Metabolic modeling using statistical and spreadsheet software: Application to the glucose minimal model.

Kinetic non-linear metabolic models are used extensively in medical research and increasingly for clinical diagnostic purposes. An example of such a model is the Glucose Minimal Model by Bergman and colleagues [1]. This model is similar to pharmacokinetic/pharmacodynamic models in that like pharmacokinetic/pharmacodynamic models, it is based on a small number of fairly simple ordinary differential equations and it aims to determine how the changing concentration of one blood constituent influences the concentration of another constituent. Although such models may appear prima facie, to be relatively simple, they have gained a reputation of being difficult to fit to data, especially in a consistent and repeatable fashion. Consequently, researchers and clinicians have generally relied on dedicated software packages to do this type of modeling. This article describes the use of statistical and spreadsheet software for fitting the Glucose Minimal Model to data from an insulin modified intravenous glucose tolerance test (IM-IVGTT). A novel aspect of the modeling is that the differential equations that are normally used to describe insulin action and the disposition of plasma glucose are first solved and expressed in their explicit forms so as to facilitate the estimation of Glucose Minimal Model parameters using the nonlinear (nl) optimization procedure within statistical and spreadsheet software. The most important clinical parameter obtained from the Glucose Minimal Model is insulin sensitivity (SI). Using IM-IVGTT data from 42 horses in one experiment and 48 horses in a second experiment, we demonstrate that estimates of SI derived from the Glucose Minimal Model fitted to data using STATA and Excel, are highly concordant with SI estimates obtained using the industry standard software, MinMod Millennium. This work demonstrates that there is potential for statistical and spreadsheet software to be applied to a wide range of kinetic non-linear modeling problems.

Asthma and incident cardiovascular disease: the Atherosclerosis Risk in Communities Study.

BACKGROUND: A possible association between asthma and cardiovascular disease has been described in several exploratory studies. METHODS: The association of self-reported, doctor diagnosed asthma and incident cardiovascular disease was examined in a biracial cohort of 45-64 year old adults (N = 13501) followed over 14 years. RESULTS: Compared with never having asthma, the multivariate adjusted hazard ratio (HR) of stroke (n = 438) was 1.50 (95% CI 1.04 to 2.15) for a baseline report of ever having asthma (prevalence 5.2%) and 1.55 (95% CI 0.95 to 2.52) for current asthma (prevalence 2.7%). The relative risk of stroke was 1.43 (95% CI 1.03 to 1.98) using a time dependent analysis incorporating follow up reports of asthma. Participants reporting wheeze attacks with shortness of breath also had greater risk for stroke (HR = 1.56, 95% CI 1.18 to 2.06) than participants without these symptoms. The multivariate adjusted relative risk of coronary heart disease (n = 1349) was 0.87 (95% CI 0.66 to 1.14) for ever having asthma, 0.69 (95% CI 0.46 to 1.05) for current asthma at baseline, and 0.88 (95% CI 0.69 to 1.11) using the time dependent analysis. CONCLUSIONS: Asthma may be an independent risk factor for incident stroke but not coronary heart disease in middle aged adults. This finding warrants replication and may motivate a search for possible mechanisms that link asthma and stroke.

Sleep disorders in regional sleep centers: a national cooperative study. Coleman II Study Investigators.

STUDY OBJECTIVE: In the last two decades there has been an increase in the awareness of and professional expertise in sleep disorders. The objective of this study was to determine the spectrum of sleep-related disorders diagnosed in regional sleep centers and compare this to a previous survey published in 1982. DESIGN: A two-month prospective point-prevalence survey SETTING: Nineteen accredited regional sleep centers in the United States. PARTICIPANTS: Patients evaluated at regional sleep centers during a two-month period. INTERVENTIONS: NA. RESULTS: Obstructive sleep apnea, narcolepsy, and restless legs syndrome were the top three reported primary diagnoses with a prevalence of 67.8%, 4.9%, and 3.2%, respectively. The entire range of sleep disorders, however, was represented in the study sample. Nearly a third of patients had either a primary or secondary diagnosis of a non-respiratory sleep disorder. Referral physicians were most likely to be from internal medicine, pulmonary medicine, and otolaryngology. Compared to the previous survey from 1982, there has been an absolute increase in patient referrals/center with a two- to four-fold increase in the number of patients/center with a final diagnosis of a non-respiratory sleep-related problem. Moreover, there has been a greater than twenty-fold increase in the diagnosis of obstructive sleep apnea. CONCLUSION: Regional sleep centers are encountering increasing patient referrals and a broad range of sleep-related disorders. The predominant reasons for referral are related to obstructive sleep apnea, narcolepsy, and restless legs syndrome.

