D. Holiday

PRESS-related statistics : regression tools for cross-validation and case diagnostics

In the health science literature, a common approach of validating a regression equation is data-splitting, where a portion of the data fits the model (fitting sample) and the remainder (validation sample) estimates future performance. The R2 and SEE obtained by predicting the validation sample with the fitting sample equation is a proper estimate of future performance, tending to correct for the natural upward bias of the R2 and SEE obtained from fitting sample alone. Data-splitting has several disadvantages, however. These include: 1) difficulty, arbitrariness, and inconvenience of matching samples; 2) the need to report two sets of statistics to determine homogeneity; and 3) the lack of equation stability due to diluted sample size. The PRESS statistic and associated residuals do not require the data to be split, yield alternative unbiased estimates of R2 and SEE, and provide useful case diagnostics. This procedure is easy to use, is widely available in modern statistical packages, but is rarely utilized. The two methods are contrasted here using a simulation from original data for predicting body density from anthropometric measurements of a group of 117 women. The PRESS approach is particularly appropriate for smaller datasets; methods of reporting these statistics are recommended.

Taxol-induced cell cycle arrest and apoptosis: dose-response relationship in lung cancer cells of different wild-type p53 status and under isogenic condition

The effective dose, schedule, molecular basis of the cytotoxicity of taxol and their dependence on the genetic background in tumor cells are still not well understood. Here, we examined how the dose-response relationship for taxol varies in lung cancer cells with different p53 status and under isogenic conditions. DNA content analyses in A 549 (p53, +/+) and H 1299 (p53, -/-) cells, showed that taxol progressively induced G2/M arrest in both cell lines in a concentration-dependent manner, which was accompanied by a parallel decrease in the G1 population. G2/M arrest, however, occurred at a lower concentration in A 549 cell lines than in H 1299 cells. The S-phase population in A 549 cells was not significantly changed up to 0.025 microM, but dropped by six-fold at 1.0 microM taxol, in contrast to that in H 1299 cells. A sub-G1 apoptotic population was present at 24 h, even at 0.002 microM taxol, when G2/M arrest was not appreciably detected. In both cell lines, the maximum apoptosis of about 28% was achieved at 0.025 microM taxol, implicating that wild-type p53 does not modulate the level of taxol-induced apoptosis. When we examined the role of the wild-type p53 in isogenic cell lines developed in a H 1299 background, the maximum level of apoptosis was in the range of 28-34% at a drug concentration around 0.03 microM, not significantly different from that observed in parental H 1299 cells. We conclude that taxol is effective in inducing apoptosis at very low doses (0.020-0.035 microM), and that the presence or absence of the wild-type p53 does not make a statistically significant difference in the level of apoptotic cell death in these lung cancer cell lines, but the maximum is attained at a lower drug concentration in the presence of p53.

Criterion validity of the Duke Activity Status Index for assessing functional capacity in patients with chronic obstructive pulmonary disease

PURPOSE This study evaluated the concurrent criterion validity of the Duke Activity Status Index (DASI) with respect to standard physiologic work capacity indices in patients with chronic obstructive pulmonary disease (COPD) and compared its performance with similar surrogates. METHODS 119 patients with moderate to severe COPD (86 men, 33 women) completed medical and smoking histories, physical examination, pulmonary function testing (PFT), cycle ergometry (CE), arm ergometry (AE), and 6-minute walk distance (6MWD), DASI, the Sickness Impact Profile-68 (SIP-68) and the Chronic Respiratory Disease Questionnaire (CRDQ). Correlation methods were used to assess the validity of the potential surrogates DASI and the domain scores for SIP-68 and CRDQ, with the standards CE, AE, PFT, and 6MWD (as a standard). RESULTS The mean DASI score was 33.4 +/- 13.0. Significant Pearson correlations (P <.01) were observed between the DASI and PFT outcomes maximum voluntary ventilation (r =.28); peak expiratory flow (r =.21); diffusion capacity of lung for carbon monoxide (r =.30). For CE, the correlations with DASI were oxygen consumption (VO(2))(r =.34); minute ventilation (r =.25); watts (r =.37). For AE, the correlations with DASI were VO(2) (r=.38); watts (r =.47). For 6MWD, the correlation was r =.53. Higher correlations were obtained for the distance completed during the first minute of the 6MWD and ergometric indices as well as DASI scores: watts(AE) (r =.39); VO(2AE) (r =.45); watts(CE) (r =.50); VO(2CE) (r =.44). Correlation coefficients for all SIP-68 and CRDQ domain and total scores were lower than corresponding correlations obtained for the DASI. Regression analysis demonstrated that the DASI and 6MWD were important (P <.05) for predicting VO(2) or work for CE while DASI and SIP range or CRDQ dyspnea entered for AE, when gender, age, BMI, and the FEV1 were forced into the model. In forward stepwise analyses, DASI entered first for AE, and 6MWD entered first for CE. The DASI was selected in 3 of 4 models with R(2) values ranging from.47 to.70. SIP-68 and CRDQ subscores were significant as additional predictors. CONCLUSIONS DASI has high criterion validity for predicting CE and/or AE outcomes in the COPD population. It is warranted in addition to the 6MWD, and its predictive significance and simplicity recommends it over several other self-administered instruments for evaluating functional capacity.

