Measurement of liver fat by MRI and its reduction by dexfenfluramine in NIDDM.

OBJECTIVE: To validate a variation of a well-established magnetic resonance imaging (MRI) technique to detect liver fat and use it to monitor liver fat changes after treatment with dexfenfluramine in men with non-insulin dependent diabetes mellitus (NIDDM). DESIGN: (a) Simple correlation study of MRI Liver Fat Index with liver biopsy results; (b) Open Study of 10 men with NIDDM treated with dexfenfluramine for 12 weeks in addition to their 'usual' therapy. SUBJECTS: (a) 19 patients (3F; 16M) with abnormal liver function tests undergoing liver biopsy; (b) 10 men, Body Mass Index (BMI) < 30, Waist to hip ratio (WHR) > 0.90 with poorly controlled NIDDM despite oral sulphonylurea therapy. MEASUREMENTS: (a) MRI liver fat; standard liver biopsy; (b) MRI visceral fat, MRI liver fat, euglycaemic clamp, plasma lipids, fasting glucose and c-peptide levels. RESULTS: In the validation group, there was a strong relationship between the MRI Liver Fat Index and histopathological assessment of the liver biopsies (r = 0.87, < 0.0001). During treatment with dexfenfluramine the mean Liver Fat Index reduced from 10.6 +/- 3.4 to 6.6 +/- 2.8 (P = 0.05). The reduction in Liver Fat Index correlated with the reduction in visceral fat (r = 0.84, P = 0.001) as well as with the improvement in insulin sensitivity (r = 0.62, P = 0.05). Using partial correlation analysis, the relationship between the change in visceral adipose tissue and the improvement in insulin sensitivity was weaker if the Liver Fat Index was kept constant (r = 0.76 decreased to r = 0.56). CONCLUSIONS: In this group of subjects MRI Liver Fat Index correlated well with liver fat as seen on biopsy. The Liver Fat Index reduced after 12 weeks therapy with dexfenfluramine suggesting a role for hepatic steatosis in the complex interaction between visceral adipose tissue and insulin sensitivity.

Influence of moderate physical exercise on insulin-mediated and non-insulin-mediated glucose uptake in healthy subjects.

To establish the relative importance of insulin sensitivity and glucose effectiveness during exercise using Bergman's minimal model, 12 nontrained healthy subjects were studied at rest and during 95 minutes of moderate exercise (50% maximum oxygen consumption [VO2max]). Each subject underwent two frequently sampled intravenous glucose tolerance tests (FSIGTs) for 90 minutes, at rest (FSIGTr) and during exercise (FSIGTe). Plasma glucose, insulin, and C-peptide were determined. Insulin sensitivity (S(I)), glucose effectiveness at basal insulin (S(G)), insulin action [X(t)], and first-phase (phi1) and second-phase (phi2) beta-cell responsiveness to glucose were estimated using both minimal models of glucose disposal (MMg) and insulin kinetics (MMi). Glucose effectiveness at zero insulin (GEZI), glucose tolerance index (K(G)), and the area under the insulin curve (AUC(0-90)) were also calculated. Intravenous glucose tolerance improved significantly during physical exercise. During exercise, S(I) (FSIGTr v FSIGTe: 8.5 +/- 1.0 v 25.5 +/- 7.2 x 10(-5) x min(-1) [pmol x L(-1)]-1, P < .01), S(G) (0.195 +/- 0.03 v 0.283 +/- 0.03 x 10(-1) x min(-1), P < .05), and GEZI (0.190 +/- 0.03 v 0.269 +/- 0.04 x 10(-1) x min(-1), P < .05) increased; however, no changes in phi1 and phi2 were found. Despite a significant decrease in the insulin response to glucose (AUC0-90, 21,000 +/- 2,008 v 14,340 +/- 2,596 pmol x L(-1) x min, P < .01), insulin action [X(t)] was significantly higher during the FSIGTe. These results show that physical exercise improves mainly insulin sensitivity, and to a lesser degree, glucose effectiveness. During exercise, the insulin response to glucose was lower than at rest, but beta-cell responsiveness to glucose did not change.

Ketosis resistance in fibrocalculous pancreatic diabetes: II. Hepatic ketogenesis after oral medium-chain triglycerides.

