Outcome-oriented cutpoints in analysis of quantitative exposures

Nutritional index for immune-checkpoint inhibitor in patients with metastatic gastro-esophageal junction/gastric cancer

BACKGROUND

Nutritional status is strongly associated to prognosis in metastatic gastrooesophageal junction (mGOJ)/gastric cancer (GC) patients. The aim of the present study was to develop an immune-checkpoint inhibitor (ICI)-specific nutritional index (NI).

METHODS

Ten serum and anthropometric nutritional markers derived from blood tests or CT scans were analyzed at baseline in patients treated with second-line ICI and correlated with overall survival (OS). An ICI-specific NI (the NUTRIICI) was developed with its specificity assessed in an independent group of patients treated with standard second-line chemotherapy.

RESULTS

From June 2014 to December 2018, 57 mGOJ/GC patients (14 females, 43 males) with a median(m) age of 61 years (range 29-85) received ICI as second-line therapy (Pembrolizumab n=26, Nivolumab n=16, Avelumab n=15). Among the 10 analyzed variables, Onodera's prognostic NI (PNI) ≤33 and waist-to-hip (WHR) <1 were independent predictors of OS and used to build the NUTRIICI. Patients with both favorable factors (i.e., PNI >33 and WHR ≥1, comparator group) had a mOS of 18.0 vs. 6.7 months of patients with one unfavorable factor (either PNI ≤33 or WHR <1, Hazard Ratio, HR 3.06), vs. 1.3 months of patients with both unfavorable factors (HR 17.56), overall P<0.0001. In the independent group of patients treated with standard chemotherapy NUTRIICI was not associated with prognosis (P=0.57).

CONCLUSIONS

NUTRIICI is the first ICI-specific NI for mOGJ/GC patients receiving second-line ICI. A validation in larger cohorts is strongly encouraged.

Reappraisal of the T Category for Solitary Intrahepatic Cholangiocarcinoma by Tumor Size in 611 Early-Stage (T1-2N0M0) Patients After Hepatectomy: a Surveillance, Epidemiology, and End Results (SEER) Analysis

BACKGROUND

The association between tumor size and survival in patients with intrahepatic cholangiocarcinoma (ICC) after hepatectomy is controversial, and the T category in the American Joint Committee on Cancer (AJCC) stage for ICC is a topic of debate.

METHODS

Data from 611 T1-2N0M0 ICC patients classified by the AJCC 8th edition who underwent hepatectomy were extracted from the Surveillance, Epidemiology, and End Results (SEER) database during 1988-2015. Cancer-specific survival was evaluated using Kaplan-Meier analysis. The optimal cutoff value of solitary tumor size was used an adjusted p value approach to discriminating patient survival.

RESULTS

In the AJCC 8th staging system, using a 5-cm cut-off value of tumor size for solitary ICC without vascular invasion (S/VI-) was not associated with survival in T1 category (p = 0.201), and multifocal ICC with vascular invasion had a worse survival than solitary ICC with vascular invasion (S/VI+) in T2 category (p = 0.014). Tumor size was a prognostic factor for both S/VI- and S/VI+, the optimal cutoff value of tumor size was obtained 8 cm for S/VI- and 3 cm for S/VI+. S/VI- ≤ 8 cm had a similar survival to S/VI+ ≤ 3 cm (p = 0.126), S/VI- > 8 cm had a similar survival to S/VI+ > 3 cm (p = 0.655), and multifocal ICC had a similar survival with S/VI- > 8 cm (p = 0.159) and S/VI+ > 3 cm (p = 0.196). When the cohort was divided into two groups-new T1 (S/VI- ≤ 8 cm and S/VI+ ≤ 3 cm) and new T2 (S/VI- > 8 cm, S/VI+ > 3 cm and multifocal ICC)-significant survival difference was observed (p < 0.0001).

CONCLUSIONS

The discriminatory power of the AJCC 8th edition for solitary ICC could be further enhanced by subdividing tumors according to size and vascular invasion (8 cm for S/VI- and 3 cm for S/VI+).

