Framework for quality assessment of knowledge

One of the key issues in the development (and subsequent application) of medical knowledge-be it in terms of a KBS or otherwise-is the assessment of its quality. We present a framework for how to manage and make measurable the quality of the semantic as well as pragmatic aspects of the knowledge embedded in classification models during the development of such models.

KAVAS-2: Knowledge Acquisition, Visualization and Assessment System

The objective of KAVAS-2 is the development of a tool, named KAVIAR, with which domain experts can make their knowledge explicit. It contains components for (computer assisted) knowledge elicitation and for machine learning. A key issue in KAVAS is the assessment of the quality of the classification and domain models built. Various quality measures are available and implemented in KAVIAR to assess the quality of models, specifically those developed from data bases by machine learning techniques.

On the quality of neural net classifiers

This paper describes several concepts and metrics that may be used to assess various aspects of the quality of neural net classifiers. Each concept describes a property that may be taken into account by both designers and users of neural net classifiers when assessing their utility. Besides metrics for assessment of the correctness of classifiers we also introduce metrics that address certain aspects of the misclassifications. We show the applicability of the introduced quality concepts for selection among several neural net classifiers in the domain of thyroid disorders.

OpenLabs: the application of advanced informatics and telematics for optimization of clinical laboratory services

OpenLabs has four major objectives: to improve the efficiency and effectiveness of clinical laboratory services by the integration of Knowledge Based Systems (KBSs) with Laboratory Information Systems (LISs) and equipment; to provide and implement standard solutions for Electronic Data Interchange (EDI) between laboratories and other medical systems; to specify a fully Open architecture for an integrated Clinical LIS and demonstrate the integration of various KBS modules on the open architecture platform; and to demonstrate the integration of OpenLabs modules with existing LISs.

Circadian rhythmicity of interferon-gamma production in antigen-stimulated whole blood

Recently, assessment of T cell function has been refined by the ability to measure cytokines produced by activated T cells. We developed a whole blood assay to detect antigen-activated T cells that produce IFN-gamma. With this assay we have found a large circadian variation in tetanus- (acrophase 00(00) p < 0.001) and PPD- (acrophase 00(08) p < 0.001) stimulated IFN-gamma production. IFN-gamma production is inversely correlated with plasma cortisol (r = -0.5), suggesting that variation in IFN-gamma production may be secondary to circadian variation in plasma cortisol levels (acrophase 11(06)). The demonstration of circadian rhythmicity in antigen-stimulated IFN-gamma production is relevant to the diagnostic use of whole blood assays and, in addition, may have implications for the therapy of immuno-inflammatory diseases.

A methodology for evaluation of knowledge-based systems in medicine

Evaluation is critical to the development and successful integration of knowledge-based systems into their application environment. This is of particular importance in the medical domain--not only for reasons of safety and correctness, but also to reinforce the users' confidence in these systems. In this paper we describe an iterative, four-phased development evaluation cycle covering the following areas: (i) early prototype development, (ii) validity of the system, (iii) functionality of the system, and (iv) impact of the system.

Do local hospitals meet the analytical goals for the use of common reference intervals?

Blood samples were obtained from 183 blood donors and measured in two hospitals in the same geographical area using identical reference intervals and measuring equipment. The analytical bias of the clinical chemical analyses was investigated through paired measurements. The results showed that out of 15 routine analyses only four meet the goals for acceptance of common reference intervals, of which at least three were of clinical significance. The problems were reflected in the quality assessment results but, nevertheless, no action seems to have been taken.

Information technology factors in transferability of knowledge based systems in medicine

The history of knowledge based systems in medicine has been that they are generally very localised, serving a special need in a single setting. Very few have proven to be capable of transfer to a distant environment. With the advent of tele-medical services and the associated transfer of data and knowledge in such services, the ability of medical KBS to transfer will be crucial to the success of tele-medical services. Differences in knowledge acquisition methods, knowledge representation techniques and in the epidemiological composition of training databases may influence viable transfer of knowledge based systems. Through experiments we demonstrate how rule-based systems may impose inflexible demands on data, how different knowledge acquisition techniques acquire different aspects of knowledge, though trained on a common training database, and how different knowledge acquisition techniques show varying degrees of robustness to slight changes in training databases.

