Confidence intervals after multiple imputation: combining profile likelihood information from logistic regressions.

In the logistic regression analysis of a small-sized, case-control study on Alzheimer's disease, some of the risk factors exhibited missing values, motivating the use of multiple imputation. Usually, Rubin's rules (RR) for combining point estimates and variances would then be used to estimate (symmetric) confidence intervals (CIs), on the assumption that the regression coefficients were distributed normally. Yet, rarely is this assumption tested, with or without transformation. In analyses of small, sparse, or nearly separated data sets, such symmetric CI may not be reliable. Thus, RR alternatives have been considered, for example, Bayesian sampling methods, but not yet those that combine profile likelihoods, particularly penalized profile likelihoods, which can remove first order biases and guarantee convergence of parameter estimation. To fill the gap, we consider the combination of penalized likelihood profiles (CLIP) by expressing them as posterior cumulative distribution functions (CDFs) obtained via a chi-squared approximation to the penalized likelihood ratio statistic. CDFs from multiple imputations can then easily be averaged into a combined CDF c , allowing confidence limits for a parameter ß at level 1 - α to be identified as those ß* and ß** that satisfy CDF c (ß*) = α ∕ 2 and CDF c (ß**) = 1 - α ∕ 2. We demonstrate that the CLIP method outperforms RR in analyzing both simulated data and data from our motivating example. CLIP can also be useful as a confirmatory tool, should it show that the simpler RR are adequate for extended analysis. We also compare the performance of CLIP to Bayesian sampling methods using Markov chain Monte Carlo. CLIP is available in the R package logistf.

Confidence intervals for ratios of AUCs in the case of serial sampling: a comparison of seven methods.

Pharmacokinetic studies are commonly performed using the two-stage approach. The first stage involves estimation of pharmacokinetic parameters such as the area under the concentration versus time curve (AUC) for each analysis subject separately, and the second stage uses the individual parameter estimates for statistical inference. This two-stage approach is not applicable in sparse sampling situations where only one sample is available per analysis subject similar to that in non-clinical in vivo studies. In a serial sampling design, only one sample is taken from each analysis subject. A simulation study was carried out to assess coverage, power and type I error of seven methods to construct two-sided 90% confidence intervals for ratios of two AUCs assessed in a serial sampling design, which can be used to assess bioequivalence in this parameter.

Transmyocardial laser revascularization: epicardial ECG detection provides efficient R-wave triggering during mobilization of the heart.

OBJECTIVE: In order to achieve an accurate intraoperative ECG detection, a new technique in detecting the trigger-signal was developed. In contrast to the traditional three-lead ECG-configuration, the left leg electrode was connected to a transient epicardial pacemaker electrode on the left-ventricular surface. BACKGROUND DATA: The Holmium:YAG-Laser for Transmyocardial Laser Revascularization (TMLR) is R-wave-triggered, providing the release of energy only during the refractory period of the heart cycle. However, an exact ECG-triggering during mobilization of the apex and/or posterior wall is difficult to achieve by using conventional ECG-configuration, therefore increasing the risk for mistriggering and induction of arrhythmias during TMLR. MATERIALS AND METHODS: Two groups of patients, all undergoing stand alone TMLR-procedures via left minithoracotomy, were compared. Ten patients were operated with the conventional ECG configuration (group 1) and ten patients with the modified epicardial ECG configuration (group 2). RESULTS: In patients of group 1, as a result of a loss of the trigger signal or due to the triggering of artifacts, the incidence of correctly triggered QRS-complexes was 56% of all documented QRS-complexes. In contrast, an excellent triggering was observed in 98% (p < 0.001) in group 2, resulting in a reduction of laser operative time by 35% (p < 0.001) and a decrease in the incidence of intraoperative ventricular fibrillation (0 vs. 3). CONCLUSION: In conclusion, this new ECG configuration is a simple but effective method in achieving an excellent ECG signal during all stages of TMLR. As a consequence, a reduction in operative time and incidence of ventricular fibrillation can be achieved.

Resection of hilar cholangiocarcinomas: pivotal prognostic factors and impact of tumor sclerosis.