Low Serum Levels of DKK2 Predict Incident Low-Impact Fracture in Older Women

There are currently no robust noninvasive markers of fragility fractures. Secreted frizzled related protein‐1 (sFRP‐1), dickkopf‐related protein 1 (DKK1) and DKK2, and sclerostin (SOST) inhibit Wnt signaling and interfere with osteoblast‐mediated bone formation. We evaluated associations of serum levels of sFRP‐1, DKK1, DKK2, and SOST with incident low‐impact fracture and BMD in 828 women aged ≥65 years from EpiDoC, a longitudinal population‐based cohort. A structured questionnaire during a baseline clinical appointment assessed prevalent fragility fractures and clinical risk factors (CRFs) for fracture. Blood was collected to measure serum levels of bone turnover markers and Wnt regulators. Lumbar spine and hip BMD were determined by DXA scanning. Follow‐up assessment was performed through a phone interview; incident fragility fracture was defined by any new self‐reported low‐impact fracture. Multivariate Cox proportional hazard models were used to analyze fracture risk adjusted for CRFs and BMD. During a mean follow‐up of 2.3 ± 1.0 years, 62 low‐impact fractures were sustained in 58 women. A low serum DKK2 level (per 1 SD decrease) was associated with a 1.5‐fold increase in fracture risk independently of BMD and CRFs. Women in the two lowest DKK2 quartiles had a fracture incidence rate of 32 per 1000 person‐years, whereas women in the two highest quartiles had 14 fragility fractures per 1000 person‐years. A high serum sFRP1 level was associated with a 1.6‐fold increase in fracture risk adjusted for CRFs, but not independently of BMD. Serum levels of SOST (r = 0.191; p = 0.0025) and DKK1(r = −0.1725; p = 0.011) were correlated with hip BMD, but not with incident fragility fracture. These results indicate that serum DKK2 and sFRP1 may predict low‐impact fracture. The low number of incident fractures recorded is a limitation and serum levels of Wnt regulators should be further studied in other populations as potential noninvasive markers of fragility fractures. © 2019 The Authors. JBMR Plus published by Wiley Periodicals, Inc. on behalf of American Society for Bone and Mineral Research.

Added predictive value of high uric acid for cardiovascular events in the Ambulatory Blood Pressure International Study

The prognostic value of uric acid (UA) for cardiovascular events (CVE) is still debated. Our purpose was to investigate the association between UA and CVE in 5243 participants of the ABP-International study with the main aim of identifying optimal sex-specific cut-points. In multivariable Cox analyses, the relationship between CVE and UA as a continuous variable was modeled by including both linear and nonlinear terms. Survival models were also estimated with UA as a categorical variable. Optimal UA cut-points were determined using an outcome-oriented approach. During a median follow-up of 5.9 years, there were 423 CVE (93 fatal). In age- and sex-adjusted Cox models, UA as a continuous variable was a significant predictor of CVE in all individuals and in men and women considered separately. The relationship between UA and CVE was linear (P-value for nonlinearity 0.54 and 0.80 for men and women, respectively). For each 1 mg/dL increase in UA, the relative hazard increase was 16% in men and 19% in women. In fully adjusted models, UA remained a significant predictor of CVE in the whole study cohort. The optimal cut-point best separating patients at low and high risk of CVE was 6.3 mg/dL for men and 4.4 mg/dL for women. Subjects with high UA had a 38% greater risk of CVE. In a sex-specific analysis, the association remained significant only in men (hazard ratio, 1.47; P < 0.01). In conclusion, high UA is an independent predictor for subsequent CVE and significantly improves risk discrimination and reclassification over the baseline multivariable model.

Seasonal development of the coccidian parasite Goussia bayae and hepatobiliary histopathology in white perch Morone americana from Chesapeake Bay

The coccidium Goussia bayae infects the gallbladder and bile ducts of white perch Morone americana from Chesapeake Bay, USA. Seasonal changes in coccidian infections were analyzed from bile specimens of 1588 fish from the Choptank River during 2016-2018 using wet mount preparations with a Sedgwick-Rafter counting chamber. Histopathology of the gallbladder and liver was analyzed from a subset (n = 480) of these fish. Maximum parasite prevalence (100%) and intensities in the gallbladder occurred during the fish spawning season in March and April. Asynchronous coccidian development and prevalence of infections in fish increased gradually during autumn and winter, but coccidian intensity increased sharply 2-4 wk prior to the onset of fish spawning activity and decreased after spawning activity concluded. Sporulation was internal, and the gallbladder was the primary reservoir for oocysts. Two previously undescribed species of coccidia were observed in the intestine. Lesions in the gallbladder were rare and included cholecystitis and epithelial necrosis. Intrahepatic bile duct lesions were more common and included distension, cholangitis, epithelial erosion and necrosis, cholestasis, hyperplasia, and neoplasia. Cholangitis and necrosis of intrahepatic bile ducts were significantly associated with coccidial infections, while plasmodia of a myxosporean (spore morphology consistent with the genera Myxidium and Zschokella) were significantly associated with bile duct hyperplasia. Biliary neoplasia included cholangiomas (5% prevalence) and cholangiocarcinomas (1% prevalence). No association was detected between G. bayae and biliary neoplasms, but an association may exist between these lesions and the myxosporean plasmodia.