Clinical utility of an international thyroid database

Assessment of thyroid function is prone to errors from several sources. Confusion is most often due to inappropriate use of tests, especially in patients with acute non-thyroid illness. With these problems in mind we have designed an international prospective thyroid database in three countries, which registers clinical and laboratory data on new patients with suspected thyroid disease. An additional aim was to assess the use of a computerised decision support tool to interpret thyroid data. The database permits rapid access to temporal trends in thyroid tests, which is useful in monitoring therapy and in follow up for hyperthyroidism. Marked contrasts in local clinical practices have highlighted the challenge in providing a valid decision tool to serve all clinical needs. Experience with multi-centre databases such as this hold promise in the drive to coordinate the disciplines of laboratory analysis and clinical decision making.

Problem-oriented management of laboratory work through dynamic test scheduling

This paper gives an overview of problems inherent in dynamic test scheduling together with some model solutions. Dynamic test schedules are decision tree-like protocols for cost-efficient management of analytical production in a clinical laboratory. The present analysis is based on previous practical experience and concludes that it is not feasible to introduce dynamic test scheduling on a large scale without computer-based support, because of the increase in complexity of the laboratory work processes. Further, our experience is that it is extremely complex to incorporate the dynamic test scheduling functionality into an existing Laboratory Information System (LIS). The approach pursued in the OpenLabs (A2028) AIM Project for implementing dynamic test scheduling is to provide the necessary functionality as a stand-alone module interconnected with a LIS in an open systems solution.

Bone metabolism in obesity: changes related to severe overweight and dietary weight reduction

A non-invasive evaluation of bone metabolism was performed in 44 morbidly obese patients before and after a mean weight loss of 22.4 kg (range 7.9-43.4 kg) after 2 months and a further weight loss of 7.3 kg after 8 months (0.8-20.0 kg). This weight reduction was obtained by a nutritionally adequate very-low-calorie diet. Before treatment the bone mineral content of the distal forearm was increased compared to normals (51.9 U vs. 43.7 U, p < 0.001). Bone formation was evaluated by serum alkaline phosphatase and serum osteocalcin. Serum alkaline phosphatase was increased (187.8 U/l vs 147.4 U/l, p < 0.001) while serum osteocalcin was lower than in the controls (0.67 nmol/l vs 0.98 nmol/l, p < 0.01). Bone resorption, as measured by the urinary hydroxyproline/creatinine ratio, was not increased in the obese patients (19.2 molar ratio x 10(-3) vs 16.7 molar ratio x 10(-3), NS). After 2 months, the bone mineral content had declined by 3.3%. Serum alkaline phosphatase remained unchanged (187.8 U/l vs 186.9 U/l, NS) but serum osteocalcin demonstrated a significant rise (3.94 nmol/l vs 10.53 nmol/l, p < 0.001), parallel to changes in the hydroxyproline/creatinine ratio (19.2 molar ratio x 10(-3) vs 25.2 molar ratio x 10(-3), p < 0.001). At 8 months, no further change in the bone mineral content was seen. The hydroxyproline/creatinine ratio did still increase (from 25.8 molar ratio x 10(-3) to 30.1 molar ratio x 10(-3), p < 0.05), while serum alkaline phosphatase and serum osteocalcin remained unchanged.(ABSTRACT TRUNCATED AT 250 WORDS)

Influence of oral magnesium supplementation on cardiac events among survivors of an acute myocardial infarction

OBJECTIVE

To investigate the effect of long term oral magnesium treatment on incidence of cardiac events among survivors of an acute myocardial infarction.

DESIGN

Double blind, placebo controlled parallel study in which patients were randomised to treatment or placebo.

SETTING

Two coronary care units and corresponding outpatient clinics.

SUBJECTS

468 survivors of an acute myocardial infarction (289 men and 178 women) aged 31-92.

INTERVENTIONS

One tablet of 15 mmol magnesium hydroxide or placebo daily for one year.

MAIN OUTCOME MEASURES

Incidences of reinfarction, sudden death, and coronary artery bypass grafting in one year.

RESULTS

There was no significant difference between treatment and placebo groups in the incidence of each of the three cardiac events, but when the events were combined and drop outs were excluded from calculations there was a significantly higher incidence of events in the treatment group (56/167 v 33/153; relative risk 1.55 (95% confidence interval 1.07 to 2.25); p = 0.02). When the timing of events was incorporated by means of a Kaplan-Meier plot the treatment group showed a significantly higher incidence of events whether drop outs were included or excluded (p < 0.025).

