Three different test meals containing 800 mg Ca from different sources labelled with ⁴⁷Ca were tested in 10 subjects using whole body counting three times during a 4-week period after the ingestion of the test meal. The three test meals were tested according to a double-blinded randomised crossover design, where the subjects were randomly assigned to the sequence of the test meals and there was a wash out period of 7 weeks between the three test meals (Figure 1). To maintain the subjects' calcium intake constant at the time of the test meals, a standardised experimental diet was provided for two days before and three days after ingestion of the test meal corresponding to the individual subjects' habitual Ca intake. In the remaining time between the test meals, the subjects consumed a self-selected diet. The study was approved by the Municipal Ethical Committee of Copenhagen and Frederiksberg (KF-01270319), and the National Institute of Radiation Hygiene, Denmark.

Eleven young healthy men (age: 26 ± 4y; BMI: 24 ± 2 kg/m²) were recruited for this experimental study through postings at universities in the Copenhagen area (Table 1). The number of subjects was chosen based on experience from previous studies 1213. The subjects were all non-smokers, none of the subjects exercised >10 h/week, took dietary supplements two months prior to start of the study and during the study, were lactose intolerant, had previously participated in studies using radioisotopes, or suffered from any chronic illnesses of relevance to the study. One subject dropped out during the first experimental period for reasons not related to the study. The subjects were given oral and written information about the study, and written consent was obtained from all subjects. After completion of the three meal tests, the subjects received 3000 DKK (US$ ~500).

To keep the subjects' Ca intake constant, a standardised experimental diet was provided two days before the test meal and three days after. The diet was designed to match the subjects' individual habitual calcium intake, which was assessed using a food frequency questionnaire aimed at calcium intake 14 and had an energy distribution corresponding to the habitual Danish diet (15 E% from protein, 35 E% from fat and 50 E% from carbohydrates). The subjects had the same food items for lunch, dinner and snack meal whereas their breakfasts differed due to the different Ca intakes. The diet consisted of common Danish food items such as rye bread, pasta and rice dishes, fruits, juice and candy. Breakfast meals differed and consisted of breakfast cereals, milk, juice, fruit, breads, cheese and jam. The energy intake was estimated by calculation of the subjects basal metabolic rate (BMR = 0.064*W + 2.84) 15 multiplied by a physical activity level (PAL) factor assessed individually based on a questionnaire on physical activity. All meals were prepared in advance in an experimental kitchen unit especially equipped to avoid trace mineral contamination. One meal was daily served at the department and the remaining foods were provided for home consumption where the hot meal could be microwave heated. The subjects were provided with mineral water (14.2 mg Ca/l, Aqua D'or, Brande, Denmark) ad libitum which they also were allowed to use for making coffee and tea. The subjects were instructed to register the amount of water consumed daily and were not allowed to drink or eat anything which was not provided for them from the department.

In addition to the standardised experimental diet, the subjects were instructed to take a daily vitamin D₃ supplement of 10 μg (Naturdrogeriet, Herning, Denmark) for one month before the first test meal and throughout the study period to ensure an adequate vitamin D status. This was necessary as the study ran from October through March were the sun exposure in Denmark is not sufficient to maintain adequate vitamin D production in the skin.

The three test meals contained 800 mg of Ca from three different sources: enzymatically treated cod bones; enzymatically treated salmon bones, and control (CaCO₃), which were baked into bread rolls. The Ca dose was chosen to reflect a daily dose of a Ca supplement. In addition to the Ca containing bread roll, the test meals consisted of butter and raspberry jam (to camouflage the crunchy fish bones) and was served with 400 ml of ultra pure water. The test persons were monitored during the meal and were given 10-15 minutes to eat. They were instructed to alternate between eating and drinking and to finish their plate. The meal was served at the Department of Human Nutrition, University of Copenhagen on the morning of the test day after a 12 hour fast and the subjects were instructed not to eat or drink for four hours after intake of the test meal.

The isotopes (⁴⁷Ca) were provided by Risø National Laboratory, Radiation Research Department (Roskilde, Denmark) through the Copenhagen University Hospital. The Copenhagen University Hospital received the isotopes the day before dosage of the test meals, prepared doses of 0.2 MBq/ml and transported these to Department of Human Nutrition, University of Copenhagen. Due to problems with the reactor where the isotope was produced, a lower dose was provided in the last period (0.108 MBq ⁴⁷Ca/ml), but this was corrected for in the calculation of calcium absorption. The doses were used to extrinsically label the test meals by drop-wise adding 1 ml of isotope solution (0.2 MBq ⁴⁷Ca) onto the cut surfaces of the rolls. The bread rolls were labelled in the afternoon on the day before intake of the test meal and stored at 5°C until consumption. The external tag method has previously been evaluated 9.