Using machine learning to examine medication adherence thresholds and risk of hospitalization

BACKGROUND

Quality improvement efforts are frequently tied to patients achieving ≥80% medication adherence. However, there is little empirical evidence that this threshold optimally predicts important health outcomes.

OBJECTIVE

To apply machine learning to examine how adherence to oral hypoglycemic medications is associated with avoidance of hospitalizations, and to identify adherence thresholds for optimal discrimination of hospitalization risk.

METHODS

A retrospective cohort study of 33,130 non-dual-eligible Medicaid enrollees with type 2 diabetes. We randomly selected 90% of the cohort (training sample) to develop the prediction algorithm and used the remaining (testing sample) for validation. We applied random survival forests to identify predictors for hospitalization and fit survival trees to empirically derive adherence thresholds that best discriminate hospitalization risk, using the proportion of days covered (PDC).

OUTCOMES

Time to first all-cause and diabetes-related hospitalization.

RESULTS

The training and testing samples had similar characteristics (mean age, 48 y; 67% female; mean PDC=0.65). We identified 8 important predictors of all-cause hospitalizations (rank in order): prior hospitalizations/emergency department visit, number of prescriptions, diabetes complications, insulin use, PDC, number of prescribers, Elixhauser index, and eligibility category. The adherence thresholds most discriminating for risk of all-cause hospitalization varied from 46% to 94% according to patient health and medication complexity. PDC was not predictive of hospitalizations in the healthiest or most complex patient subgroups.

CONCLUSIONS

Adherence thresholds most discriminating of hospitalization risk were not uniformly 80%. Machine-learning approaches may be valuable to identify appropriate patient-specific adherence thresholds for measuring quality of care and targeting nonadherent patients for intervention.

Categorisation of built environment characteristics: the trouble with tertiles

Background

In the analysis of the effect of built environment features on health, it is common for researchers to categorise built environment exposure variables based on arbitrary percentile cut-points, such as median or tertile splits. This arbitrary categorisation leads to a loss of information and a lack of comparability between studies since the choice of cut-point is based on the sample distribution.

Discussion

In this paper, we highlight the various drawbacks of adopting percentile categorisation of exposure variables. Using data from the SocioEconomic Status and Activity in Women (SESAW) study from Melbourne, Australia, we highlight alternative approaches which may be used instead of percentile categorisation in order to assess built environment effects on health. We discuss these approaches using an example which examines the association between the number of accessible supermarkets and body mass index.

Summary

We show that alternative approaches to percentile categorisation, such as transformations of the exposure variable or factorial polynomials, can be implemented easily using standard statistical software packages. These procedures utilise all of the available information available in the data, avoiding a loss of power as experienced when categorisation is adopted.We argue that researchers should retain all available information by using the continuous exposure, adopting transformations where necessary.

Body mass index categories in observational studies of weight and risk of death

The World Health Organization (Geneva, Switzerland) and the National Heart, Lung, and Blood Institute (Bethesda, Maryland) have developed standard categories of body mass index (BMI) (calculated as weight (kg)/height (m)(2)) of less than 18.5 (underweight), 18.5-24.9 (normal weight), 25.0-29.9 (overweight), and 30.0 or more (obesity). Nevertheless, studies of BMI and the risk of death sometimes use nonstandard BMI categories that vary across studies. In a meta-analysis of 8 large studies that used nonstandard BMI categories and were published between 1999 and 2014 and included 5.8 million participants, hazard ratios tended to be small throughout the range of overweight and normal weight. Risks were similar between subjects of high-normal weight (BMI of approximately 23.0-24.9) and those of low overweight (BMI of approximately 25.0-27.4). In an example using national survey data, minor variations in the reference category affected hazard ratios. For example, choosing high-normal weight (BMI of 23.0-24.9) instead of standard normal weight (BMI of 18.5-24.9) as the reference category produced higher nonsignificant hazard ratios (1.05 vs. 0.97 for men and 1.06 vs. 1.02 for women) for the standard overweight category (BMI of 25.0-29.9). Use of the standard BMI groupings avoids problems of ad hoc and post hoc category selection and facilitates between-study comparisons. The ways in which BMI data are categorized and reported may shape inferences about the degree of risk for various BMI categories.