CONCLUSION

Long term oral treatment with 15 mmol magnesium daily doses not reduce the incidence of cardiac events in survivors of an acute myocardial infarction and, indeed, seems to increase the risk of developing a cardiac event. Consequently, this treatment cannot be recommended as secondary prophylaxis for such patients.

Cica-clamp technique: a method for quantifying parathyroid hormone secretion: a sequential citrate and calcium clamp study

Objectives of the present study were to establish and investigate a standardized method for quantifying of intact parathyroid hormone secretion during sequential induction of hypo- and hypercalcaemia, and to explore the applicability to these data of a mathematical model derived from in vitro studies as presented in the literature. Twenty-two healthy volunteers aged 20-80 years participated in one or more experiments. The experiments comprised three different protocols of sequential induction of a regular hypocalcaemic (citrate) clamp followed by increases in blood ionized calcium, ending in a regular hypercalcaemic (calcium) clamp. During protocol I, the induction of hypocalcaemia, blood ionized calcium 0.21 mmol l-1 (SD 0.01, n = 76) below baseline, the release of serum parathyroid hormone rapidly increased to a concentration of four to seven times above baseline. The serum parathyroid hormone declined gradually to a steady state of about two to three times above baseline. During stepwise increases in blood ionized calcium, the serum parathyroid hormone rapidly declined to new steady state concentrations. When a hypercalcaemia of 0.20 mmol l-1 (SD 0.02, n = 76) above baseline was reached, serum parathyroid hormone was suppressed to about one-fourth of baseline concentration. Protocol II, the Cica-clamp, and protocol III, are short versions of protocol I using a slow and gradual increase in blood ionized calcium from hypo- to hypercalcaemia.(ABSTRACT TRUNCATED AT 250 WORDS)

The effects of intraarticular administration of hyaluronan in a model of early osteoarthritis in sheep. I. Gait analysis and radiological and morphological studies

Using a model of early osteoarthritis (OA) induced in ovine joints by medial meniscectomy, the intraarticular effects of two hyaluronic acid (HA) preparations (AHA and DHA) were investigated. DHA was an HA preparation with an average molecular weight (MW) of approximately 2.0 x 10(6) d, and AHA had a MW of approximately 8 x 10(5) d. Animals (n = 5) were injected intraarticularly with 1 mL (10 mg/mL) of either HA preparation once a week for 5 weeks beginning 16 weeks after initiation of arthropathy. Meniscectomized, saline (1.0 mL)-injected animals (n = 5) and nonoperated sheep (n = 5) were used for controls. Force-plate analysis of gait and radiographic changes in joints were evaluated in these groups before and after intraarticular treatment. At necropsy, cartilage gross morphology, osteophyte development, and cartilage histopathology were examined. Meniscectomized joints were characterized by erosions and fissuring of cartilage of the medial compartment with areas of decreased matrix staining for proteoglycans. Osteophytes were present at the medial joint margins. Saline-treated meniscectomized animals showed reduced loading of the operated limb using the force plate. Force-plate analysis of walking animals before and after treatment with either AHA or DHA indicated some normalization of joint loading. However, osteophyte scores for meniscectomized joints injected with AHA and DHA were higher after treatment than those of the corresponding saline-treated group. Although the gross cartilage damage was lower than in saline-treated controls for both the HA-treated groups, the histological scores did not support this conclusion. Indeed, the tibial score for the DHA group was higher than for the AHA group (P < .05). These studies confirmed previous reports that meniscectomy of sheep stifle (knee) joints resulted in matrix changes similar to those described for early OA in humans. Both HA preparations appeared to improve gait, suggesting decreased lameness. Increased joint loading associated with gait improvement may account for the higher osteophyte scores in the treated groups. However, cartilage damage with DHA was found to be higher than when the lower-MW HA preparation (AHA) was used.