Since each subject received three labelled meals, a maximum dose of 0.6 MBq was received during the study period, which resulted in an estimated maximal radiation dose of 1.5 mSV.

The whole-body counter, located at the National University Hospital in Copenhagen, consists of a lead-lined steel chamber with four plastic scintillator blocks (NE110, Nuclear Enterprises Limited, Edinburgh) connected to conventional nuclear electronic modules and a multichannel analyzer system. The counting efficiency and the settings of the energy window were established though measurements of three water-filled phantoms each containing ⁴⁷Ca activity corresponding to a meal and with weights 55, 66, and 77 kg. The individual measurements lasted for 10 min. To minimize contamination by atmospheric background activity, the subjects had a shower and hair wash and were dressed in hospital clothing before each measurement.

Four different Ca supplements were obtained from a local pharmacy and health store, analysed and evaluated as source for Ca control. The source of Ca in the products differed; being apatite, oyster shell, algae (Lithothamnion calcareum), and CaCO₃. We decided on using CaCO₃ as control in the present study.

Four different fish bone sources were analysed and evaluated as Ca source for the experimental diets. The sources were salmon and cod bones where soft tissue was removed by use of two different methods, boiling or by use of industrialised produced enzymes. The fish bones rinsed by use of enzymes were chosen as Ca source in the present study. The cod bone was provided by Maritex (Sortland, Norway), and the bones had been rinsed enzymatically by use of Papain (6000 NFPU/mg, Biochem Europe, Mons, Belgium). Bones from fresh salmon (Salmo salar) were provided by BioMega A/S, Sotra, Norway. The frames were treated at BioMega A/S with industrially produced enzymes (proteases) 2, similar to protamex (Novozymes A/S, Bagsværd, Denmark). The salmon bones were then rinsed for soft tissue in running cold water using forceps. The cleaned bone samples were freeze dried (Hetosicc, Heto, Denmark) for 74 hours and thereafter ground (Knifetec mill, Foss Tecator). A detailed example of the enzymatic process is described elsewhere 1. CaCO₃ produced for human consumption (Merck KGaA, Darmstadt, Germany) served as a control source of Ca.

Representative samples of the Ca supplements and fish bone were analysed in duplicate. The total fat content of the samples was determined gravimetrically after extraction with ethyl acetate 16. The ash concentration was likewise determined gravimetrically by burning all organic substances in a programmable furnace where the temperature was gradually increased from ambient to 550°C and held at this temperature for 20 hours until constant weight. Total nitrogen was determined using a nitrogen element analyser (LECO, FP-528 N-analyser, Leco Corporation Svenska AB, Sweden). Protein was calculated as N × 6.25.

Prior to the determination of the elements, subsamples were submitted to microwave-assisted wet digestion using 2.0 mL HNO₃ (ultra pure quality) and 0.5 mL H₂O₂, in an Ethos Pro microwave system (Milestone, Holger Teknologi, Oslo, Norway). Flame Atomic Absorption Spectrometry (Perkin-Elmer 3300 AAS, Norwalk, CT) was used for the determination of the elements Na, K, Mg, Ca and Fe (1718). Quantification was made by means of external calibration. Hallow cathode lamps (HCL) were used for magnesium, calcium and iron, whereas sodium and potassium were run in emission mode. The effect of the nitrate concentration present in the sample solutions on the Ca signal were tested 17 and no suppression effect was found. The concentration of P was determined by electrothermal atomic absorption spectrometry on a Zeeman Atomic Absorption Spectrometer (Perkin Elmer 4110 ZL, Norwalk, CT) equipped with na THGA graphite furnace and an AS 72 autosampler. Palladium and magnesium were used as matrix modifier. An Agilent Quadrupole ICP-MS 7500c instrument (Yokogawa Analytical Systems, Inc., Tokyo, Japan) was used as a specific detector for the essential elements: Cu, I, Mn, Se and Zn and the non-essential elements As, Cd, Hg and Pb Six points calibration solutions (5-100 μl/L) were prepared daily by appropriate dilution of 1000 mg/L stock solution of the elements in question (Spectrascan, Teknolab AS, Drøbak, Norway). A diluted solution of a 1000 mg/L (in 10% HCl) certified rhodium stock solution (Spectroscan) was added online and served as an internal standard. Iodine was determined by ICP-MS after the samples were extracted in tetramethylammonium hydroxide (TMAH) for 3 h at 90 °C (1920). The elemental analyses are all accredited by the Norwegian Metrology and Accreditation Service. The certified reference materials TORT-2, Lobster hepatopancreas (National Research council of Canada) and NIST Oyster Tissue (National Institute for Standards and Technology, Gaithersburg, MD, USA) were used for quality assurance of the determination of all elements analysed.

