Statistical analysis in pharmacoeconomic studies. A review of current issues and standards

The increasing number of economic evaluations of healthcare interventions, and of drug therapies in particular, has been well documented. However, surveys have demonstrated that standards of conduct of such studies have not similarly increased. Of particular concern is the lack of development or even consideration of statistical techniques in the reporting of studies. This article addresses issues that must be considered both in the conduct and in the assessment of the quality of studies. Throughout the paper, examples of pharmacoeconomic analyses are used to illustrate the points made. Recommendations for the conduct of future pharmacoeconomic studies are given. Such recommendations specifically relate to the level of testing that is conducted, the choice of statistical tests and the manner in which statistical significance is reported. In addition, existing methods for the statistical analysis of cost-effectiveness ratios and for the determination of sample size in economic evaluations are discussed, and a partial solution to this issue is offered.     

The vasa vasorum and angioplasty

Interruption of flow in the vasa vasorum may lead to medial necrosis and aneurysm formation. The purpose of this study was to determine whether angioplasty produces significant alterations in the morphology or blood flow of the vasa vasorum of the dilated artery. The morphology of the canine vasa vasorum was studied before and after angioplasty; in a separate experiment vessel wall blood flow (VWBF) in canine carotid arteries was measured after angioplasty to determine whether physiologic regulation of the blood flow was disrupted by arterial dilation. No morphologic changes could be demonstrated in the vasa vasorum of the dilated artery; however, VWBF was increased by 1194 +/- 215% (mean +/- standard error, p less than 0.01) between 90 and 120 minutes after angioplasty. VWBF in the adjacent nondilated arterial segment was also increased (720 +/- 177% between 10-30 minutes, p less than 0.01) but returned toward normal after 60 minutes. Adenosine caused a "paradoxical" decrease in VWBF (p less than 0.05) of the dilated arterial segment while causing increased VWBF (p less than 0.05) in the thoracic aorta. Angioplasty appears to produce persistent hyperemia in the dilated arterial wall. A paradoxical response to adenosine suggests that vasa vasorum in the dilated arterial segment are maximally vasodilated. This may be due to mechanical disruption of vasomotor tone or to release of vasoactive substances.     

The effects and after-effects of interferon alpha on human performance, mood and physiological functions

The effects of three doses of interferon alpha (IFNα, 1.5 Mu, 0.5 Mu and 0.1 Mu) on performance, mood and physiological function were examined in a double-blind placebo controlled trial. The subjects given an injection of 1.5 Mu showed symptoms which closely resembled those seen in influenza, although most of the symptoms had gone by the next day. All of the doses of IFN reduced subjective ratings of alertness, but again this effect was only apparent on the day of challenge. Subjects given the 1.5 Mu injection were slower on two of the performance tasks (a simple reaction time task and a pegboard task) both on the day of challenge and the following day. None of the other tasks in the battery was impaired by any of the doses of IFN, either on the day of challenge or the following day. These results suggest that IFNα has selective effects on performance, and it is possible that IFN-induced changes in CNS function provide a mechanism through which viral infections influence behaviour.     

Recruitment of patients into an internet-based clinical trials database: the experience of OncoLink and the National Colorectal Cancer Research Alliance.

PURPOSE: In March 2001, the National Colorectal Cancer Research Alliance (NCCRA) and OncoLink (http://www.oncolink.org) established a database to facilitate patient enrollment onto clinical trials. This study describes the population registering with the database and identifies discrepancies between individuals registering through the Internet and those registering through a telephone call center. METHODS: Participants registered with the NCCRA/OncoLink database through the Internet or a telephone call center. All participants entering the database completed a questionnaire regarding basic demographics, colon cancer risk factors, and indicated how they became aware of the database. Comparisons were made between individuals registering through the Internet and those registering through the telephone call center. RESULTS: A total of 2,162 participants registered during the first 16 months of the database. Most patients registered through the Internet rather than the telephone call center (88% v 12%; P < .001). More females than males registered (73% v 27%; P < .001). The majority (89%) were white. Participants registering through the Internet were younger than those registering through the call center (mean, 48.8 v 55.0 years; P < .001). There was no difference between the two groups with regard to sex or ethnicity. CONCLUSION: The Internet has the potential to increase the likelihood that interested individuals find appropriate clinical trials. Some of the discrepancies that are known to exist for access to the Internet were also seen for those registering with the database through the Internet. Despite these differences, the potential to increase clinical trial enrollment with this type of Internet-based database is high.     

