The importance of clinical variables in comparative analyses using propensity-score matching: the case of ESA costs for the treatment of chemotherapy-induced anaemia.

BACKGROUND: The erythropoiesis-stimulating agents (ESAs) epoetin alfa (EA) and darbepoetin alfa (DA) have comparable efficacy in treating chemotherapy-induced anaemia (CIA). Therapy choice depends on many factors, including cost. Previous estimates of ESA cost differences have been derived from claims data. These data lack clinical variables, such as baseline haemoglobin (Hb) level, which are likely to influence choice of ESA, dosing and costs. We estimated cost differences between DA and EA in patients with cancer receiving chemotherapy, using a propensity-score matched analysis of baseline patient characteristics with and without Hb values to assess the effect of this clinical variable on ESA cost estimates. METHODS: Data were extracted from electronic medical records in two US databases between January 2004 and December 2006. The study sample included 6743 patients receiving chemotherapy, with one or more visits during the study period, who received an ESA during a chemotherapy episode. Episodes of chemotherapy care were constructed using a 90-day gap in administration to identify the start and end. Patients receiving both DA and EA during their initial chemotherapy episode or with missing data were excluded, representing 42% of patients with CIA receiving an ESA. Drug costs were calculated from the cumulative dose multiplied by 106% of the average sales price (ASP) for DA or EA. Two propensity-score matches were conducted: first using variables available in administrative billing claims systems, then adding the baseline Hb test result. Regression-adjusted cost differences were estimated with and without baseline Hb, using generalized linear models. RESULTS: Using baseline Hb levels resulted in a better match of the baseline characteristics for the EA and DA treatment groups than the original sample or the matched sample without Hb variables. Mean ESA costs (year 2007 values) for the original sample were $US4171 for EA and $US3811 for DA (mean difference $US360; p < 0.001, standard error [SE] $US99). With propensity-score matching without Hb variables, mean estimated costs were $US3836 for EA and $US3599 for DA (mean difference $US237; p = 0.053, SE $US123). With propensity-score match including Hb variables, mean costs were $US3965 for EA and $US3536 for DA (mean difference $US429; p = 0.001, SE $US125). Cost differences in sensitivity analyses ranged between $US102 (p = 0.201) and $US261 (p = 0.003). CONCLUSIONS: Addition of baseline Hb level as a variable in propensity score and ESA cost models affects ESA treatment cost estimates in patients with cancer receiving chemotherapy. Cost comparisons based on observational data should use analytical methods that account for differences in clinical variables between treatment groups.

An experimental study on the noise properties of x-ray CT sinogram data in Radon space.

Computed tomography (CT) has been well established as a diagnostic tool through hardware optimization and sophisticated data calibration. For screening purposes, the associated x-ray exposure risk must be minimized. An effective way to minimize the risk is to deliver fewer x-rays to the subject or lower the mAs parameter in data acquisition. This will increase the data noise. This work aims to study the noise property of the calibrated or preprocessed sinogram data in Radon space as the mAs level decreases. An anthropomorphic torso phantom was scanned repeatedly by a commercial CT imager at five different mAs levels from 100 down to 17 (the lowest value provided by the scanner). The preprocessed sinogram datasets were extracted from the CT scanner to a laboratory computer for noise analysis. The repeated measurements at each mAs level were used to test the normality of the repeatedly measured samples for each data channel using the Shapiro-Wilk statistical test merit. We further studied the probability distribution of the repeated measures. Most importantly, we validated a theoretical relationship between the sample mean and variance at each channel. It is our intention that the statistical test and particularly the relationship between the first and second statistical moments will improve low-dose CT image reconstruction for screening applications.

An improved electronic colon cleansing method for detection of colonic polyps by virtual colonoscopy.

Electronic colon cleansing (ECC) aims to segment the colon lumen from a patient abdominal image acquired using an oral contrast agent for colonic material tagging, so that a virtual colon model can be constructed. Virtual colonoscopy (VC) provides fly-through navigation within the colon model, looking for polyps on the inner surface in a manner analogous to that of fiber optic colonoscopy. We have built an ECC pipeline for a commercial VC navigation system. In this paper, we present an improved ECC method. It is based on a partial-volume (PV) image-segmentation framework, which is derived using the well-established statistical expectation-maximization algorithm. The presented ECC method was evaluated by both visual inspection and computer-aided detection of polyps (CADpolyp) within the cleansed colon lumens obtained using 20 patient datasets. Compared to our previous ECC pipeline, which does not sufficiently consider the PV effect, the method presented in this paper demonstrates improved polyp detection by both visual judgment and CADpolyp measure.

Real-World Axitinib Use in the United States: A Retrospective Study Using Linked Datasets.

