Nonparametric analysis of statistic images from functional mapping experiments.

The analysis of functional mapping experiments in positron emission tomography involves the formation of images displaying the values of a suitable statistic, summarising the evidence in the data for a particular effect at each voxel. These statistic images must then be scrutinised to locate regions showing statistically significant effects. The methods most commonly used are parametric, assuming a particular form of probability distribution for the voxel values in the statistic image. Scientific hypotheses, formulated in terms of parameters describing these distributions, are then tested on the basis of the assumptions. Images of statistics are usually considered as lattice representations of continuous random fields. These are more amenable to statistical analysis. There are various shortcomings associated with these methods of analysis. The many assumptions and approximations involved may not be true. The low numbers of subjects and scans, in typical experiments, lead to noisy statistic images with low degrees of freedom, which are not well approximated by continuous random fields. Thus, the methods are only approximately valid at best and are most suspect in single-subject studies. In contrast to the existing methods, we present a nonparametric approach to significance testing for statistic images from activation studies. Formal assumptions are replaced by a computationally expensive approach. In a simple rest-activation study, if there is really no activation effect, the labelling of the scans as "active" or "rest" is artificial, and a statistic image formed with some other labelling is as likely as the observed one. Thus, considering all possible relabellings, a p value can be computed for any suitable statistic describing the statistic image. Consideration of the maximal statistic leads to a simple nonparametric single-threshold test. This randomisation test relies only on minimal assumptions about the design of the experiment, is (almost) exact, with Type I error (almost) exactly that specified, and hence is always valid. The absence of distributional assumptions permits the consideration of a wide range of test statistics, for instance, "pseudo" t statistic images formed with smoothed variance images. The approach presented extends easily to other paradigms, permitting nonparametric analysis of most functional mapping experiments. When the assumptions of the parametric methods are true, these new nonparametric methods, at worst, provide for their validation. When the assumptions of the parametric methods are dubious, the nonparametric methods provide the only analysis that can be guaranteed valid and exact.

Recall bias in a prospective cohort study of acute time-varying exposures: example from the herpetic eye disease study.

Recall bias is possible in a prospective cohort study when exposure status is transient and must be periodically recalled, and ascertainment occurs after symptom onset. We know of no published demonstration of such bias at play in a prospective cohort study. In a substudy of a randomized clinical trial, 308 participants were prospectively followed to investigate potential acute triggers of ocular herpes simplex virus (HSV) recurrences. Participants reported on the presence of systemic infection or high psychological stress (exposures) on a home log that was completed weekly for up to 15 months and mailed to the study's coordinating centers. By protocol, exposure reporting was to occur on the last day of the week (Sunday) so that a prospective 1-week lag period between exposure and outcome in the following week could be assessed. The study outcome was development of a recurrence of ocular HSV disease documented by clinical examination. Using 35 weekly reports of exposure properly completed before the week of an outcome, there was no evidence of higher risk of HSV recurrence associated with systemic infection (rate ratio = 0.62, 95% confidence interval [CI]: 0.19-2.02) or high psychological stress rate (ratio = 0.0, 95% CI: 0.0-undefined). In contrast, when the analysis was based on 26 weekly reports of exposure improperly completed on or after the date of outcome, the risk of recurrence associated with systemic infection was estimated to be 4-fold (rate ratio = 4.07, 95% CI: 1.84-8.98), and there was a suggestion of a 2-fold risk associated with high psychological stress (rate ratio = 2.02, 95% CI: 0.69-5.91). Without real-time monitoring of exposure reporting, preservation of the temporal exposure-disease relationship-an implicit assumption of the prospective cohort study design-may be particularly tenuous when transient exposures are investigated longitudinally.

Uveitis associated with topical beta-blockers.

