of the data. The PRE and TRIAL costs of the patients who received standard treatment before the clinical trial at MSKCC, as well as the POST and TRIAL costs of the patients who received standard treatment before the clinical trial, were compared using a Wilcoxon signed rank test23 (using SPSS version 9.0; SPSS Inc, Chicago, IL). A nonparametric test was chosen because even after log-transformation, the data were not normally distributed in each group.
The Wilcoxon signed rank test only considers pairs of observations (TRIAL and PRE, or TRIAL and POST). Yet, there is a subgroup of patients for which a triplet of data (PRE, TRIAL, and POST) is available. An approach that allows use of all these data simultaneously may be more efficient. One such approach, a linear mixed model24 (PROC MIXED Statistical Analysis Software Version 6.12 [SAS/STAT User's Guide Version 6]; SAS Institute, Cary, NC), allows TRIAL costs to be compared with the combination of PRE and POST costs. As we have mentioned, costs underwent log transformation to normalize the data.
While the linear mixed model technique is a more powerful statistical method, the Wilcoxon signed rank test is easier to understand. Therefore, both statistical approaches are presented. All P values correspond to two-sided testing for significance, with a significance level set at ≤ .05.
Fifty-nine patients who participated in either trial A (n = 18) or trial B (n = 41) were analyzed for DMCs. No patient participated in both studies. At the time of the initiation of investigational treatment, the median age was 53 years (range, 21 to 77 years) and the median Karnofsky performance status was 80% (range, 70% to 90%). A breakdown of tumor types by group (ie, TRIAL, PRE, POST) is presented in Table 3. Twelve (20%) of the 59 patients had a sarcoma.
Forty-one of the 59 patients had standard treatment at MSKCC before investigational treatment there (PRE group). The 41 patients received one of 21 different chemotherapy regimens (Table 4); 12 (29%) of 41 patients received a combination chemotherapy program. Twenty-nine of the 59 patients had standard treatment at MSKCC after investigational treatment there (POST group), which commenced an average of 5 weeks after removal from the phase I study. These 29 patients received one of 14 different chemotherapy regimens (Table 4), and combination chemotherapy was used in a smaller proportion (4 [14%] of 29 patients).
Twelve of the 59 patients did not receive standard chemotherapy at MSKCC before or after the investigational treatment. Twenty-three of the 59 patients had standard chemotherapy at MSKCC both before and after they received treatment on protocol, and the patients (with relevant PRE or POST standard treatment) were included in the PRE (ie, 23 of 41 patients) and POST (ie, 23 of 29 patients) groups. Accordingly, 18 (ie, 41 minus 23 patients) and six (ie, 29 minus 23 patients) patients, respectively, received standard chemotherapy either before (PRE group) or after (POST group) investigational treatment, but not in both settings.
The median time on treatment for the PRE group was 113 days (range, 21 to 1536 days). The median time on treatment for the POST group was 57 days (range, 21 to 394 days). The median time patients received investigational therapy was the shortest at 41 days (range, 1 to 415 days).
The median costs, averaged per day, and ranges for the TRIAL, PRE, and POST groups were $131 (range, $26 to $1,151), $133 (range, $23 to $5,148), and $152 (range, $32 to $844), respectively. This is summarized in Figure 1 and in Tables 5 and 6. In each group, costs ranged widely and were quite skewed. A breakdown of the major cost components showed that the largest component of DMCs by median value for the PRE group was physician billing charges (32% of total costs), while room and board charges were the second largest component (28%). However, for the TRIAL and POST groups, physician billing charges were only the second largest component of DMCs (22% and 23%, respectively), while room and board charges were the largest component (36% and 41%, respectively). For all three groups, intravenous fluid and medications were the third largest component of costs, while other components (eg, laboratory and radiological studies) made up a much smaller percent of the total DMCs.
Cost data were only collected for resources used at MSKCC, since only billing data were analyzed; therefore, the costs of all resources from outside centers were not included in our analysis. During the clinical trial, approximately 32% of computed tomography scan or magnetic resonance imaging studies, 5% of x-rays, 0% of ultrasounds, and 0% of positron emission tomography scans were done at outside centers. Required laboratory tests were done outside of MSKCC for only one patient. More comprehensive information on missing cost data of resources used while patients were receiving standard treatment are not available.