Filtering clinically significant hypercalcaemia from non-significant hypercalcaemia at the laboratory level

In an earlier study it was shown, that, in a hospital population, only about 30% of patients with hypercalcaemia had the diagnosis identified and so an alarm filter was developed, at the laboratory level, to differentiate between clinically significant and clinically non-significant hypercalcaemia. The filter correctly classified 84% and 77% of women and men with hypercalcaemia, respectively, with respect to the clinical significance of the hypercalcaemia. The sensitivity and the specificity of the discriminant functions were 90% and 73% in women and 80% and 67% in men. The alarm filter was based on a two-step procedure. The patients were primarily classified by the discriminant functions into one of the following medical diagnoses related to hypercalcaemia: primary hyperparathyroidism, malignancy, and transient hypercalcaemia. Based on this classification the patients were characterized as having either a clinically significant or non-significant hypercalcaemia. The alarm filter was based on two discriminant functions. The discriminant variables were phosphate, albumin, protein, LD, and a haemoglobin in women, and albumin, protein, ALP, and age in men. Missing values were estimated. The discriminant variables were selected from 17 possible discriminant variables, all measured by routine, and age. The study group comprised 257 patients with hypercalcaemia, consecutively registered, during half a year. The classification efficiency of the discriminant functions were based on comparison of results obtained by cross-validation of the discriminant functions, and the medical diagnosis decided by the clinicians in the respective departments of the hospital. The medical diagnoses were based on principles and definitions generally used in the departments with no knowledge of the results assessed by the discriminant functions.

Changing incidence and residual lifetime risk of common osteoporosis-related fractures

Changes in incidence and lifetime risk of fractures are of major importance in the epidemiology of osteoporosis. We focused on hip fractures in women and men and on radial and humeral fractures in women. The study subjects comprised 4500 women and men 20 years old or more with fractures. In women 1735 fractures of the distal radius, 747 fractures of the proximal humerus, 878 cervical and 635 trochanteric hip fractures were included. In men 273 cervical and 232 trochanteric hip fractures were included. The fractures were registered during the period 1976 to 1984 and changes in age-specific incidence were calculated (chi-squared test for linear trend; p-values less than 0.05 were considered significant). On the basis of life tables and population background data, the lifetime risk was estimated. The incidence of cervical hip fractures in women aged 60-89 years decreased significantly (p < 0.05) during the observation period, while no significant decrease was found in the incidence of trochanteric fractures. No significant changes in incidence were observed in women with radial or humeral fractures, or in men with hip fractures. A women 60 years old with a life expectancy of 81 years had an estimated residual lifetime risk of radial, humeral or hip fracture of 17%, 8% and 14% respectively. A man 60 years of age with a life expectancy of 77 years had an estimated risk of hip fracture of 6%.(ABSTRACT TRUNCATED AT 250 WORDS)

Radial and humeral fractures as predictors of subsequent hip, radial or humeral fractures in women, and their seasonal variation

Hip fractures are common in elderly women, and early risk assessment of future hip fractures is relevant in relation to prevention. We studied the predictive value of radial and humeral fractures in women. The influence of weather conditions on the risk was also studied. Women aged 20-99 years with a fracture of the distal radius (n = 1162) or proximal humerus (n = 406) were followed for 0 to 9 years. The relative risk (RR) and 95% confidence limits (CL) of subsequent fracture among women suffering radial or humeral fractures compared with the background population were calculated. Women 60-79 years of age who had suffered a fracture of the distal radius or proximal humerus had relative risks of sustaining a hip fracture of 1.9 (1.3-2.6, 95% CL) and 2.5 (1.3-3.6, 95% CL) respectively. The relative risk of hip fracture was highest within the first years following a fracture of the radius or the humerus. Women suffering an upper extremity fracture (radius or humerus) in snowy or icy weather had a marginally increased risk (RR = 1.3, 0.4-2.3, 95% CL and RR = 1.8, 0.3-3.4, 95% CL) for a later hip fracture. A woman 50 years old with a radial or a humeral fracture had an estimated residual lifetime risk of sustaining a subsequent hip fracture of 17% and 16% respectively compared with 11% for the background population.(ABSTRACT TRUNCATED AT 250 WORDS)

Animal models of early osteoarthritis: their use for the evaluation of potential chondroprotective agents

Medial meniscectomy was undertaken in adult merino sheep and after 16 weeks exercise each group was administered five weekly intra-articular injections of saline, pentosan polysulphate (PPS), hyaluronic acid (HA) or a combination of PPS + HA. Gait analysis and x-rays were undertaken before and after drug treatment. At sacrifice (26 weeks), joints were examined for gross pathological and histochemical changes. Only the PPS-treated group showed an improvement in gait, with low radiological and histology scores. The HA-treated group showed similar but less significant changes to these parameters.