A majority of patients with fibrocalculous pancreatic diabetes (FCPD) do not become ketotic even in adverse conditions. It is not clear whether this ketosis resistance is due to reduced fatty acid release from adipose tissue or to impaired hepatic ketogenesis. We tested hepatic ketogenesis in FCPD patients using a ketogenic challenge of oral medium-chain triglycerides (MCTs) and compared it with that in matched insulin-dependent diabetes mellitus (IDDM) patients and healthy controls. After oral MCTs, FCPD patients showed only a mild increase in blood 3-hydroxybutyrate (3-HB) concentrations (median: fasting, 0.13 mmol/L; peak, 0.52) compared with IDDM patients (fasting, 0.44; peak, 3.39) and controls (fasting, 0.04; peak, 0.75). Plasma nonesterified fatty acid (NEFA) concentrations were comparable in the two diabetic groups (FCPD: fasting, 0.50 mmol/L; peak, 0.79; IDDM: fasting, 0.91; peak, 1.04). Plasma C-peptide concentrations were low and comparable in the two diabetic groups. Plasma glucagon concentrations were higher in IDDM patients in the fasting state, but declined to levels comparable to those in FCPD patients after oral MCTs. Plasma carnitine concentrations were comparable in the two groups of patients. It is concluded that the failure to stimulate ketogenesis under these conditions could be partly due to inhibition of a step beyond fatty acid entry into the mitochondria.

Reduction of visceral adipose tissue and improvement of metabolic indices: effect of dexfenfluramine in NIDDM.

Increased visceral adipose tissue is thought to contribute to impaired glucose tolerance. We studied 10 men with non-insulin dependent diabetes (NIDDM) before and after a 12-week intervention study using dexfenfluramine. Subjects had a mean body mass index (BMI) of 26.4 +/- 1.7 kg/m2 and had an abdominal distribution of body fatness (waist-to hip ratio > 0.9). Anthropometric indices, biochemistry, macronutrient intake from 7-day food records as well as a euglycaemic glucose clamp and magnetic resonance imaging (MRI) were performed at week 0 and week 12. Abdominal adipose tissue area measured by MRI was reduced from 854 +/- 270 cm2 to 666 +/- 231 cm2 (p = 0.003) due mainly to a selective 32% reduction in visceral fat area from 484 +/- 230 cm2 to 333 +/- 72 cm2 (p = 0.002). Insulin sensitivity improved from 0.29 +/- 0.13 [min-1 (mU/L)] to 0.54 +/- 0.21 [min-1 (mU/L)] (p = 0.01) and C-peptide levels reduced from 0.77 +/- 0.24 mumol/L to 0.58 +/- 0.15 mumol/L (p = 0.002). The reductions in fasting glucose and glycated haemoglobin failed to achieve significance. Fasting total cholesterol and triglyceride levels significantly reduced (p = < 0.001 and p = 0.021 respectively). There was a reduction in total energy intake (p = 0.005) due to a significant reduction in calories obtained from fat (p < 0.001). Thus dexfenfluramine was shown to be a useful adjunct therapy for the reduction of visceral fat in abdominally-obese men with NIDDM with an associated improvement in insulin sensitivity.

Problems associated with using in vivo proton (1H) magnetic resonance spectroscopy to quantify liver fat.

In-vivo 1H magnetic resonance (MR) spectra of the liver were obtained in 8 patients admitted for liver biopsy. These patients had abnormal liver function and the presumptive diagnosis of fatty liver prior to biopsy. Two patients with NIDDM were also studied but liver biopsies were not performed as liver function was normal. The MR spectra, obtained on a 60 cm clear-bore 1.9 tesla superconducting magnet showed two 1H resonances, one from water and the other from repeating methylene protons - (CH2)n - in triglyceride. The lipid: water signal ratio was used to characterize tissues as subcutaneous fat (high lipid:water ratio), normal liver (low lipid: water ratio) and fatty liver (intermediate lipid: water ratio). The spectra obtained at the greatest depth from the probe surface ~4.5 cm) was used as it was most likely to represent liver tissue. Although all 8 patients were expected to have fatty liver only 2 had evidence of significant fatty changes on microscopy. This was assessed by counting the vacuoles of fat over the area of the biopsy specimen and quantitated as 'fat vacuoles per high power field' (f/hpf). In the 2 patients with NIDDM, unusual stack plots suggested technical difficulties with 1H MR spectroscopy for in-vivo assessment of fatty liver. The first patient, PT had a significant increase in lipid:water ratio on the spectra thought to represent liver (lipid:water ~ 65% cf levels <3% in norma liver and 12.6% + 26.5% in those patients subsequently found to have fat on biopsy). This was later found on MR imaging to represent omental fat lying between the liver and muscle layer. The second patient, OM had a large amount of subcutaneous fat overlying the area assessed. As seen on the stack plot, the probe depth was not great enough to pass through the subcutaneous fat and muscle layer to penetrate liver tissue. There was a significant correlation between the lipid:water signal ratio and visible fat on biopsy in those patients who underwent liver biopsy. Difficulties experienced with probe depth suggests imaging would be necessary prior to spectroscopy to ensure liver tissue is actually assessed.