Do optimal prognostic thresholds in continuous physiological variables really exist? Analysis of origin of apparent thresholds, with systematic review for peak oxygen consumption, ejection fraction and BNP

Background

Clinicians are sometimes advised to make decisions using thresholds in measured variables, derived from prognostic studies.

Objectives

We studied why there are conflicting apparently-optimal prognostic thresholds, for example in exercise peak oxygen uptake (pVO₂), ejection fraction (EF), and Brain Natriuretic Peptide (BNP) in heart failure (HF).

Data Sources and Eligibility Criteria

Studies testing pVO₂, EF or BNP prognostic thresholds in heart failure, published between 1990 and 2010, listed on Pubmed.

Methods

First, we examined studies testing pVO₂, EF or BNP prognostic thresholds. Second, we created repeated simulations of 1500 patients to identify whether an apparently-optimal prognostic threshold indicates step change in risk.

Results

33 studies (8946 patients) tested a pVO₂ threshold. 18 found it prognostically significant: the actual reported threshold ranged widely (10–18 ml/kg/min) but was overwhelmingly controlled by the individual study population's mean pVO₂ (r = 0.86, p<0.00001). In contrast, the 15 negative publications were testing thresholds 199% further from their means (p = 0.0001). Likewise, of 35 EF studies (10220 patients), the thresholds in the 22 positive reports were strongly determined by study means (r = 0.90, p<0.0001). Similarly, in the 19 positives of 20 BNP studies (9725 patients): r = 0.86 (p<0.0001).

Second, survival simulations always discovered a “most significant” threshold, even when there was definitely no step change in mortality. With linear increase in risk, the apparently-optimal threshold was always near the sample mean (r = 0.99, p<0.001).

Limitations

This study cannot report the best threshold for any of these variables; instead it explains how common clinical research procedures routinely produce false thresholds.

Key Findings

First, shifting (and/or disappearance) of an apparently-optimal prognostic threshold is strongly determined by studies' average pVO₂, EF or BNP. Second, apparently-optimal thresholds always appear, even with no step in prognosis.

Conclusions

Emphatic therapeutic guidance based on thresholds from observational studies may be ill-founded. We should not assume that optimal thresholds, or any thresholds, exist.

Well-being and employee health-how employees' well-being scores interact with demographic factors to influence risk of hospitalization or an emergency room visit

The goal of this study was to determine the relationship between individual well-being and risk of a hospital event in the subsequent year. The authors hypothesized an inverse relationship in which low well-being predicts higher likelihood of hospital use. The study specifically sought to understand how well-being segments and demographic variables interact in defining risk of a hospital event (inpatient admission or emergency room visit) in an employed population. A retrospective study design was conducted with data from 8835 employees who completed a Well-Being Assessment questionnaire based on the Gallup-Healthways Well-Being Index. Cox proportional hazards models were used to examine the impact of Individual Well-Being Score (IWBS) segments and member demographics on hazard ratios (HRs) for a hospital event during the 12 months following assessment completion. Significant main effects were found for the influence of IWBS segments, sex, education, and relationship status on HRs of a hospital event, but not for age. However, further analysis revealed significant interactions between age and IWBS segments (P=0.005) and between age and sex (P<0.0001), indicating that the effects for IWBS segments and sex on HRs of a hospital event are mediated through their relationship with age. Overall, the strong relationship between low well-being and higher risk of an event in employees ages 44 years and older is mitigated in younger age groups. These results suggest that youth attenuates the risk engendered in poor well-being; therefore, methods to maintain or improve well-being as individuals age presents a strong opportunity for reducing hospital events.

Dichotomization: 2 x 2 (x2 x 2 x 2...) categories: infinite possibilities

BACKGROUND

Consumers of epidemiology may prefer to have one measure of risk arising from analysis of a 2-by-2 table. However, reporting a single measure of association, such as one odds ratio (OR) and 95% confidence interval, from a continuous exposure variable that was dichotomized withholds much potentially useful information. Results of this type of analysis are often reported for one such dichotomization, as if no other cutoffs were investigated or even possible.

METHODS

This analysis demonstrates the effect of using different theory and data driven cutoffs on the relationship between body mass index and high cholesterol using National Health and Nutrition Examination Survey data. The recommended analytic approach, presentation of a graph of ORs for a range of cutoffs, is the focus of most of the results and discussion.