Cardiovascular risk factors during estrogen-norethindrone and cholecalciferol treatment

The effect of cholecalciferol and estrogen-norethindrone treatment on total cholesterol level, high-density lipoprotein cholesterol level, blood pressure, and body mass index was investigated in 74 postmenopausal women in a double-blind, randomized trial. Blood pressure and body mass index did not change throughout the study. We demonstrated a decrease (11%) in serum cholesterol level after 1 year of treatment with estrogen-norethindrone. When this treatment was combined with cholecalciferol, a similar decrease (13%) was observed. The hypocholesterolemic effect was correlated to body mass index in a way that indicated the most pronounced decrease in lean women. The high-density lipoprotein cholesterol/total cholesterol fraction increased by 45% after 1 year of estrogen-norethindrone treatment, while an increase of 25% after 1 year was seen when cholecalciferol was added to the treatment. The latter increase was not different from a similar increase in the placebo group. The possible dyslipidemic effect of cholecalciferol, along with the risk of hypercalcemia, emphasizes the caution necessary in cholecalciferol treatment.

Calcium clamp technique: suppression of serum intact PTH by induced hypercalcaemia in normal man and primary hyperparathyroidism

The aim of the study was to investigate the interrelation between induced hypercalcaemia and serum intact parathyroid hormone (S-PTH(1-84)) in normal man and in patients with primary hyperparathyroidism (PHPT) by measuring blood ionized calcium (B-Ca++) and S-PTH(1-84) before and during a controlled calcium infusion. Guided by frequent measurements of B-Ca++, we adjusted the calcium infusion rate continuously, thereby keeping B-Ca++ in a steady state at a pre-determined level approximately 0.25 mmol l-1 above baseline values. This calcium clamp technique (CCT) applied to 14 normal volunteers for 120 min established a standardized reference for parathyroid suppression and the renal physiological PTH response. The reproducibility of the method and the results obtained by the CCT were satisfactorily assessed in six of the 14 normal subjects. In normal subjects B-Ca++ was raised from 1.25 +/- 0.3 mmol l-1 (mean +/- SD) to 1.49 +/- 0.02 mmol l-1 suppressing S-PTH(1-84) to 264 +/- 9.9% of pre-infusion levels. We applied the CCT to 10 patients with PHPT for 120 min raising B-Ca++ from 1.41 +/- 0.09 mmol l-1 to 1.69 +/- 0.08 mmol l-1, thereby suppressing S-PTH(1-84) to 47.9 +/- 16.3% of pre-infusion levels. The renal handling of calcium and phosphate during CCT demonstrates the biological effects of suppressed activity of PTH on the renal tubules showing increments in the maximal tubular phosphate reabsorption in relation to the glomerular filtration rate (TmP/GFR) and decreased tubular reabsorption fraction of calcium. The described CCT is a safe and reliable dynamic test.(ABSTRACT TRUNCATED AT 250 WORDS)

Normal pattern of parathyroid response to blood calcium lowering in primary hyperparathyroidism: a citrate clamp study

OBJECTIVE

The objective of the present study was to elucidate the parathyroid responsiveness by measurements of blood ionized calcium and serum intact parathyroid hormone (PTH) concentrations, before and during trisodium citrate induced hypocalcaemia.

PATIENTS AND CONTROLS

Sixteen patients with primary hyperparathyroidism and 32 healthy volunteers.

DESIGN

Blood ionized calcium concentration was lowered by about 0.20 mmol/l and maintained at this level for 2 hours by blood ionized calcium controlled trisodium citrate infusion.

MEASUREMENTS

Serum PTH(1-84) was measured by an immunoradiometric assay.

RESULTS

In patients and controls, baseline measurements of blood ionized calcium were 1.39 +/- 0.07 vs 1.24 +/- 0.04 mmol/l (mean +/- SD) (P < 0.001) and of serum PTH (1-84) 9.7 +/- 5.4 vs 3.2 +/- 1.1 pmol/l (P < 0.001). During a trisodium citrate clamp, serum PTH(1-84) rose to a maximal concentration after 5-10 minutes in both groups, the patients to 2-10 times baseline, whereas controls rose to 4-7 times baseline values. In both groups the peak of serum PTH(1-84) declined to a steady state concentration around 2-4 times baseline.

CONCLUSIONS

In conclusion, adenoma cells seem to react in almost the same way as normal parathyroid cells. They respond to initiation of hypocalcaemia by the release of preformed PTH(1-84), and continue to secrete increased amounts of PTH(1-84) during the maintenance of relative hypocalcaemia. The increased baseline concentrations of blood ionized calcium and serum PTH(1-84) and the serum PTH(1-84) response during blood ionized calcium lowering all suggest a shift upwards in the calcium set point.