Glycaemic responses in type 2 diabetic patients to various mixed meals taken at home.

Eight Type 2 diabetic patients ate and prepared five different meals at home, taking each meal on two separate occasions. They measured their blood glucose just before eating and 30, 60, 120, and 180 min after the meal. The meals varied in energy and dietary fibre content and in the ratio (by energy) of carbohydrate to fat. Total energy content of the meals had little effect on the postprandial glycaemic responses nor were the responses reduced by meals with high dietary fibre content. The ratio of carbohydrate to fat did not significantly affect postprandial glycaemic responses when meals were low in fibre. However, postprandial glycaemic responses were significantly greater in the meal with a high ratio of carbohydrate to fat, high in fibre and low in energy compared with those after the equicaloric meal low in carbohydrate to fat ratio and low in fibre (area under the curve 683 +/- 131 vs 306 +/- 55 mmol l-1 min-1, p < 0.05). Fat intake of 35% of energy may be compatible with improved postprandial blood glucose concentrations. Many meal combinations need to be studied in order to provide reliable information for diabetic patients. The method outlined proved producible (within patient coefficient of variation 13%), easy to perform and inexpensive.

The effects of changes in plasma nonesterified fatty acid levels on oxidative metabolism during moderate exercise in patients with non-insulin-dependent diabetes mellitus.

Blood levels of intermediary metabolites were measured and indirect calorimetry was performed in 10 otherwise healthy, non-insulin-dependent diabetic (NIDDM) patients before, during, and after 30 minutes of moderate exercise on three occasions in random order at weekly intervals with (1) heparin treatment to increase preexercise plasma nonesterified fatty acid (NEFA) levels (HEPARIN); (2) acipimox, a nicotinic acid analogue, to reduce preexercise plasma NEFA levels (ACIPIMOX); and (3) no manipulation of preexercise plasma NEFA levels (NIL). With ACIPIMOX, preexercise blood levels were significantly reduced for NEFAs and glycerol (P < .01) and marginally reduced for acetoacetate and 3-hydroxybutyrate (NS) compared with preexercise levels for the other two treatments; these low levels seen with acipimox treatment increased only slightly during exercise and the postexercise period. Plasma NEFA levels increased by approximately 150% (P < .001) with HEPARIN at the same times. The levels of ketone bodies during either NIL or HEPARIN increased rapidly postexercise by approximately 90% to 110% for both acetoacetate and 3-hydroxybutyrate (both P < .01). Plasma insulin levels tended to be lowest (despite similar plasma glucose levels during the three treatments) with ACIPIMOX, while growth hormone (hGH) and, perhaps, noradrenaline levels were highest both during and after exercise. The respiratory quotient (RQ) was highest with ACIPIMOX (P < .05 for exercise and postexercise periods compared with the other two treatments), which, compared with NIL, reduced fat oxidation by 27% and 60% and increased carbohydrate oxidation by 29% and 74% during and after exercise, respectively (all P < .05). These changes in substrate oxidation due to ACIPIMOX were almost opposite to those observed with HEPARIN.(ABSTRACT TRUNCATED AT 250 WORDS)

The relationship between obesity, plasma immunoreactive insulin concentration and blood pressure in newly diagnosed Indian type 2 diabetic patients.