RESULTS

These cutoff variations resulted in ORs between 1.1 and 1.9. This allows investigators to select a result that either strongly supports or provides negligible support for an association; a choice that is invisible to readers. The OR curve presents readers with more information about the exposure disease relationship than a single OR and 95% confidence interval.

CONCLUSION

As well as offering results for additional cutoffs that may be of interest to readers, the OR curve provides an indication of whether the study focuses on a reasonable representation of the data or outlier results. It offers more information about trends in the association as the cutoff changes and the implications of random fluctuations than a single OR and 95% confidence interval.

Proteasome enzymatic activities in plasma as risk stratification of patients with acute myeloid leukemia and advanced-stage myelodysplastic syndrome

PURPOSE

Cytogenetic abnormalities are currently the most important predictors of response and clinical outcome for patients with acute myeloid leukemia (AML) or advanced-stage myelodysplastic syndrome (MDS). Because clinical outcomes vary markedly within cytogenetic subgroups, additional biological markers are needed for risk stratification.

EXPERIMENTAL DESIGN

We assessed the utility of measuring pretreatment proteasome chymotrypsin-like, caspase-like, and trypsin-like activities in plasma to predict response and survival of patients with AML (n = 174) or advanced-stage MDS (n = 52).

RESULTS

All three enzymatic activities were significantly (P < 0.001) increased in the plasma of patients with AML and MDS compared with normal controls. Both chymotrypsin-like and caspase-like activities, but not trypsin-like activity, correlated with outcome. Chymotrypsin-like and caspase-like activities, but not trypsin-like activity, predicted response in univariate analysis (P = 0.002). However, only chymotrypsin-like activity was independent predictor of response from age grouping (<70 versus > or =70 years), cytogenetics, and blood urea nitrogen in multivariate analysis. Similarly, both chymotrypsin-like and caspase-like activities, but not trypsin-like activity, were predictors of overall survival in univariate analysis (P < 0.0001), but only chymotrypsin-like activity was independent of cytogenetics, age, performance status, blood urea nitrogen, and beta(2)-microglobulin in multivariate Cox regression models. Chymotrypsin-like activity was also a strong independent predictor of survival in patients with intermediate karyotype (n = 124).

CONCLUSIONS

Measuring plasma chymotrypsin-like activity may provide a powerful biomarker for risk stratification in patients with AML and advanced-stage MDS, including those with normal karyotype.

Avoiding infinite estimates of time-dependent effects in small-sample survival studies

We address the phenomenon of monotone likelihood in Cox regression with time-dependent effects. Monotone likelihood occurs in the fitting process of a Cox model if at least one parameter estimate diverges to +/- infinity. We show that the probability of monotone likelihood is increased by the inclusion of time-dependent effects, particularly in small samples with several unbalanced and highly predictive covariates, and with a high percentage of censoring. Firth's bias reduction procedure was shown to provide an ideal solution to monotone likelihood. Here we extend his idea to Cox regression with time-dependent effects. By penalized maximum likelihood estimation, finite hazard ratio estimates of constant and time-dependent effects can be obtained. Penalized likelihood ratio tests and profile penalized likelihood confidence intervals are proposed as tools for inference. A Monte Carlo study of Cox regression with time-dependent effects confirms advantages of Firth-corrected (FC) over standard Cox analysis in terms of average bias and median absolute deviation. We also compare the FC and standard Cox approaches by means of analyses of two studies with time-dependent effects. An SAS macro and an R package for FC Cox regression with time-varying covariates and time-dependent effects are available at: http://www.muw.ac.at/msi/biometrie/programs.

Baseline 6-min walk distance predicts survival in lung transplant candidates

In a large, prospectively followed, two-center cohort of patients listed for lung transplantation (n = 376), we used Cox proportional hazards models to determine the importance of baseline 6-min walk distance (6MWD) in predicting patient survival. 6MWD used as a continuous variable was a significant predictor of survival after adjusting for other important covariates when transplant was considered as a time-varying covariate (HR for each 500 ft increase in 6MWD = 0.57, 95% CI: 0.43-0.77, p = 0.0002). 6MWD remained an important predictor of survival in models that considered only survival to transplant (HR for each 500 ft increase in 6MWD = 0.41, 95% CI: 0.27-0.62, p < 0.0001) or survival only after transplant (HR for each 500 ft increase in 6MWD = 0.40, 95% CI: 0.22-0.72, p = 0.002). Unadjusted Kaplan-Meier analysis demonstrates significantly different survival by 6MWD tertiles (<900, 900-1200, or >1200 ft, p-value = 0.0001). In the overall model, 6MWD prediction of survival was relatively homogeneous across disease category (6MWD by disease interaction term, p-value = 0.63). Our results demonstrate a significant relationship between baseline 6MWD and survival among patients listed for lung transplantation that exists across all native disease categories and extends through transplantation. The 6MWD is thus a useful measure of both urgency and utility among patients awaiting lung transplantation.