The well-known poor prognosis of proximal bile duct cancer is due to its unfortunate anatomical location and its late diagnosis. Successful tumor resection, which is considered to be optimal treatment, depends on many factors. Eighty-eight patients suffering from proximal bile duct cancer underwent surgical exploration at our institution between 1977 and 1998. In 37 patients the tumor was resectable; in the remaining 51 patients exploratory laparotomy or a palliative operation was performed. The median survival after tumor resection was 18.6 months, but median survival after a palliative procedure or an exploratory laparotomy was only 3.4 months (p < 0.001). A curative R0 resection was possible in 11 patients, an R1 resection was performed in 22 patients, and 4 patients had an R2 resection. The median survival rate after R0 resection was 83.6 months, 12.3 months after R1 resection, and 2.7 months after R2 resection (p < 0.001). Survival after resection in patients with negative lymph nodes (n = 30) was significantly longer than in those with positive lymph nodes (n = 7) (p = 0.022). Grade of tumor sclerosis tended to have an influence on resectability rate (p = 0.076). The pattern of tumor growth was without statistical influence. Multivariate analysis revealed resection (p < 0.001) as the only significant prognostic marker for patient survival. Radical resection is the only therapy that provides a chance for long-term survival, with sclerosis of the cancer tending to have an influence on univariate analysis.

Fixing the nonconvergence bug in logistic regression with SPLUS and SAS.

When analyzing clinical data with binary outcomes, the parameter estimates and consequently the odds ratio estimates of a logistic model sometimes do not converge to finite values. This phenomenon is due to special conditions in a data set and known as 'separation'. Statistical software packages for logistic regression using the maximum likelihood method cannot appropriately deal with this problem. A new procedure to solve the problem has been proposed by Heinze and Schemper (Stat. Med. 21 (2002) pp. 2409-3419). It has been shown that unlike the standard maximum likelihood method, this method always leads to finite parameter estimates. We developed a SAS macro and an SPLUS library to make this method available from within one of these widely used statistical software packages. Our programs are also capable of performing interval estimation based on profile penalized log likelihood (PPL) and of plotting the PPL function as was suggested by Heinze and Schemper (Stat. Med. 21 (2002) pp. 2409-3419).

Case-based distance measurements for the selection of controls in case-matched studies: application in coronary interventions.

In case-based studies, controls are retrospectively assigned to patients in order to permit a statistical evaluation of the study results through a comparison of the main outcome measures for the patient and retrieved control groups. Inappropriate selection of the controls by using false retrieval parameters or a false algorithm might lead to an incorrect data analysis and a false interpretation of the results. The main hypothesis of the presented study was that, if the matching baseline parameters were at least nearly perfectly selected, then the outcomes of the matched pairs should be similar, or no significant differences in study outcome should be observed between the patients and the matched controls. In the present study, the case-based domain was created from 1566 patients who had undergone intracoronary stent implantation. Uni- and multivariate logistic regression analysis determined nine significant predictors (matching variables) for the occurrence of major adverse cardiac events. An additional 425 consecutive patients undergoing intracoronary stent implantation were then matched with all the potential controls from the database by calculating the individual distance between the patient and the matched control on the basis of the elaborated retrieval algorithm. To test the matching hypothesis, different distance measurements, and an altered outcome and retrieval process were examined. With "flexible" distance measurements, the mean distance between the patients and the first matched controls was 1.31. The major adverse cardiac events were compared in the patient and matched control groups. The best sensitivity and specificity values of the matching system could be achieved in matched pairs with the distance < or =3.00 (95.1% of all patients). On the further stepwise exclusion of the matched pairs exhibiting a distance greater than 2 and 1, then the number of "matchable" controls and the specificity of our matching concept decreased considerably. When the short- or mid-term outcome was compared by using the long-term follow-up matching parameters, no correlation could be found between the treated subjects and controls, indicating that for the other study, main measures of other appropriate parameters must be selected, and demonstrating the importance of the outcome-oriented selection of the retrieval parameters. Furthermore, the outcome measures of the patients and randomly (non-systematically) selected controls did not correlate, revealing the impossibility of drawing correct study conclusions from a non-matched, randomly assigned pairs. In conclusion, the sensitivity and specificity of the matching program, and the study conclusions depend on the appropriately predefined matching parameters and retrieval algorithm.