Modeling hypersomnolence in sleep-disordered breathing. A novel approach using survival analysis.

The etiology of excessive daytime sleepiness in patients with sleep-disordered breathing (SDB) is not well defined. In this study, we examined the relationships between several clinical and polysomnographic parameters and the degree of hypersomnolence in 741 patients with SDB (apnea-hypopnea index [AHI] >/= 10 events/h). The study sample was obese (body mass index [BMI]: 35.3 +/- 8.5 kg/m2) and had evidence of moderate SDB (AHI: 47.6 +/- 29.3 events/h). Hypersomnolence was quantified with the multiple sleep latency test (MSLT) and survival analysis was used to assess the risk factors for hypersomnolence. In a multivariate proportional hazards model, AHI and nocturnal hypoxemia were independent predictors of hypersomnolence (MSLT < 10 min). The adjusted relative risks (RR) of hypersomnolence were 1.00, 1.30, and 1.65 for patients with an AHI of 10 to 29.9, 30 to 59.9, and >/= 60 events/h, respectively. A positive association between hypersomnolence and oxyhemoglobin desaturation (DeltaSaO2) was observed with RR of 1.00, 1.18, 1.43, and 1.94 for a DeltaSaO2 of 15%, respectively. Sleep fragmentation, as assessed by the distribution of sleep stages, was also an independent predictor of hypersomnolence. Using stage 1 sleep as a reference, an increase in stage 2 and slow wave sleep (SWS) were protective from hypersomnolence. For a 10% increase in stage 2 or SWS the adjusted RR for hypersomnolence were 0.93 and 0.79, respectively. REM sleep showed no significant association with the degree of hypersomnolence. These results suggest that AHI, nocturnal hypoxemia, and sleep fragmentation are independent determinants of hypersomnolence in SDB.

Correction of single-breath helium lung volumes in patients with airflow obstruction.

STUDY OBJECTIVE: To determine whether alveolar volume (V(A)) measured during the single-breath diffusing capacity for carbon monoxide (DCO) can be used as a substitute measure for the multiple-breath total lung capacity (TLC) in subjects with and without airways obstruction. DESIGN: Retrospective review of pulmonary function test (PFT) results. SETTING: Pulmonary function laboratories at the Johns Hopkins Hospital (JHH) and the Johns Hopkins Asthma and Allergy Center (HAAC). PARTICIPANTS: Patients referred for spirometry, helium lung volumes, and Dco during a single visit between November 1993 and November 1996. RESULTS: JHAAC patients (n=2,477) were used to assess the relationship between V(A) and TLC. In patients with an FEV1/FVC > or = 0.70, there was good agreement between V(A) and TLC (V(A)/TLC=0.97 to 0.99). However, in patients with an FEV1/FVC <0.70, V(A) systematically underestimated TLC (V(A)/TLC=0.67 to 0.94). The degree of underestimation was related to the severity of airflow obstruction. To predict TLC using V(A) a regression equation was used to "correct" V(A) for the severity of obstruction. This equation was used to predict the multiple-breath TLC for JHH patients (n=2,892). Patients with an FEV1/FVC > or = 0.70 showed a high degree of correlation between V(A) and TLC (Pearson's correlation coefficient [r]=0.96 to 0.99; p<0.05). After adjusting for the severity of airflow obstruction, patients with an FEV1/FVC in the range of 0.40 to 0.70 also had a strong correlation between the corrected V(A) and the multiple-breath TLC (r=0.83 to 0.94; p<0.05). CONCLUSIONS: V(A) accurately predicts TLC in patients with mild or no airflow obstruction. For patients with moderate to severe obstruction, correcting V(A) for the severity of obstruction improves the accuracy of this relatively simple and rapid technique for measuring TLC.