OBJECTIVE: To assess the incidence of uveitis associated with topical beta-blockers. DESIGN: Retrospective cohort study. SETTING: A computerized database of a health maintenance organization was used to identify patients who were prescribed metipranolol or other beta-blockers. PATIENTS: Two groups: 1928 patients using 0.3% metipranolol and 3903 patients using other beta-blockers. MAIN OUTCOME MEASURE: Cases of uveitis were identified by noting a prescription of a topical corticosteroid and then conducting a chart review. RESULTS: No cases of uveitis were identified among the 5831 patients. The upper limits of the 95% confidence intervals for the incidence of uveitis in cohorts of this size are 1.87 cases per 1000 person-years of exposure for metipranolol and 1.38 for other beta-blockers. CONCLUSION: Drug-induced uveitis from metipranolol or other beta-blocker administration is a rare event, even if a causal relationship exists.

The amblyopia treatment study visual acuity testing protocol.

OBJECTIVE: To evaluate the reliability of a new visual acuity testing protocol for children using isolated surrounded HOTV optotypes. METHODS: After initial pilot testing and modification, the protocol was evaluated using the Baylor-Video Acuity Tester (BVAT) to present isolated surrounded HOTV optotypes. At 6 sites, the protocol was evaluated for testability in 178 children aged 2 to 7 years and for reliability in a subset of 88 children. Twenty-eight percent of the 178 children were classified as having amblyopia. RESULTS: Using the modified protocol, testability ranged from 24% in 2-year-olds to 96% in 5- to 7-year-olds. Test-retest reliability was high (r = 0.82), with 93% of retest scores within 0.1 logMAR unit of the initial test score. The 95% confidence interval for an acuity score was calculated to be the score +/-0.125 logMAR unit. For a change between 2 acuity scores, the 95% confidence interval was the difference +/-0.18 logMAR unit. CONCLUSIONS: The visual acuity protocol had a high level of testability in 3- to 7-year-olds and excellent test-retest reliability. The protocol has been incorporated into the multicenter Amblyopia Treatment Study and has wide potential application for standardizing visual acuity testing in children.

Infusion of total parenteral nutrition via the umbilical artery.

The administration of total parenteral nutrition through umbilical artery catheters in 48 neonates (birth weight 1.7 +/- 0.58 kg) was compared with administration via central venous catheters in 26 infants (birth weight 2.05 +/- 0.89 kg). There was no significant difference in the amount of calories delivered (72 +/- 12 vs 78 +/- 18 cal/kg/day) or in the mean daily weight gain (16.6 +/- 13.3 vs 18 +/- 13.9 g/day). The incidence of sepsis was significantly lower in the umbilical artery catheter group (10.4% vs 15.4%) but there was no significant difference in the rate of infection when adjustment was made for number of days of catheter life (1 per 224 days of catheter life in the umbilical artery group vs 1 per 199 days in the central venous catheter group). Other major complications included transient hypertension in 2 (4%) of the 48 umbilical artery catheter infants and in 1 (3.8%) of the central venous catheter group, aortic thrombosis in 1 (2%) of the 48 umbilical artery catheter infants and a tricuspid vegetation in 1 (3.8%) of the central venous catheter group. Results suggest that the umbilical artery is a reasonable route for the infusion of total parenteral nutrition in low birth weight infants who require arterial access for blood gas analysis. Use of the umbilical artery catheter for parenteral alimentation may avoid the need for surgical placement of central venous lines and the risk of the attendant complications. Nevertheless, safer routes and improved methods of infusion of parenteral infusion must continue to be developed.

The small abdominal aortic aneurysm: the eternal dilemma.