The association of blood pressure with clinical and biochemical measures was studied in 185 newly diagnosed Type 2 diabetic patients, 74 impaired-glucose-tolerant (IGT) and 128 non-diabetic control subjects. Hyperglycaemic subjects were older than control subjects (controls 40 (24-59) years, IGT 48 (29-64) years, diabetic 43 (29-60) years, median (5th-95th centile) both p < 0.05). They were also more obese (body mass index (BMI) controls 23.5 kg m-2 (17.2-29.9), IGT 26.0 kg m-2 (19.8-33.9), diabetic 24.2 kg m-2 (19.3-32.2)) and with a greater waist-hip ratio (controls 0.83 (0.70-0.98), IGT 0.88 (0.75-0.98), diabetic 0.89 (0.75-1.00)). Blood pressure was significantly higher in both IGT (systolic 127 mmHg (108-162), diastolic 84 mmHg (66-99)) and diabetic patients (systolic 130 mmHg (104-160), diastolic 84 mmHg (66-102)) compared to non-diabetic controls (systolic 120 mmHg (100-151), diastolic 80 mmHg (60-94)). Univariate analysis showed that in diabetic patients systolic blood pressure was related to age (r = 0.17, p < 0.05), BMI (r = 0.23, p < 0.01) and plasma immunoreactive insulin (fasting and post glucose, r = approximately 0.25, p < 0.01) but not to C-peptide concentrations; diastolic blood pressure to BMI (r = 0.35, p < 0.001), waist-hip ratio (r = 0.23, p < 0.01) and plasma immunoreactive insulin (fasting r = 0.30, p < 0.001, post glucose r = approximately 0.20, p < 0.05) but not to C-peptide concentrations.(ABSTRACT TRUNCATED AT 250 WORDS)

Distribution of type II diabetes in nuclear families.

Type II diabetes has a substantial genetic component, but the mode of inheritance and the molecular basis of this inheritance are uncertain. This study documents the familial distribution of the disease in the parents and siblings of a consecutive series of type II diabetic subjects. We studied 66 first-degree relatives of 20 white subjects with type II diabetes and both parents alive. They were tested with a continuous infusion of glucose (5 mg.kg IBW-1.min-1) (n = 49) or FPG and hemoglobin A1c (n = 17). Seven probands had neither parent affected with diabetes or IGT, 10 had one parent affected (6 with diabetes and 4 with IGT), and 3 had both parents affected. The probands with affected and those with unaffected parents were phenotypically similar. These findings indicate that a sizable subgroup of type II diabetic subjects may have neither parent affected with a demonstrable abnormality of glucose tolerance. The assumption of autosomal dominance with complete penetrance is not supported, although it remains possible that a dominant gene of low penetrance may play a role in some pedigrees. Polygenic inheritance would appear likely, and genetic heterogeneity may occur. The inheritance of diabetic traits from phenotypically normal parents needs to be considered in the analysis of genetic linkage with type II diabetes.

Adipose tissue metabolism in obesity: lipase action in vivo before and after a mixed meal.

Physiological actions of insulin include suppression of fat mobilization from adipose tissue and activation of adipose tissue lipoprotein lipase. Here, we report measurements of adipose tissue hormone-sensitive lipase (HSL) and lipoprotein lipase (LPL) action in vivo in 10 normal and eight obese subjects, with the latter group having varying degrees of glucose intolerance. HSL and LPL actions (per gram of adipose tissue) were similar in the two groups, after an overnight fast. In the normal subjects, HSL action was suppressed after a meal (by 75% +/- 6% between 60 to 300 minutes, P less than .01), and the action of LPL was increased (clearance of circulating triacylglycerol [TAG] increased by 140% +/- 57% at 300 minutes, P less than .05). Despite hyperinsulinemia, these responses were blunted in the obese subjects (P less than .05 for each change being less than in normal group). The adipose tissue of the obese subjects showed continued nonesterified fatty acid (NEFA) release at a time when NEFA mobilization was completely suppressed in the normal group. Both impaired suppression of HSL and low fractional retention of fatty acids for reesterification within the adipose tissue contributed to this abnormal NEFA release. Impaired activation of LPL was associated with a greater absolute increase in plasma TAG concentration postprandially in the obese. In obese subjects, adipose tissue HSL and LPL fail to respond to immunoreactive insulin postprandially, which may be an important maladaptation in terms of lipoprotein metabolism and risk of coronary heart disease.

The ketosis-resistance in fibro-calculous-pancreatic-diabetes. 1. Clinical observations and endocrine-metabolic measurements during oral glucose tolerance test.