Systematic review and meta-analysis of real-world adherence to drug therapy for osteoporosis

OBJECTIVE

To quantify the adherence of patients to drug therapy for osteoporosis in real-world settings via a systematic review and meta-analysis of observational studies.

METHODS

The PubMed and Cochrane databases were searched for English-language observational studies published from January 1, 1990, to February 15, 2006, that assessed patient adherence to drug therapy for osteoporosis using the following medical subject headings and keywords: drug therapy, medication adherence, medication persistence, medication possession ratio, patient compliance, and osteoporosis. Studies were stratified into 3 groups: persistence (how long a patient continues therapy), compliance (how correctly, in terms of dose and frequency, a patient takes the medication), and adherence (a combination of persistence and compliance). A random-effects model was used to pool results from the selected studies.

RESULTS

Twenty-four studies were included in the meta-analysis. The pooled database-derived persistence rate was 52% (95% confidence interval [CI], 44%-59%) for treatment lasting 1 to 6 months, 50% (95% CI, 37%-63%) for treatment lasting 7 to 12 months, 42% (95% CI, 20%-68%) for treatment lasting 13 to 24 months, returning to 52% (95% CI, 45%-58%) for treatment lasting more than 24 months. Pooled adherence rates decreased from 53% (95% CI, 52%-54%) for treatment lasting 1 to 6 months to 43% for treatment lasting 7 to 12 months (95% CI, 38%-49%) or 13 to 24 months (43%; 95% CI, 32%-54%). The pooled refill compliance estimate was 68% (95% CI, 63%-72%) for treatment lasting 7 to 12 months and 68% (95% CI, 67%-69%) for treatment lasting 13 to 24 months. The pooled self-reported compliance rate was 62% (95% CI, 48%-75%) for treatment lasting 1 to 6 months and 66% (95% CI, 45%-81%) for treatment lasting 7 to 12 months.

CONCLUSION

One-third to half of patients do not take their medication as directed. Nonadherence occurs shortly after treatment initiation. Terms and definitions need to be standardized to permit comparability of technologies designed to improve patient adherence. Prospective trials are needed to assess the relationship between adherence and patient outcomes.

Categorizing BMI may lead to biased results in studies investigating in-hospital mortality after isolated CABG

OBJECTIVE

To investigate how categorizing body mass index (BMI) into weight classes can impact the assessment of the relationship between BMI and in-hospital mortality after coronary artery bypass graft (CABG) surgery.

STUDY DESIGN AND SETTING

BMI-mortality (in-hospital) relationship was assessed in 5,762 patients who underwent isolated CABG at Baylor University Medical Center (Dallas, TX) from January 1, 1997 to November 30, 2003. Different ways of modeling BMI were used to investigate this association in a propensity-adjusted model, controlling for risk factors identified by the Society of Thoracic Surgeons (STS) and other clinical/nonclinical details.

RESULTS

A highly significant (P=0.003) association between BMI (modeled with a restricted cubic spline) and mortality was found. Reduced risk of in-hospital mortality was observed for subjects with BMI in the low-30s as compared with patients with BMI in the mid-20s or over 40 kg/m(2). Results were strongly affected by the way BMI was specified in the multivariable model. Only five of the 10 BMI categorizations considered produced significant results, and these results did not fully determine the effect of BMI on mortality.

CONCLUSIONS

BMI categorization critically impacts study results. Conceivably, findings of other studies investigating BMI and adverse outcomes after CABG may be similarly affected. Investigators should consider smoothing techniques to avoid categorization.