In order to evaluate morbidity and mortality after elective resection of abdominal aortic aneurysms (AAA) as it relates to aneurysm size, a retrospective review of 111 elective aneurysmectomies over a 5 year period was undertaken in a VA population. Thirty seven AAA's measured < 5 cm in diameter and 74 were > or = 5 cm by CT scan. Patients with small AAA (S-AAA) were significantly younger (mean 64 years) than those with large AAA (L-AAA) (mean 69 years) (p < 0.003). Both groups were similar with respect to prevalence of cardiovascular, pulmonary and renal disease. Aortic cross-clamping time was significantly shorter in L-AAA, possibly because those with S-AAA had a higher prevalence of associated occlusive disease requiring more femoral anastomoses (p < 0.04). Postoperatively six patients (8%) had a myocardial infarction (MI) in the L-AAA group and four (5%) of these died. In contrast no patient with S-AAA suffered a postoperative MI. The rates of non-cardiac complications and length of hospital stay were not significantly different between the two groups. However, the patients with L-AAA stayed longer in ICU (p < 0.05) and the overall combined morbidity rate was significantly higher in this group (p < 0.02). Our results suggest that resection of S-AAA upon diagnosis in acceptable risk patients appears to be the safest overall therapeutic plan.

Comparison of the cytology brush with the Dacron swab for detecting Chlamydia trachomatis by enzyme immunoassay in female university students.

Previous studies comparing the cytology brush (Cytobrush) with swabs for the detection of Chlamydia trachomatis from the uterine cervix have yielded divided results. The authors carried out a retrospective analysis at the University of South Florida Student Health Service to evaluate the effect of a change in specimen collection technique on C trachomatis prevalence estimates in female university students. Samples were collected from 1,003 women in 1989, using a small Dacron swab, and 1,229 women in 1990, using a brush; the samples were tested by means of the Pathfinder Enzyme Immunoassay (EIA) detection kit. Positive test results were 5.88% for 1989 and 11.66% for 1990. Actual C trachomatis prevalence for 1989 and 1990 was stable, as determined by three control groups. The results suggest that the cytology brush is more effective than the Dacron swab for the detection of C trachomatis in the uterine cervix when it is used with the Pathfinder EIA test kit.

Cardiovascular and sympathetic reactions to in-flight emergency responses among base fire fighters.

The responses of 12 healthy male fire fighters to simulated and actual in-flight emergencies were investigated. Subjects were aroused from sleep during all emergency responses. The ECG, blood pressure, heart rate (HR), rate-pressure product (RPP), and plasma norepinephrine (NE) responses were determined. Despite a mean HR increase of 112% (75 beats/min) and a 145% rise in RPP, there was no significant elevation in plasma NE concentration during the emergency response. Legible ECG tracings showed no abnormal ST segment deviations or arrhythmias. The HR, BP, and RPP results indicated a greater cardiovascular response to emergencies among inactive fire fighters than among those who were physically active. Based on the observed differences between emergency and simulated emergency responses, we concluded that the physiological reactions during emergency responses were due primarily to the arousal response. When suddenly aroused from sleep, the fire fighter's response to in-flight emergencies produces significant elevations in HR and myocardial oxygen consumption which were unrelated to increases in sympathetic activity.

Increasing consumer satisfaction. One social service and public health initiative shows how social marketing can increase consumer satisfaction.

The key to a successful social marketing approach to health care is continually listening to consumers' feedback and being willing to change the health product or service according to their needs and preferences. This approach can increase the likelihood of consumers being satisfied with, and continuing to utilize or provide the particular health service.

Comparison of two tests useful in situations where treatment is expected to increase variability relative to controls.

The type I error and power characteristics of the modified t test were compared with those of the generalized t test. Results suggested that, in contrast to the generalized t test, the modified t test can be seriously non-robust to departures from population normality, with such departures often producing anti-conservative results. Neither test held an absolute power advantage over the other when responses were from normal distributions, but the modified t test was generally more powerful for these conditions. In comparison with the pooled samples t test, both tests were usually much more efficient when treatment caused increases both in mean response and between-subject variance, and suffered only small disadvantages when between-subject variance was unchanged by treatment. Given these results and other considerations, recommendations for use of these recently devised tests are given.

A note on the operating characteristics of the modified F test.