Of the 41 patients who received standard treatment before the clinical trial, the median and mean PRE and TRIAL costs were similar ($133 v $123 and $267 v $219, respectively), as summarized in Table 5. However, the standard deviation of the PRE costs ($789) was much higher than the standard deviation of the TRIAL costs ($265). In 22 of 41 patients, standard treatment costs were greater than clinical trial costs, while in the remaining 19 patients, clinical trial costs were greater than standard treatment costs. There was no significant difference (P = .53) by the Wilcoxon signed rank test.
Of the 29 patients who received standard treatment after the clinical trial, the median and mean POST and TRIAL costs were also similar ($152 v $157 and $226 v $226, respectively), as summarized in Table 6. The SD of the POST costs ($225) and the TRIAL costs ($270) were similar. In 13 of 29 patients, standard treatment costs were greater than clinical trial costs, while in the remaining 16 patients, clinical trial costs were greater than standard treatment costs. There was no significant difference (P = .66) by the Wilcoxon signed rank test.
Using a linear mixed model approach, no significant difference of average DMC was found between investigational therapy (TRIAL group) and standard therapy (PRE and POST groups; P = .54). There was also no difference when comparing the PRE and POST groups individually to the TRIAL group (P = .45 and P = .77, respectively). Interestingly, the cost of the POST group was also found not to be significantly different from the PRE group (P = .72).
Using a similar linear mixed model approach (after log transformation, “days on trial” was substituted for “costs”), the time on clinical trial was significantly less than the time for receiving standard therapy (P < .001). As noted previously, those disparities were addressed in our cost figures by averaging costs per day (Tables 5 and 6).
We demonstrated the feasibility of applying a new, efficient design where patients serve as their own controls for the assessment of clinical trial costs, and the results of our pilot study show no significant difference from a payer's perspective in the cost of participating in either of two phase I trials compared to standard treatment. Our pilot analysis was done from the payer's perspective,25 given the significant impact payers have on reimbursement policies for investigational therapies. Insurers are expecting the research community to justify their actions not only scientifically, but economically.26 However, studies examining clinical trial costs through other and broader perspectives (eg, patient, hospital, society) remain important and are warranted. Indeed, one would anticipate that if all costs are taken into account (eg, the cost of the investigational agent, the cost of pharmacy time, the cost of research study assistants and nurses, physician time discussing the clinical trial), the cost of doing a clinical trial may be more than the cost of standard treatment. For example, the estimated cost for the research study assistants utilized while the 59 patients received investigational therapy was $16 per day, a cost payers may never see that is nonetheless relevant from broader economic perspectives. Understanding the incremental cost of doing clinical trials would assist those responsible for these additional costs (eg, the pharmaceutical industry, federal granting agencies, the academic center, the principal investigator) in the determination of whether these added costs will be offset by the anticipated benefits from a study.
By using the cost of a patient's own treatment with standard chemotherapy as a control for the cost of phase I trial participation, many potential confounders that will affect matched-pair case-control studies are addressed. Concerns regarding changes in costs related to the progressive debilitation of the patient with additional treatment were addressed by considering both pretrial and posttrial standard therapies. There was a large range in the cost of each treatment group, related in part to the heterogeneity of patients and solid tumor types entered in these two phase I trials, as well as outliers against the backdrop of a relatively small denominator. Also, a variety of standard treatment regimens were used in the PRE and POST groups. Nonetheless, clinical trial costs were not significantly greater than standard treatment costs either before and/or after the clinical trial.
There are limitations to this study and its design. The first is that only a comparison of the cost during a single period of treatment, not throughout the entire care of the cancer patient, was done. Therefore, the impact that participation in a phase I trial has on the collective cost of palliative care was not addressed, nor how cost compared with supportive care alone. However, the data and method used provide an important first step toward addressing these larger issues.
A second limitation was that only medical costs accrued at MSKCC were analyzed in this pilot retrospective study. Radiographic and laboratory studies done outside of the hospital were not included. Our analysis of the patients on trial demonstrated that virtually all studies were done at MSKCC, with the exception of a minority (32%) of computed tomography and magnetic resonance imaging scans. Unfortunately we could not comprehensively collect such information for standard treatments since this information is not always readily accessible in the chart. However, the radiological and laboratory studies only accounted for a small fraction of the total DMC, and the vast majority of resources were used at MSKCC. Furthermore, we anticipate that physicians are more flexible in allowing patients off protocol to have these tests at a location of greatest convenience. Patients participating in clinical trials are often encouraged to have their tests done at the investigating hospital, especially diagnostic imaging studies, to facilitate efficient review by the reference radiologist. Accordingly, the omission of these costs would likely lead to a greater underestimate of the cost of standard therapy than the cost of investigative treatment. These “missing” costs should be investigated in a future study in order to examine the validity of this assumption.