An S-Plus function to calculate relative risks and adjusted means for regression models using natural splines

We provide for generalized linear regression models that use natural cubic splines to model predictors an S-Plus function to calculate relative risks (RR), log relative risk (logRR), mean percent change (MPC) for continuous covariates modeled using a logarithmic link as well as adjusted means differences (MD) for the identity link. The function makes explicit use of the natural spline basis functions, the estimated coefficients for each natural spline basis function, and the fitted correlation matrix for the estimated coefficients and can thus accommodate any number of degrees of freedom. The main function produces a publication-quality graph of all of these quantities as compared to a user-specified reference value as well as the associated confidence limits. In another function, specific values of these statistics comparing a vector of values of the independent variable to the reference value can be calculated rather than plotted.

Subcellular localization and protein levels of cyclin-dependent kinase inhibitor p27 independently predict for survival in epithelial ovarian cancer

PURPOSE

p27 protein is regarded as a valuable prognostic biomarker in cancer with a potential use as a molecular target. However, different methods of immunohistochemical assessment have yielded conflicting results. Here, we sought to determine the prognostic value of p27 in ovarian cancer using a novel method of compartmentalized in situ protein analysis.

EXPERIMENTAL DESIGN

A tissue array composed of 150 advanced stage ovarian cancers uniformly treated, with surgical debulking followed by platinum-paclitaxel combination chemotherapy, was constructed. For evaluation of p27 protein expression, we used an immunofluorescence-based method of automated in situ quantitative measurement of protein analysis [automated quantitative analysis (AQUA)].

RESULTS

The mean follow-up time of the patients was 34.3 months. Patients with low Fédération Internationale des Gynaecologistes et Obstetristes stage were more likely to have low nuclear p27 expression (P = 0.008). Low nuclear p27 expression was associated with improved 3-year overall survival (66% versus 20%, P = 0.0047) and disease-free survival (27% versus 12%, P = 0.022). In multivariable analysis, adjusting for well-characterized prognostic variables, low nuclear p27 expression level was the most significant prognostic factor for both disease-free and overall survival.

CONCLUSIONS

Our results indicate that quantitative assessment of nuclear p27 expression level by automated in situ quantitative analysis is a strong predictor for outcome in ovarian cancer.

Confidence intervals for the effect of a prognostic factor after selection of an ‘optimal’ cutpoint

When investigating the effects of potential prognostic or risk factors that have been measured on a quantitative scale, values of these factors are often categorized into two groups. Sometimes an 'optimal' cutpoint is chosen that gives the best separation in terms of a two-sample test statistic. It is well known that this approach leads to a serious inflation of the type I error and to an overestimation of the effect of the prognostic or risk factor in absolute terms. In this paper, we illustrate that the resulting confidence intervals are similarly affected. We show that the application of a shrinkage procedure to correct for bias, together with bootstrap resampling for estimating the variance, yields confidence intervals for the effect of a potential prognostic or risk factor with the desired coverage.

An epidemiologic approach to drug prescribing quality assessment: a study in primary care practice in France

OBJECTIVE

To assess drug prescribing by primary care physicians in France for various types of conditions, and to identify patterns and risk factors for poor prescribing quality.

METHODS

The orders (n = 23,080) written for patients with five target diseases (acute nasopharyngitis, acute tonsillitis, essential hypertension, osteoarthrosis, and back and periarticular disorders), by primary care physicians (n = 1,049) were extracted from a nationwide prescription database and analyzed according to 17 quantitative indicators of drug prescribing quality constructed on explicit a priori criteria.

RESULTS

Ineffective drugs were prescribed in 32% to 88% of orders according to the target disease. Six percent to 40% of orders resulted in drug interactions, age problems, and overdosage. A consistent pattern of associations between indicators was found, which suggests that drug prescribing quality is multidimensional and is composed of at least five dimensions: placebo, novelty, exoticism, misdosage, and interaction. Several factors associated with indicators were also identified, some of them defining groups of patients at risk (women, elderly, and less educated), physicians at risk (women, aged, and isolated), and contexts at risk (patient's home and disease frequently treated by the physician) of poor drug prescribing quality.

CONCLUSIONS

Drug prescribing by French primary care physicians appears nonoptimal, in terms of both risk of iatrogeny and waste of money. This study further documents the complexity and the multidimensionality of drug prescribing quality. It suggests that more attention must be paid to patients' and physicians' risk factors for poor drug prescribing quality if educational programs and regulatory processes are to succeed in promoting safer and more cost-effective practices.