Brownie, Boos, and Hughes-Oliver (1990, Biometrics 46, 259-266) suggested a modification to the fixed-effects analysis of variance (ANOVA) F test for use in situations where treatments are likely to affect mean response while simultaneously increasing between-subject variability. These authors suggest that the modified test (1) is robust against departures from population normality, (2) is more powerful than its classical counterpart when the expected alternative form is encountered, and (3) will suffer only minor power losses relative to the usual test when alternatives are in the form of simple shifts in location. This study shows that the modified test is not generally robust when sampling is from nonnormal distributions and is likely to produce unacceptably high Type I error inflations when sampling is from heavy-tailed or skewed distributions. Some power results are also provided for the multisample situation to supplement those of Brownie et al., who generated results only for the two-sample case.

On the use of the generalized t and generalized rank-sum statistics in medical research.

We have used Monte Carlo methods to compare the type I error properties of the conditional and unconditional versions of the generalized t and the generalized rank-sum tests to those of the independent samples t and Wilcoxon rank-sum tests. Results showed inflated type I errors for the conditional generalized tests but not for the unconditional tests. We also compared the power of the unconditional generalized tests to that of the t and Wilcoxon tests under a variety of conditions. Results showed the generalized tests to be much more efficient than their traditional counterparts in some circumstances, but substantially less powerful in others. Based on these and other considerations, we conclude that the application of these newer statistics in medical research needs further consideration.

Topographical and temporal stability of the P300.

Sixteen normal adults listened to a standard oddball auditory paradigm 3 times, each time separated by 15 min, and this protocol was then repeated 1-2 months later. Stability of the wave forms was measured between 250 and 500 msec for each subject and for each of 16 electrodes by the cross-correlation technique described by Glaser and Ruchkin (1976) and Gevins (1987) in which points on 2 digitized wave forms are paired to produce a correlation coefficient. For each subject, the correlation coefficient was generated for each electrode and then averaged across the 2 parietal and 2 central electrodes to produce a single stability measure for the central electrodes, while the cross-correlation coefficients for the 12 remaining electrodes were averaged for a peripheral stability measure. Three-way repeated measures ANOVAs were performed to determine the significance of cross-correlation coefficient differences. The stability over 15 min for the central electrodes was 0.80, indicating that the P300 was very stable over a short time. The peripheral electrodes were significantly less stable than the central electrodes (P = 0.001). The stability of the wave forms was virtually unchanged when assessed over 1 month (P = 0.9). The target wave forms were significantly more stable than the difference waves (target minus non-target) for both the central and peripheral electrodes (P = 0.04 and 0.01). When the 3 blocks within each session were averaged, there was a significant increase in stability (P less than 0.0005). The wave form cross-correlation coefficients can be used as a measure of the stability of a topographical map over time. The map of the target P300 wave form is very stable, showing no loss in stability from 15 min to 1 month, is more stable than the difference wave form and significantly increases in stability when separate trial blocks are averaged together.

An alternative method for significance testing of waveform difference potentials.

Guthrie and Buchwald (1991) proposed an ad hoc procedure for assessing the statistical significance of waveform difference potentials that may arise in a variety of psychophysiology research contexts. In our paper, an alternative method is presented and demonstrated that has fewer underlying assumptions than does the Guthrie-Buchwald test and may, therefore, produce better results in some situations. In particular, the test proposed here (a) is distribution free, (b) requires no assumption of an underlying correlation structure (e.g., first-order autoregressive), (c) requires no estimate of the population autocorrelation coefficient, (d) is exact, (e) produces p values for any number of subjects and time points, and (f) is highly intuitive as well as theoretically justifiable. This procedure may be used to carry out multiple comparisons with exact specification of experiment-wise error, however, this test is based on permutation principles and may require large amounts of computer time for its implementation.

An exact statistical method for comparing topographic maps, with any number of subjects and electrodes.