A third limitation was that our analysis only examined 59 patients, of whom 47 received standard chemotherapy at MSKCC before and/or after their participation in one of two phase I trials. In order to ensure that no potential type II error has occurred, and determine the applicability of our results to the larger universe of phase I trials, we plan using this same methodology to examine a larger number of patients participating in a more comprehensive spectrum of phase I studies in the future. A much larger effort using a case-control methodology, the Cost of Cancer Treatment Study,27 has begun to obtain cost information for phase II and phase III, but not phase I, National Cancer Institute–sponsored clinical trials, and will use a matched case-control methodology. Using patients as their own controls represents an alternative approach to address questions of this nature, although post-treatment controls would likely be much more common than pretreatment ones given the more restrictive pretreatment eligibility criteria in phase II and III trials compared to phase I trials. This last observation highlights that there are settings where using patients as their own controls may apply less well. Accordingly, delineating and comparing the strengths and weaknesses of case-control and using patients as their own controls methodologies, represents another important direction of future research.
The authors indicated no potential conflicts of interest.
Fig 1. Direct Medical Costs per day; Log10 scale.
|
| Protocol A: Perillyl Alcohol | Protocol B: Flavopiridol and Paclitaxel |
|---|---|
| Karnofsky performance status < 70% | Karnofsky Performance Status < 60% |
| Peripheral leukocyte count < 3,000/μL | |
| Hemoglobin < 8 g/dL | Peripheral leukocyte count < 3,500/μL |
| Absolute neutrophil count < 1,500/μL | |
| Serum creatinine > 1.6 mg/dL | Serum creatinine > 1.5 mg/dL |
| Total serum bilirubin > 2 mg/dL | |
| Alkaline phosphatase > 230/μL | Serum AST and ALT > 2.5× upper limit of normal |
| Serum AST > 74/μL | |
| Patient taking antiseizure medications known to be metabolized by cytochrome P-450 enzyme family | History of cardiac disease within previous 6 months |
| Central nervous system metastasis or primary | |
| Pre-existing neurotoxicity of grade 3+ or greater | |
| HIV disease |
|
| Extended Services | Cost-Charge Ratio |
|---|---|
| Blood and blood processing | 0.6590 |
| Cardiopulmonary/ECG | 0.3162 |
| Drugs/IV/clinical nutrition | 0.5814 |
| Laboratory | 0.4489 |
| Medical supplies | 0.0348 |
| MRI/CT/ultrasound | 0.2833 |
| Operating room/recovery room | 0.8855 |
| Radiation therapy | 0.7847 |
| Radioisotopes | 0.6657 |
| Radiology diagnostic | 1.0759 |
| Room and board | 1.2441 |
| Speech therapy/PT/OT/respiratory therapy | 0.3026 |
| Total | 0.7464 |
Abbreviations: IV, intravenous; MRI, magnetic resonance imaging; CT, computed tomography; PT, physical therapy; OT, occupational therapy.
|
| Disease | TRIAL | PRE | POST | ||||||
|---|---|---|---|---|---|---|---|---|---|
| No. of Patients | % | No. of Patients | % | No. of Patients | % | ||||
| Sarcoma | 12 | 20 | 6 | 15 | 3 | 10 | |||
| Non–small-cell lung cancer | 9 | 15 | 9 | 22 | 5 | 17 | |||
| Colon | 8 | 14 | 6 | 15 | 6 | 21 | |||
| Stomach | 6 | 10 | 3 | 7 | 2 | 7 | |||
| Prostate | 4 | 7 | 3 | 7 | 3 | 10 | |||
| Esophagus | 4 | 7 | 3 | 7 | 2 | 7 | |||
| Melanoma | 3 | 5 | 3 | 7 | 1 | 3 | |||
| Pancreas | 3 | 5 | 2 | 5 | 1 | 3 | |||
| Hepatocellular carcinoma | 2 | 3 | 0 | 0 | 1 | 3 | |||
| Other | 8 | 14 | 6 | 15 | 5 | 17 | |||
| Total | 59 | 100 | 41 | 100 | 29 | 100 | |||
Abbreviations: TRIAL, one of the two phase I trials; PRE, before protocol period; POST, after protocol period.