Rapid risk household screening by neonatal arm circumference: results from a cohort study in rural Burkina Faso

Neonatal arm circumference (NAC) and other attributes of the newborn and its household were analysed as potential predictors of child death in a cohort of 1367 newborn children representing the majority of births in a rural area of Burkina Faso from 1992 to 1994. During 3872 person years observed 264 children died, resulting in an average mortality rate of 6.8% per year. 90 mm was chosen as the best cut-off to differentiate low NAC associated with high mortality from normal NAC. The hazard ratio of children with low NAC (15.7%) compared to others was 1.7 (P < 0.001) in Cox regression. Kaplan-Meier curves of cumulative survival showed that this higher risk lasted throughout the first two years of life. Multivariate Cox regression comparing NAC with other variables known or suspected to influence child survival yielded a model including mother's death, twin birth, affiliation to a particular health centre, home delivery and birth during the rainy or harvest season as other significant risk factors beside NAC. Protective factors were mother's participation in antenatal care despite considerable distance to the health centre, medium household size (5-7 members) and household cash crop production. We propose a simple risk score for rapid household screening in rural Burkina Faso and comparable settings elsewhere for identifying households at risk of experiencing child death. As much of the other variables' contribution to the explanation of survival pattern is absorbed by NAC in more parsimonious models, even simpler screening strategies based on NAC make sense. In the study area risk households will be offered periodical home visits by the local nurse promoting immunization, treatment of illness and strengthening the mothers' competence to recognize and manage frequent health problems of their children as part of a 'Shared Care' concept.

Statistical methods in epidemiology: a comparison of statistical methods to analyze dose-response and trend analysis in epidemiologic studies

Evaluation of various statistical methods to describe accurately associations between exposures and disease are constantly being explored. Spline regression has been suggested as an alternative to using categorized variables in studies of disease etiology, as it uses all data points to estimate the shape of the association between a given exposure and disease outcome. It has been proposed that this method is especially beneficial when associations are concentrated in a small range of the overall distribution of the exposure. In this study, we use data from a large case-control study of colon cancer to evaluate associations obtained from logistic regression models that use spline regression for main exposure and confounder effects with those that use categorized variables for main exposure. Our results show that for variables for which the association appears to be linear, such as body size and dietary intake of calcium, fiber, and cholesterol, associations are similar when estimates are generated from spline or categorized variable models. For other variables, such as total energy intake, for which associations appear to be strongest in the upper end of the distribution, estimates of association appear to be conservative when using categorized variables. The data also suggest that selection of cut points for the categorized variables may have an impact on the associations observed. Spline regression appears to be useful to estimate the shape of the association between a given exposure and disease and may provide guidance as to the appropriateness of using categorized variables. The risk estimates from spline regression appear to be similar to those from traditional categorical methods. When effects are large or rapidly changing, spline models may more appropriately describe the association.

Factors affecting survival after orthotopic liver transplantation in infants

The technical and medical management of small infants requiring orthotopic liver transplantation remains a challenge. The present study examined 117 orthotopic liver transplantations performed in 101 infants from <1 to 23 months of age between March 1988 and February 1995 to determine factors that influence patient and graft outcome. Factors analyzed included etiology of liver disease, recipient and donor age and weight, United Network for Organ Sharing (UNOS) status, retransplantation, ABO-compatibility, full-size (FS) versus reduced-size grafts, vascular thrombosis (VT), including hepatic artery and portal vein (PVT), and the presence of lymphoproliferative disease (LPD). UNOS status 1, fulminant hepatic failure, and the development of Epstein-Barr virus-associated LPD were each associated with 10-20% lower patient and graft survival rates. Of 101 infants, 11 (11%) developed LPD with an associated 36% mortality. VT occurred in 10 (9 hepatic artery and 1 portal vein) of 117 orthotopic liver transplantations (9%), all less than 1 year of age, and was associated with significantly poorer 1-year (50% vs. 85% no VT, P<0.01) and 5-year patient survival rates (50% vs. 83% no VT, P<0.01). One-year graft survival rates for FS grafts in recipients <12 months versus 12-23 months were 67% vs. 94% (P<0.01); the patient survival rate was also significantly lower in FS graft recipients <12 months (76% vs. 100%, P<0.05). Recipients <5 months of age had the worst survival rates: 1-year and 5-year patient survival rates were 65% and 46% for recipients 0-4 months (n=17) versus 82% and 82% for recipients 5-11 months (n=56), and 93% and 93% for recipients age 12-23 months (n=28; P<0.05). In summary, factors associated with reduced survival rates include recipient age <5 months, recipient age <12 months who received FS grafts, development of VT and donor weight <6 kg. There was a trend for UNOS status 1, fulminant hepatic failure, and presence of LPD to be associated with reduced survival rates.