Regulation of acute parathyroid hormone release in normal humans: combined calcium and citrate clamp study

The objective of the present study was to elucidate the dynamics of parathyroid hormone regulation, with particular reference to the mechanism controlling the acute parathyroid hormone release. Through utilization of the citrate clamp technique and the calcium clamp technique we were able, in a standardized way, to stimulate and suppress the parathyroid hormone secretion. Precise bedside measurements of blood ionized calcium and measurements of intact parathyroid hormone were performed. Twelve healthy young volunteers participated in two trials 6-12 wk apart, a citrate clamp (delta-blood ionized calcium -0.19 mmol/l) and a calcium plus citrate clamp (delta-blood ionized calcium +0.22 mmol/l and -0.19 mmol/l). During the citrate clamp, preceded by normal calcemia, serum intact parathyroid hormone peaked to a maximum after 5-10 min, four to six times above baseline concentration and then declined to a steady state two to three times above baseline concentration. During the citrate clamp, preceded by hypercalcemia induced by a calcium clamp, serum intact parathyroid hormone also peaked immediately to about five to nine times above its suppressed level, approximately two times above the baseline concentration. Subsequently, serum intact parathyroid hormone declined to a steady state just below the baseline concentration. In conclusion, within the range studied, the mechanism eliciting the acute serum intact parathyroid hormone release from its depot is a fall in blood ionized calcium, not the absolute concentration of ionized calcium.

The effect of noise and biases on the performance of machine learning algorithms

This paper describes the results of experiments with a machine learning algorithm for the induction of classification trees. We mainly address the impact of noise on the resulting classification tree and on the classification results obtained with the derived tree. We use the domain of the biochemical assessment of thyroid diseases as an example. Some suggestions for quality assessment are outlined that should be available in tools that assist users in deriving classification trees in noisy domains.

Hypocalcemia and parathyroid hormone responsiveness in diabetes mellitus: a tri-sodium-citrate clamp study

The aim of this study was to elucidate the diabetic hypocalcemia and PTH responsiveness, investigated by measuring blood ionized calcium and serum intact parathyroid hormone (S-PTH(1-84)) concentrations, before and during an induced and maintained controlled hypocalcemia. In 15 patients with insulin-dependent diabetes mellitus and 19 healthy volunteers the blood ionized calcium concentration was lowered by about 0.20 mmol/l and maintained at this level by blood ionized calcium controlled tri-sodium-citrate infusion. In patients vs controls, baseline measurements averaged for blood ionized calcium (mmol/l) 1.18 +/- 0.08 vs 1.24 +/- 0.03 (p less than 0.01), for S-magnesium (mmol/l) 0.73 +/- 0.07 vs 0.81 +/- 0.07 (p less than 0.01) and for S-PTH (1-84) (pmol/l) 3.0 +/- 1.0 vs 3.1 +/- 1.0 (p greater than 0.75). During the clamp, S-PTH (1-84) peaked to comparable maximums after 5-10 min in both groups and then declined to constant concentrations two to three times above their control levels. In conclusion, we found a diabetic hypocalcemia and hypomagnesemia, though baseline levels of PTH and PTH responsiveness were normal. This may be taken to indicate a mild shift downwards in the set-point for PTH secretion in patients with insulin-dependent diabetes mellitus.

Does Calcium Supplementation Reduce the Risk of Urinary Oxalate Calculi after Jejunoileal Bypass for Morbid Obesity?

Twelve patients having received an end-to-side jejunoileal bypass operation for morbid obesity 6-10 years previously were studied over three periods of 2 weeks each. The first period was used for baseline observations without any treatment. For the second period patients were randomly assigned to equimolar (75 mmol/day) oral calcium supplements administered either as a slow release or as a bolus calcium preparation. During the third period these treatments were crossed over. The calcium preparations used caused equal reductions in frequency of bowel movements. Both preparations increased serum calcium and serum phosphate, but significance (p < 0.05) was reached only during supplementation with the slow-release preparation. Despite the Intended raising of serum calcium levels, none of the preparations influenced the 24-hour urine oxalate to creatinine ratio or the urine stone index. There are several other reasons for supplying extra calcium after intestinal bypass procedures, but our data do not support the concept of preventing renal stone formation by means of calcium supplementation.