|
| PRE | POST | ||||
|---|---|---|---|---|---|
| Regimen | No. of Patients | Regimen | No. of Patients | ||
| Carboplatin/paclitaxel | 1 | Doxorubicin | 1 | ||
| CAV | 1 | Cisplatin/fluorouracil | 1 | ||
| Cisplatin/fluorouracil | 2 | Fluorouracil ± leucovorin | 6 | ||
| Cisplatin/gemcitabine | 1 | Gemcitabine | 2 | ||
| Cisplatin/paclitaxel | 1 | Ifosfamide | 1 | ||
| Cisplatin/topotecan | 1 | Irinotecan | 4 | ||
| Cyclophosphamide | 1 | MAID | 1 | ||
| Dartmouth | 2 | Mitomycin | 3 | ||
| Doxorubicin | 4 | Mitomycin/fluorouracil | 1 | ||
| Estramustine/vinblastine | 1 | Mitomycin/vinblastine | 1 | ||
| Fluorouracil ± leucovorin | 5 | Methotrexate | 2 | ||
| Gemcitabine | 4 | Paclitaxel | 3 | ||
| Ifosfamide | 1 | Docetaxel | 1 | ||
| Irinotecan | 3 | Vinorelbine | 2 | ||
| Mitomycin | 2 | ||||
| Methotrexate | 2 | ||||
| Methotrexate/fluorouracil | 1 | ||||
| Paclitaxel | 4 | ||||
| Docetaxel | 2 | ||||
| Docetaxel/vinorelbine | 1 | ||||
| Vinblastine | 1 | ||||
Abbreviations: PRE, before protocol period; POST, after protocol period; CAV, cyclophosphamide/doxorubicin/vincristine; Dartmouth, carmustine/cisplatin/dacarbazine/tamoxifen; MAID, mesna/doxorubicin/ifosfamide/dacarbazine.
|
| PRE Standard Treatment | Phase I Treatment | |
|---|---|---|
| Median time on treatment, days | 113 | 35 |
| Median total direct medical costs per day, $ | 133 | 123 |
| Total direct medical costs per day, $ | ||
| Mean | 267 | 219 |
| 95% CI | 18 to 516 | 135 to 302 |
NOTE. This Table compares the 41 patients who received treatment on the clinical trial and also received standard treatment prior to the clinical trial.
Abbreviation: PRE, before protocol period.
|
| POST Standard Treatment | Phase I Treatment | |
|---|---|---|
| Median time on treatment, days | 57 | 41 |
| Median total direct medical costs per day, $ | 152 | 157 |
| Total direct medical costs per day, $ | ||
| Mean | 226 | 226 |
| 95% CI | 140 to 311 | 123 to 329 |
NOTE. This Table compares the 29 patients who received treatment on the clinical trial and also received standard treatment after the clinical trial.
Abbreviation: POST, after protocol period.
Presented in part at the 36th Annual Meeting of the American Society of Clinical Oncology, New Orleans, LA, May 20-23, 2000.
Dr Sherman is the recipient of the 2000 ASCO Fellowship Award for the highest-ranked abstract submitted by a fellow.
Authors' disclosures of potential conflicts of interest are found at the end of this article.
| 1. | Brown ML: The national economic burden of cancer. J Natl Cancer Inst 82::1811,1990-1814, Crossref, Medline, Google Scholar |
| 2. | Wade JL, Weinberg PD, Bennett CL: Managed care's effects on community clinical research. Oncology News International, volume 8, 1999. http://www.cancernetwork.com/journals/oncnews/n9905a.htm Google Scholar |
| 3. | Arnold K, Vastag B: Medicare to cover routine care costs in clinical trials. J Natl Cancer Inst 92::1032,2000, Crossref, Medline, Google Scholar |
| 4. | Wagner JL, Alberts SR, Sloan JA, et al: Incremental costs of enrolling cancer patients in clinical trials: A population-based study. J Natl Cancer Inst 91::847,1999-853, Crossref, Medline, Google Scholar |
| 5. | Fireman BH, Fehrenbacher L, Gruskin EP, et al: Cost of care for patients in cancer clinical trials. J Natl Cancer Inst 92::136,2000-142, Crossref, Medline, Google Scholar |
| 6. | Bennett CL, Stinson TJ, Vogel V, et al: Evaluating the financial impact of clinical trials in oncology: Results from a pilot study from the Association of American Cancer Institutes/Northwestern University Clinical Trials Costs and Charges Project. J Clin Oncol 18::2805,2000-2810, Link, Google Scholar |