We measured circulating levels of C-peptide, pancreatic glucagon, cortisol, growth hormone and metabolites (glucose, non-esterified fatty acids, glycerol and 3-hydroxybutyrate) in fibro-calculous-pancreatic diabetic (FCPD, n = 28), insulin-dependent diabetic (IDDM, n = 28) and non-diabetic control (n = 27) subjects during an oral glucose tolerance test. There was no difference in the two diabetic groups in age (FCPD 24 +/- 2, IDDM 21 +/- 2 years, mean +/- SEM), BMI (FCPD 16.0 +/- 0.6, IDDM 15.7 +/- 0.4 kg/m2), triceps skinfold thickness (FCPD 8 +/- 1, IDDM 7 +/- 1 mm), glycaemic status (fasting plasma glucose, FCPD 12.5 +/- 1.5, IDDM 14.5 +/- 1.2 mmol/l), fasting plasma C-peptide (FCPD 0.13 +/- 0.03, IDDM 0.08 +/- 0.01 nmol/l), peak plasma C-peptide during OGTT (FCPD 0.36 +/- 0.10, IDDM 0.08 +/- 0.03 nmol/l) and fasting plasma glucagon (FCPD 35 +/- 4, IDDM 37 +/- 4 ng/l). FCPD patients, however, showed lower circulating concentrations of non-esterified fatty acids (0.73 +/- 0.11 mmol/l), glycerol (0.11 +/- 0.02 mmol/l) and 3-hydroxybutyrate (0.15 +/- 0.03 mmol/l) compared to IDDM patients (1.13 +/- 0.14, 0.25 +/- 0.05 and 0.29 +/- 0.08 mmol/l, respectively). This could be due to enhanced sensitivity of adipose tissue lipolysis to the suppressive action of circulating insulin and possibly also to insensitivity of hepatic ketogenesis to glucagon. Our results also demonstrate preservation of alpha-cell function in FCPD patients when beta-cell function is severely diminished, suggesting a more selective beta-cell dysfunction or destruction than hitherto believed.

Central rather than generalized obesity is related to hyperglycaemia in Asian Indian subjects.

The relationship of body mass index and waist-hip ratio with plasma glucose concentrations during an oral glucose tolerance test (OGTT) was studied in native Indian (Asian) subjects. A total of 389 subjects (131 non-diabetic, 74 impaired glucose tolerant (IGT) and 184 Type 2 diabetic (newly diagnosed and untreated] were studied. Prevalence of obesity (BMI greater than or equal to 27.0 kg m-2 in men and greater than or equal to 25.0 kg m-2 in women, 21% and 47%, respectively) was lower in people with Type 2 diabetes than that reported in white Caucasian and migrant Asian populations. Body mass index was highest in IGT subjects (26.1 (19.7-34.3) kg m-2, median (5-95th centile] and was higher in diabetic subjects (24.2 (19.3-32.2) kg m-2) than in non-diabetic control subjects (23.5 (17.1-30.0) kg m-2). However, waist-hip ratio was higher in both IGT (0.88 (0.75-0.98)) and diabetic subjects (0.88 (0.75-1.00)) than in non-diabetic control subjects (0.83 (0.70-0.97)), with no difference between the hyperglycaemic groups. On multivariate analysis, fasting as well as 2-h plasma glucose concentrations during OGTT were found to be related to waist-hip ratio (p less than 0.01) and subscapular fat thickness (p less than 0.01) but not to body mass index (or triceps fat thickness). Thus, in native Indians central obesity seems to be a more important association of hyperglycaemia than generalized obesity.

Factors controlling fat mobilization from human subcutaneous adipose tissue during exercise.

To investigate possible factors that limit fat utilization during exercise, arteriovenous differences of plasma nonesterified fatty acids (NEFA) and glycerol were measured across the subcutaneous adipose tissue of the anterior abdominal wall in nine subjects who exercised for 60 min at 50-70% of their maximal O2 consumption. The large gradient of NEFA concentration from adipose tissue venous to arterial plasma increased throughout the exercise period. Maximal plasma NEFA concentrations in adipose venous drainage were reached postexercise (median 3,800 mumol/l), with a median NEFA-to-albumin molar ratio of 5.7. Fractional reesterification of fatty acids within the tissue (assessed from the ratio of NEFA to glycerol release) was 20-30% in the basal state and declined during exercise. After exercise there was apparently negative reesterification, implying release of NEFA retained in adipose tissue during exercise. Although these findings challenge current views on the regulation of NEFA release, they are in agreement with the concept of supply of fatty acids from adipose tissue as the major factor limiting fat oxidation during sustained exercise.