BACKGROUND: Axitinib is approved by the FDA for the treatment of advanced renal cell carcinoma (RCC) after failure of 1 previous systemic therapy and is distributed primarily through specialty pharmacies. Although the efficacy and safety of axitinib have been established in clinical trials, information from real-world populations will help to elucidate patients' clinical profiles and utilization patterns. Prescription records alone provide limited information on patient characteristics and other treatment experiences. Expansion of these data with information from medical claims databases should yield observational real-world data that may help to optimize therapy for patients with advanced RCC. OBJECTIVE: To link information from a specialty pharmacy database with information from medical and pharmacy claims databases to characterize real-world treatment patterns of axitinib as subsequent systemic therapy in patients with RCC in the United States. METHODS: This retrospective, observational, cohort study linked de-identified patient-level data from 22 specialty pharmacies that dispense axitinib with databases of longitudinal medical and pharmacy claims. Eligible patients had a diagnosis of RCC (> 1 claim for RCC defined as ICD-9-CM code 189.0), previously received > 1 systemic therapy, had the first prescription for axitinib dispensed between May 2012 and April 2013 (index), and had consistent claims reporting by pharmacies and physicians. All treatment data were used to calculate cycle, line of therapy, and duration of therapy; prescription data were used to determine axitinib dose modifications. Multivariate and logistic regression analyses were conducted to assess the effect of patient/prescriber characteristics on duration of axitinib therapy and dose modifications, respectively. RESULTS: In all, 1,175 patients met the study inclusion criteria and had data present in specialty pharmacy and claims databases. Most patients (74%) were male, and 68% were aged 55-74 years. Mean (SD) Charlson Comorbidity Index score was 2.7 (± 1.1); the most common comorbidity was hypertension (in 199 patients, 17%). Based on Rx-Risk-V, the most frequent concomitant conditions were pain (40%) and ischemic heart disease/hypertension (30%); the most frequent concomitant medications were antihypertensive medications (46%) and opiates (40%). Most prescribers (63%) were affiliated with an academic center, and all U.S. geographic regions were represented. In all, 847 patients (72%) had commercial insurance. Axitinib was prescribed as second-line therapy in 659 patients (56%), as third-line therapy in 326 patients (28%), and as fourth-line or later therapy in 190 patients (16%). In the overall population, mean (SD) duration of axitinib therapy was 168.6 (± 148.4) days. Axitinib treatment duration was 21 days longer in males than females (P = 0.013); 28 days longer in patients in the Northeast than in the Midwest or West (P = 0.010 and P = 0.016, respectively); and 26 days longer in patients receiving baseline hypothyroidism treatment (P = 0.004). In patients receiving second-line axitinib, the most common first-line therapy was sunitinib (56%), followed by pazopanib (16%) and everolimus (12%). Mean (SD) duration of second-line axitinib treatment was 172.3 (± 150.6) days and ranged from 127 days in patients who previously received temsirolimus to 196 days in those who previously received sorafenib. Of 1,025 patients who initiated axitinib at the standard 5 mg twice daily starting dose, 70% remained at this dose throughout treatment, whereas 10% had a dose increase. Younger age and gender (male) were associated with dose increases (OR = 0.958, 95% CI = 0.941-0.975 and OR = 0.573, 95% CI = 0.364-0.903, respectively). Baseline hypothyroidism treatment was associated with dose decreases and increases (OR = 1.662, 95% CI = 1.088-2.539 and OR = 2.149, 95% CI = 1.353-3.413, respectively). CONCLUSIONS: This analysis demonstrates the feasibility and utility of linking specialty pharmacy data to other longitudinal databases to better understand patient, provider, and reimbursement characteristics. These data provide insight into routine clinical use of axitinib as subsequent RCC therapy in the United States in the period following FDA approval, as well as additional information on sequencing of targeted agents in patients with advanced RCC. DISCLOSURES: This study was sponsored by Pfizer. MacLean and Cisar are employees of and hold stock in Pfizer. At the time of this analysis, Mehle, Eremina, and Quigley were employees of IMS Health who were paid consultants to Pfizer during the conduct of this study and in connection with the development of this manuscript. MacLean and Cisar contributed to study design and manuscript development. Mehle, Eremina, and Quigley contributed to study design, analysis, and manuscript development.

An improved electronic colon cleansing method for detection of polyps by virtual colonoscopy.

Electronic colon cleansing (ECC) aims to segment the colon lumen from the patient abdominal image acquired with colonic material tagging by oral contrast and other means, so that a virtual colon model can be constructed. Virtual colonoscopy (VC) navigates through the colon model looking for polyps in a similar manner as the fiber optic colonoscopy does. We had built an ECC pipeline for the commercial VC system of Viatronix Inc. In this paper, we present an improved ECC method. It is based on a partial -volume image -segmentation framework, which is derived using the well-established statistical expectation-maximization algorithm. The presented ECC method was evaluated by both visual inspection on the cleansed colon lumens and computer-aided detection of polyps (CADpolyp) using 20 patient datasets. Compared to our previous ECC pipeline, this presented new method demonstrates improvement in both visual judgment and CADpolyp.