Long-term endothelial changes in phakic eyes with posterior chamber intraocular lenses.

PURPOSE: To evaluate the long-term endothelial cell changes in phakic eyes after implantation of a posterior chamber phakic intraocular lens to correct high ametropia. SETTING: Department of Ophthalmology, University Hospital of Vienna, Vienna, Austria. METHODS: Thirty-four eyes of 21 patients having implantation of an implantable contact lens (ICL, Staar Surgical) for high myopia or hyperopia were prospectively examined. The minimum follow-up was 2 to 4 years. Preoperative and serial postoperative specular microscopy (Noncon Robo SP 8000, Konan) was performed to evaluate the long-term endothelial cell changes. RESULTS: The mean preoperative endothelial cell density was 2854 cells/mm(2). The mean endothelial cell loss from preoperatively was 1.8% at 3 months, 4.2% at 6 months, 5.5% at 12 months, 7.9% at 2 years (n = 34), 12.9% at 3 years (n = 13), and 12.3% at 4 years (n = 11). All other endothelial cell characteristics remained stable during the 4-year follow-up. CONCLUSIONS: Continuous endothelial cell loss was observed after ICL implantation during a 4-year follow-up. There was rapid cell loss until 1 year postoperatively, after which the rate of loss was no longer statistically significant. The percentage of hexagonal cells (polymorphism) and the coefficient of variation (polymegethism) remained stable during the 4-year follow-up.

TP53 genotype but not p53 immunohistochemical result predicts response to preoperative short-term radiotherapy in rectal cancer.

OBJECTIVE: To evaluate and compare the predictive power of p53 gene analysis versus p53 immunohistochemical staining in terms of response to preoperative short-term radiotherapy using 25 Gy in operable rectal cancer. SUMMARY BACKGROUND DATA: Recent studies show that p53 may be a determinant of radiosensitivity being required for induction of apoptosis in case of radiation-induced DNA damage. METHODS: Preirradiation biopsy samples of 64 patients with rectal carcinoma were analyzed. Genetic alterations of the p53 gene were detected by complete direct sequencing of exons 2 to 10. Expression of the nuclear phosphoprotein p53 was assessed by immunohistochemical staining. Results were correlated with histopathology of resected specimens and follow-up data, respectively. RESULTS: Mutations of the p53 gene were present in 45% of tumors. Patients with a normal p53 gene had a significant survival advantage. Comparing pre- and postradiotherapy T category, a reduction was seen in patients with normal p53 genotype only. A mutant p53 genotype was highly specific in indicating stable disease concerning T category after irradiation. Protein overexpression was detected in 61%. Overexpression of the p53 protein was not related to survival or response. The concordance between immunohistochemistry and sequencing was only 0.51. CONCLUSIONS: The authors show that downstaging after short-term radiation may occur but is seen in tumors with normal p53 gene only. Moreover, p53 genotype but not p53 immunohistochemistry is predictive for response to preoperative short-term radiotherapy and patient survival.

The value of switching from cyclosporine to tacrolimus in the treatment of refractory acute rejection and obliterative bronchiolitis after lung transplantation.

Standard cyclosporine-based immunosuppression is ineffective in the treatment of refractory acute rejection (RAR) and obliterative bronchiolitis (OB) that follows lung transplantation. The aim of this study was to evaluate the results of switching from cyclosporine to tacrolimus in the treatment of these situations. Nineteen patients entered the study. The indication for switching was OB in 11 patients and RAR in 8. Mean age was 41.3 +/- 13.1 years. In patients with RAR, the number of acute rejections was 1.5 +/- 0.7 and there were zero episodes per patient per 100 days before and after switching, respectively ( P = 0.02). There was no significant reduction of the decline of forced expiratory volume (FEV(1)) within 6 months after switching in patients with OB. We conclude that the conversion from cyclosporine to tacrolimus was associated with favourable results in the treatment of RAR. Further studies are required to assess the influence of this approach in the treatment of OB.