Comparison of indices of islet B-cell function in type 2 diabetes in relation to insulin effectiveness and clinical outcome.

The most appropriate way to estimate islet B-cell function in Type 2 diabetes is unclear, and this has led to many different techniques of measurement being used. We have examined the associations to two fasting and four glucose-stimulated indices of islet B-cell function in members of a group of 249 Type 2 patients, seeking correlations with concurrent glucose tolerance and antilipolytic effect, and with subsequent clinical outcome. The six B-cell indices were interrelated to variable degrees (rs -0.21 to +0.92). Early glucose-stimulated insulin output (incremental 1st-phase insulin area) was not significantly positively correlated with the fasting plasma concentration of immunoreactive insulin at any time. Fasting immunoreactive insulin and 'minimal model' islet B-cell parameters were poorly related to the rate constant for glucose clearance and the degree of antilipolysis (rs values between -0.13 and +0.40). Homeostatic model assessment of the fasting islet B-cell function was more consistently related to these metabolic effects. Incremental first-phase insulin area was the islet B-cell index most consistently related to metabolic abnormalities (rs up to +0.56), and to subsequent need for oral hypoglycaemic or exogenous insulin therapy. No index of islet B-cell function was consistently associated with the subsequent development of diabetic tissue damage.

Breath hydrogen excretion or plasma acetate levels during the lactulose tolerance test?

Since both acetate and hydrogen are produced by colonic bacterial fermentation, the clinical utility of the measurement of either parameter in nine subjects for the lactulose tolerance test was tested. The fasting plasma acetate concentration (mean +/- s.d., mmol/l) of 0.11 +/- 0.06 increased to peak levels between 150 min (0.23 +/- 0.12) and 180 min (0.23 +/- 0.09), both P less than 0.01, after ingesting 20 g lactulose. In one subject with previous gastrectomy and intestinal hurry, the peak was at 30 min. Mean post-lactulose acetate levels (0.21 +/- 0.09) were higher than fasting levels (P less than 0.03). Breath hydrogen excretion exhibited a similar trend. Indeed, a significant correlation (rs 0.39, P less than 0.01) was demonstrated between the acetate and hydrogen values. It is therefore concluded that patients for the lactulose breath test show fairly similar changes in plasma acetate and breath hydrogen excretion after lactulose ingestion. Either measurement could thus be used in assessing colonic fermentation in humans.

The formation of acetate from ethanol with and without prior chlorpropamide intake in diabetic and non-diabetic subjects.

It has been suggested that raised post-ethanol plasma acetaldehyde levels, from inhibition of aldehyde dehydrogenase, underlie the liability to chlorpropamide, alcohol flushing (CPAF). We tested the hypothesis that acetate formation from acetaldehyde, the reaction catalysed by that enzyme, was also likely to be affected by chlorpropamide (CP) medication. In six healthy non-diabetic 'non-flushers', fasting acetate (Ac +/- s.d. mmol/l) was 0.22 +/- 0.12, and increased by 0.47 +/- 0.14 to peak levels by 30 min after intake of 40 ml dry sherry, which increased plasma ethanol (mmol/l) levels to 10.2 +/- 6.0. After 5 days of CP (250 mg daily), fasting Ac (0.17 +/- 0.05) and increase to peak of Ac and ethanol after 40 ml sherry (0.56 +/- 0.12 and 8.9 +/- 7.2 respectively), were not changed (P n.s.). There was no correlation between Ac and ethanol at any time point. When the studies were repeated in five non-insulin-dependent diabetic 'flushers', both on regular CP medication and after 3 days without CP, there was again no significant difference in fasting and post-ethanol Ac levels between the two studies (fasting 0.18 +/- 0.04 v. 0.17 +/- 0.02, and increase to peak 0.62 +/- 0.13 v. 0.72 +/- 0.18, P n.s.). These results indicate that the conversion of ethanol to acetate is unaffected by CP medication, and furthermore that post-ethanol acetate levels do not predict liability to CPAF.