Peak Physiologic Responses to Arm and Leg Ergometry in Male and Female Patients With Airflow Obstruction

STUDY OBJECTIVE To investigate differences in work capacity for the arms and legs in patients with moderate-to-severe COPD. DESIGN Cross-sectional investigation. PATIENTS One hundred twenty-four patients (90 men and 34 women) aged 45 to 81 years with moderate-to-very severe COPD. FEV(1) ranged from 0.70 to 2.79 L/min (FVC, 1.73 to 5.77 L; FEV(1)/FVC, 24 to 69%). All patients were in stable condition at the time of testing and receiving a stable drug regime. MEASUREMENTS Each patient completed a demographic and medical history questionnaire, pulmonary function studies (spirometry, lung volumes, and diffusion capacity), peak exercise ergometry with gas exchange for the arms and legs; they also rated their subjective assessment of perceived dyspnea and extremity fatigue using Borg scores during exercise. RESULTS Patients were of comparable age, with men taller and heavier than women. Smoking history was significantly less for women (47.9 pack-years vs 66.6 pack-years for men) even though each group presented with equivalent age (p > 0.05). Women were less obstructed than men, with FEV(1)/FVC (mean +/- SD) of 46.5 +/- 10.9% vs 40.2 +/- 9.3%, respectively. Ventilatory limitation during exercise was noted for all patients studied. Peak work capacity was greater for men, and leg peak responses were greater than arm values for each gender. As airway obstruction increased, work capacity became more limited. Peak arm work achieved was 38.9 +/- 19.6 W, oxygen uptake (VO(2)) was 903.9 +/- 263.5 mL/min, and minute ventilation (VE) was 33.7 +/- 9.5 L. Peak leg work value was 62.9 +/- 24.8 W, VO(2) was 1,091.4 +/- 321.5 mL/min, and VE was 39.3 +/- 12.0 L. Hence, arm values were 62%, 83%, and 85% of the measured leg values, respectively. Dyspnea and extremity effort scores were similar for men and women, and for arms and legs. Regression analysis was used to derive prediction equations for arm work from measured leg ergometry testing. For watts of work, a three-variable model emerged explaining 66% of the variance; VO(2) yielded a four-variable model with 80% of the variance explained; and VE yielded a three-variable model explaining 72% of the variance. CONCLUSION Arm work is reduced by 38% that of the legs, while more modest reductions are noted for VO(2) and VE, suggesting greater mechanical efficiency for leg work as compared to arm work. These data also suggest greater metabolic demand for respiratory muscles and arm ergometry. Dyspnea and extremity Borg scores were equivalent for each modality and level of airway obstruction studied, suggesting that perception plays an important role in limiting exercise, and that a threshold for termination of exercise may exist. Further, peak leg ergometry results can be used with pulmonary function indexes to predict peak arm workload in watts, VO(2), and VE. These data may be used to assist the clinician in prescribing rehabilitation or estimating arm exercise ability when arm testing is unavailable.

Up-regulation of neutral endopeptidase (CALLA) in human neutrophils by granulocyte-macrophage colony-stimulating factor.

Neutral endopeptidase 24.11 (NEP/CALLA/ CD10), an enzyme expressed on early lymphoid progenitors, neutrophils, and various other cell types, inactivates many biologically active peptides, including the bacterial chemo‐ tactic peptide M‐formylmethionyl‐leucyl‐phenylalanine (fMLP). Inhibition of CD10/NEP on the surface of human neutrophils (PMNs) in vitro inhibits migration toward this chemotaxin, suggesting that enzymatic inactivation by NEP regulates the neutrophil response to fMLP. Because PMNs in inflammatory sites are exposed to various cytokines, we evaluated the effects of selected cytokines on CD10/NEP activity in vitro. Of five cytokines tested—interleukin‐1 (IL‐1), IL‐6, and IL‐8, granulocyte colony‐stimulating factor, and granulocyte‐ macrophage colony‐stimulating factor (GM‐CSF) —GM‐ CSF provided the most consistent increase in surface NEP activity. Low concentrations (10−9‐10−7 M) of GM‐ CSF increased NEP activity in a time‐ and concentration‐ dependent manner to more than 225% that of control (phosphate‐buffered saline‐treated) cells. Cytofluorome‐ try of cells stained with a fluorescent antibody to CD10 indicated that GM‐CSF increased expression of surface CD10/NEP antigen in a similar manner. The effect of GM‐CSF on NEP activity was enhanced still further by simultaneous exposure to IL‐1, suggesting that combinations of cytokines may direct and regulate the neutrophil response within an inflammatory site. Rapid up‐ regulation of CD10/NEP underscores the importance of this enzyme for control of peptide mediators of inflammation.

Near optimal weights in nonparametric regression under some common restrictions

A fixed design nonparametric regression model having potentially correlated replicate measurements is considered. Optimal minimizers of the mean averaged squared error are derived under restrictions of symmetry and nonnegativity/normalization. The feasibility of a direct estimation strategy, naturally utilizing correlation, is discussed.