Vascular morphometric changes after radioactive stent implantation: a dose-response analysis.

OBJECTIVES: The goal of this study was to evaluate the dose-dependency of morphometric changes in the coronary arterial wall after radioactive stenting. BACKGROUND: Radioactive stents have been found to reduce intrastent intimal hyperplasia (IIH) but lead to a characteristic type of restenosis occurring predominantly at the stent edges. METHODS: Fifteen patients underwent intravascular ultrasound (IVUS) examination after implantation of a P-32 radioactive stent and at the six-month follow-up. The post-stent IVUS measurements on seven predefined locations of each lesion were subjected to a computer algorithm for the development of dose-volume histograms (DVH). Thus, we derived the radiation doses delivered to at least 10% and 90% of the adventitia (DV10, DV90). The IIH and vascular remodeling at follow-up were correlated with the doses in each segment. RESULTS: The IIH was most pronounced at the stent edges and lowest in the stent-body, whereas we detected a significant expansive remodeling within the stent body. The delivered doses correlated with a decreased IIH (r = 0.52, p < 0.001 for DV10 and r = 0.62, p < 0.001 for DV90) and with expansive remodeling (r = 0.48, p = 0.009 for DV10 and r = 0.50, p = 0.006 for DV90). A DV10 >90 Gy or a DV90 >15 Gy reduced IIH and induced expansive remodeling. Plaque growth was not reduced by radioactive stents. CONCLUSIONS: The DVH analysis reveals a dose-dependent increase of external elastic lamina area behind radioactive stents, whereas plaque growth is not reduced but inverted into an outward direction from the stent. A DV10 >90 Gy or a DV90 >15 Gy results in a beneficial long-term outcome after radioactive stenting.

SAS and SPLUS programs to perform Cox regression without convergence problems.

When analyzing survival data, the parameter estimates and consequently the relative risk estimates of a Cox model sometimes do not converge to finite values. This phenomenon is due to special conditions in a data set and is known as 'monotone likelihood'. Statistical software packages for Cox regression using the maximum likelihood method cannot appropriately deal with this problem. A new procedure to solve the problem has been proposed by G. Heinze, M. Schemper, A solution to the problem of monotone likelihood in Cox regression, Biometrics 57 (2001). It has been shown that unlike the standard maximum likelihood method, this method always leads to finite parameter estimates. We developed a SAS macro and an SPLUS library to make this method available from within one of these widely used statistical software packages. Our programs are also capable of performing interval estimation based on profile penalized log likelihood (PPL) and of plotting the PPL function as was suggested by G. Heinze, M. Schemper, A solution to the problem of monotone likelihood in Cox regression, Biometrics 57 (2001).

Basic fibroblast growth factor and hypertensive disorders in pregnancy.

OBJECTIVE: We examined serum concentrations of basic fibroblast growth factor (bFGF) in women with hypertensive disorders in pregnancy and analyzed whether serum concentrations of bFGF can be used as a discriminator between mild and severe preeclampsia. METHODS: One hundred and twenty pregnant women were included in this prospective cohort study. We evaluated serum concentrations of bFGF in pregnant women with chronic hypertension (n=22), mild preeclampsia (n=40), severe preeclampsia (n=31), and healthy pregnant women (n=27). RESULTS: Median serum concentrations of bFGF in healthy pregnant women, women with chronic hypertension, and women with mild or severe preeclampsia were 0.0 (0-37.2), 0.0 (0-3.0), 1.7 (0-97.2), and 0.0 (0-52.0), respectively. Comparison of the median values of serum bFGF concentrations showed a significant difference between healthy pregnant women and women with mild preeclampsia (P=0.02). In a logistic regression model, we found a significant influence of bFGF serum concentrations on the diagnosis of mild preeclampsia (P=0.01), but not on the diagnosis of chronic hypertension (P=0.19) or severe preeclampsia (P=0.41). CONCLUSIONS: Elevated serum concentrations of bFGF are associated with mild preeclampsia, but are not discriminatory for the distinction between mild and severe preeclampsia.