| 7. | Smith TL, Lee JJ, Kantarjian HM, et al: Design and results of phase I cancer clinical trials: Three-year experience at M.D. Anderson Cancer Center. J Clin Oncol 14::287,1996-295, Link, Google Scholar |
| 8. | Weiss NS, Heckbert SR: Patients as their own controls in studies of therapeutic efficacy. J Gen Intern Med 3::381,1988-383, Crossref, Medline, Google Scholar |
| 9. | Harrison WM, Endicott J, Nee J: Treatment of premenstrual dysphoria with alprazolam: A controlled study. Arch Gen Psychiatry 47::270,1990-275, Crossref, Medline, Google Scholar |
| 10. | Phillips BA, Schmitt FA, Berry DT, et al: Treatment of obstructive sleep apnea: A preliminary report comparing nasal CPAP to nasal oxygen in patients with mild OSA. Chest 98::325,1990-330, Crossref, Medline, Google Scholar |
| 11. | Fauvel JP, Velon S, Berra N, et al: Effects of losartan on renal function in patients with essential hypertension. J Cardiovasc Pharmacol 28::259,1996-263, Crossref, Medline, Google Scholar |
| 12. | Frederiksen SG, Johansson J, Johnsson F, et al: Neither low-calorie diet nor vertical banded gastroplasty influence gastro-oesophageal reflux in morbidly obese patients. Eur J Surg 166::296,2000-300, Crossref, Medline, Google Scholar |
| 13. | Mahon J, Laupacis A, Donner A, et al: Randomised study of n of 1 trials versus standard practice. BMJ 312::1069,1996-1074, Crossref, Medline, Google Scholar |
| 14. | Haines DR, Gaines SP: N of 1 randomised controlled trials of oral ketamine in patients with chronic pain. Pain 83::283,1999-287, Crossref, Medline, Google Scholar |
| 15. | Jansen IH, Olde Rikkert MG, Hulsbos HA, et al: Toward individualized evidence-based medicine: Five “N of 1” trials of methylphenidate in geriatric patients. J Am Geriatr Soc 49::474,2001-476, Crossref, Medline, Google Scholar |
| 16. | Hammond CM, Pierson JF, Grande TP, et al: Economic evaluation of risperidone in an outpatient population. Ann Pharmacother 33::1160,1999-1166, Crossref, Medline, Google Scholar |
| 17. | Karnon J, Qizilbash N: Economic evaluation alongside n-of-1 trials: Getting closer to the margin. Health Econ 10::79,2001-82, Crossref, Medline, Google Scholar |
| 18. | Azzoli CG, Miller VA, Ng KK, et al: A phase I trial of perillyl alcohol in patients with advanced solid tumors: Cancer Chemother Pharmacol 51::493,2003-498, Medline, Google Scholar |
| 19. | Schwartz GK, O'Reilly E, Ilson D, et al: Phase I study of the cyclin-dependent kinase inhibitor flavopiridol in combination with paclitaxel in patients with advanced solid tumors. J Clin Oncol 20::2157,2002-2170, Link, Google Scholar |
| 20. | Finkler SA: The distinction between cost and charges. Ann Intern Med 96::102,1982-109, Crossref, Medline, Google Scholar |
| 21. | Miller ME, Zuckerman S, Gates M: How do Medicare physician fees compare with private payers? Health Care Financ Rev 14::25,1993-39, Google Scholar |
| 22. | Sherman EJ, Pfister DG, Ruchlin HI, et al: The COIN form: A new tool to collect the invisible costs of hormone-refractory prostate cancer. Cancer 91::841,2001-853, Crossref, Medline, Google Scholar |
| 23. | Wilcoxon F: Individual comparisons by ranking methods. Biometrics Bull 1::80,1945, Crossref, Google Scholar |
| 24. | McCullagh P, Nelder J: Generalized Linear Models (ed 2) . London, United Kingdom, Chapman and Hall, 1989 Google Scholar |
| 25. | Eisenberg JM: Clinical economics: A guide to the economic analysis of clinical practices. JAMA 262::2879,1989-2886, Crossref, Medline, Google Scholar |
| 26. | Marwick C: Insurers, employers ask for better evaluation of new technologies. JNCI 94::250,2002-251, Crossref, Medline, Google Scholar |
| 27. | Goldman DP, Schoenbaum ML, Potosky AL, et al: Measuring the incremental cost of clinical cancer research. J Clin Oncol 19::105,2001-110, Link, Google Scholar |