Statistical methods for testing differences between neural images (e.g., PET, MRI or EEG maps) are problematic because they require (1) an untenable assumption of data sphericity and (2) a high subject to electrode ratio. We propose and demonstrate an exact and distribution-free method of significance testing which avoids the sphericity assumption and may be computed for any combination of electrode and subject numbers. While this procedure is rigorously rooted in permutation test theory, it is intuitively comprehensible. The sensitivity of the permutation test to graded changes in dipole location for systematically varying levels of signal/noise ratio, intersubject variability and number of subjects was demonstrated through a simulation of 70 different conditions, generating 5,000 different data sets for each condition. Data sets were simulated from a homogeneous single-shell dipole model. For noise levels commonly encountered in evoked potential studies and for situations where the number of subjects was less than the number of electrodes, the permutation test was very sensitive to a change in dipole location of less than 0.75 cm. This method is especially sensitive to localized changes that would be "washed-out" by more traditional methods of analysis. It is superior to all previous methods of statistical analysis for comparing topographical maps, because the test is exact, there is no assumption of a multivariate normal distribution or of the correlation structure of the data requiring correction, the test can be tailored to the specific experimental hypotheses rather than allowing the statistical tests to limit the experimental design, and there is no limitation on the number of electrodes that can be simultaneously analyzed.

Control of familywise errors in multiple endpoint assessments via stepwise permutation tests.

We describe permutation based sequentially rejective multiple comparison procedures useful in multiple endpoint assessments. We used Monte Carlo methods to compare the power of these newly devised tests to that of tests due to Holm and Rom as well as to the classical Bonferroni method. We illustrate applications of the methods with analysis of visual field data collected from optic neuritis patients. We conclude that the new methods are particularly useful when there are many endpoints involved, the data are significantly correlated, and/or the distributional assumptions are questionable.

The late somatosensory evoked potential in premature and term infants. II. Topography and latency development.

The maturation of latency and scalp voltage topography of the simultaneously bilateral somatosensory evoked potential was studied in 53 neurologically intact pre-term and term infants, from 31 to 40 weeks post-menstrual age. Four peaks (N1, P1, N2 and P2) were reliably identified in all infants. The latency of each peak decreased as the infants matured. Each peak had a unique voltage scalp topography that remained stable as infants matured, even though the maps changed in amplitude intensity. N2 was large, easily identifiable with a central peak, and extremely stable in topography, suggesting that it might be used to evaluate the functional status of the somatosensory cortex in pre-term and term infants who are at high risk for developing intracranial hemorrhage leading to abnormalities of tone and delays in motor development.

A Study of Multivariate Permutation Tests Which May Replace Hotelling's T2 Test in Prescribed Circumstances.

Multivariate permutation tests are described and studied which may be profitably substituted for Hotelling's one-sample P test in situations commonly arising in behavioral science research. These tests (a) may be computed even when the number of variables exceeds the number of subjects, (b) are distribution-free, (c) may be tailored for sensitivity to specific treatment alternatives, and (d) provide one-sided as well as two-sided tests of hypotheses. Power comparisons were made between the permutation tests and Hotelling's T(2) test under a variety of treatment effect model, correlation structure and number of variables combinations. Results show that the permutation tests have significant power advantages over the T(2) in a variety of circumstances, but may have considerably less power in others.

Parametric and non-parametric tests for the overall comparison of several treatments to a control when treatment is expected to increase variability.

We consider the problem of making an overall comparison of several treatments to a control where experimental units are randomly assigned to either the 'control' group which receives no treatment or to one of k-1 'treatment' groups. We assume that the effect of the treatments is, if anything, a location shift possibly accompanied by an increase in scale relative to that of the control group. The ANOVA F test loses considerable power in such circumstances. A modification of the ANOVA F test has been proposed which uses the variance estimate from the controls in place of the usual pooled variance estimate. However, this modification has shortcomings when k exceeds two and the variances of the treatment groups are not inflated. We develop a combination procedure to avoid the pitfalls of the modified and usual F tests. We then propose parametric and non-parametric implementations of a likelihood ratio test that more efficiently incorporates the assumptions of this problem, yielding a test with a high power profile over a large range of normal alternatives. We use simulations to compare the power of the competing tests against several alternatives for normal and non-normal data.