Parvovirus B19 in kidney transplant patients

Renal transplant patients were screened for the presence of parvovirus B19, before transplantation and monthly for 4 months after transplantation, by means of a sensitive nested PCR assay. Upon screening plasma from 110 patients, we found that two asymptomatic patients were B19 DNA positive. One of these patients was PCR positive in the plasma sample taken 2 months after transplantation; the plasma contained anti-B19 IgG antibodies before transplant and throughout the follow-up period, with an increase in the IgG level in the second posttransplant sample coinciding with the detection of B19 DNA. IgM antibodies to B19 were not detected in this patient. Because, for this patient, the donor's spleen DNA was also B19 DNA positive, we suspect B19 transmission from the donor and limited B19 replication, inasmuch as this patient already had a primed immune response to B19. The other patient was PCR positive in the pretransplant and in the plasma sample taken 1 month after transplant and contained a strong anti-B19 IgG response in the pretransplant sample and throughout the follow-up period-and anti-B19 IgM antibodies were not detected before or after transplantation. By testing samples taken from this patient at 2 weeks, 2 months, and 3 months before transplantation, we were able to determine that the infection occurred shortly before transplantation. Unexpectedly, this graft failed and was removed 2 days after transplantation despite a negative cross-match. A pathological examination of the kidney indicated acute vascular rejection, suggesting a possible role for B19 in this complication.     

Aberrant crypt focus promotion and glucose intolerance: correlation in the rat across diets differing in fat, n-3 fatty acids and energy

McKeown-Eyssen (Cancer Epidemiol. Biomarkers Prevent., 3, 687-695, 1994) and Giovannucci (Cancer Causes Control, 6, 164-179, 1995), noting the striking similarity in lifestyle risk factors for colorectal cancer and insulin resistance, proposed that the hyperinsulinemia, glycemia and hypertriglyceridemia associated with insulin resistance promotes colon cancer. To compare the effect of diet on colon cancer promotion and insulin resistance in the F344 rat, we assessed the effect of fat, n-3 fatty acids and energy in pairwise comparisons on average size of aberrant crypt foci (ACF) and on glucose intolerance in the same animals in a single experiment. Diets high in fat and energy increased and diets with increased n-3 fatty acids and calorie restriction decreased both ACF growth and glucose intolerance compared with control diets. The measures of promotion of colon cancer and insulin resistance were strongly correlated (n = 98, r = 0.67, P < 0.001). In addition, both were highly correlated with daily energy intake (r = 0.62 and 0.66) and were also correlated with basal (post-prandial) insulin, glucose and triglycerides (r = 0.31-0.53, P < 0.01). We concluded that ACF growth and glucose intolerance are correlated for a wide range of diets and that increased circulating energy (glucose and triglycerides) may lead to both colon cancer promotion and insulin resistance.      

Effects of differential privacy techniques: Considerations for end users

We study the effects of differentially private (DP) noise injection techniques in a survey data setting, using the release of cost of early care and education estimates from the National Survey of Early Care and Education as a motivating example. As an example of how DP noise injection affects statistical estimates, our analysis compares the relative performance of DP techniques in the context of releasing estimates of means, medians, and regression coefficients. The results show that for many statistics, basic DP techniques show good performance provided that the privacy budget does not need to be split over too many estimates. Throughout, we show that small decisions, such as the number of bins in a histogram or the scaling of a variable in a regression equation, can have sometimes dramatic effects on the end results. Because of this, it is important to develop DP techniques with an eye towards the most important aspects of the data for end users.