Total Ca content of duplicate portions of the experimental diets and test meals was analysed after homogenisation and microwave digestion (MES-1000, CEM Corporation, Matthews, North Carolina, USA) with HNO3 65%, suprapur (Merck, Darmstadt, Germany) and H2O2 30%, suprapur (Merck, Darmstadt, Germany) by atomic absorption spectroscopy SpectraAA-200 VARIAN (Varian Techtron Pty. Limited, Victoria, Australia) after dilution with lanthaniumoxide solution (Merck, Darmstadt, Germany) (lanthaniumoxide solution: 0.5% lanthaniumoxide, 1% HNO3 and H2O). Standards were prepared from a 1000 mg/L Ca standard (Tritisol^®, Merck, Darmstadt, Germany) by dilution with lanthaniumoxid solution. A reference diet (Standard Reference Material 1548a, Typical Diet, National Institute of Standards and Technology, Gaithersburg, MD, USA) was analysed in the same run. Analysed Ca content 1.943 ± 0.038 mg/g; certified value: 1.967 ± 0.113 mg/g. The coefficient of variation was 2.0% (n = 4).

Concentration of 25(OH)D was quantified using ¹²⁵I RIA kit (DiaSorin, Stillwater, USA) with an intra-assay of 10% and both cholecalciferol and ergocalciferol metabolites were measured. Intact PTH was quantified with an Immulite intact PTH immunoassay kit (Diagnostic Products Corporation, Los Angeles, Calif.) with an intra-assay precision of 5.8%.

It has long been known that retention of many bone-seeking elements is inadequately described by a single exponential function, and that a sum of multiple exponentials presents a more accurate picture 21. Retention (R) after time (t) can then be described by (1):

For large values of t, this function can be accurately approximated by a power function,

where A is the absorption, and b is the rate of excretion.

This approximation reduces the number of parameters that need to be estimated to A and b, and moreover, since the power function is linear in a log-log plot, these parameters can be estimated from empirical data using linear regression on log-transformed values 2122.

For small values of t, the approximation in (2) is less accurate. It is easy to see that as the exponent b is negative, the retention becomes arbitrarily large as t approaches zero. A more accurate function for small t is also given by Norris et al. 21 as

The parameter y can be estimated empirically; the ICRP's estimate of 0.76 days for Ca (ICRP 1972), is reported to fit well with observed data 23. However, for estimation it is simpler and more accurate to use the linear approximation (2) and samples taken sufficiently long after ingestion. Beck et al. 22 suggest sampling after day 9. In this study, measurements were taken on day 16, 23 and 26.

Whole body retention of ⁴⁷Ca was measured with a whole-body counter at the Copenhagen University Hospital (Rigshospitalet, Copenhagen). The individual initial radionuclide burden of the subjects was measured before the ingestion of the test meal. Thus, each test person was measured 12 times. All measurements of ⁴⁷Ca were corrected for background radiation and radioactive decay back to the time of administration.

All statistical analyses were performed using the Statistical Analysis System software package, version 9.1 (SAS Institute inc., Cary, N.C.). Dependent variables (absorption % and vit D concentration) were controlled for homogeneity of variance investigation and normal distribution by investigation of residual plots, normal probability plots and histograms. None of the variables required transformation prior to analyses. An ANCOVA was used to examine the effect of treatment on absorption %. This was performed in PROC MIXED, where subject was modeled as a random variable and vitamin D concentration at baseline of each period as covariates. Period and treatment × period interaction was included as fixed variables. The effect of period on vitamin D concentration was evaluated by ANCOVA, where subject was modeled as a random variable and period as a fixed variable. Posthoc pairwise comparisons were made using Tukey-Kramers adjustment. All data are presented as LS means ± SEE unless otherwise stated and the statistical significance level is defined as p < 0.05.

The Ca content of the four supplements varied (Table 2). The two products containing apatite and oyster shell were low in Ca, containing 0.04 and 24 g Ca/kg, respectively. The supplements containing algae (336 g Ca/kg) and calcium carbonate (324 g Ca/kg) had a higher Ca concentration. Contrary to the calcium supplements, the fish samples also contained fat and proteins. It should be recorded that the salmon and cod frames treated with enzymes had a higher Ca content and a lower fat content compared to the salmon and cod frames rinsed by boiling. Also, some differences in the concentration of various trace elements between the two differently rinsed bones were found (Table 2). The concentration of the non-essential elements in the salmon and cod frames was low. For the present trial, CaCO₃ was chosen as control, while cod and salmon bones rinsed by use of enzymes were used as Ca source in the experimental groups.

Mean Ca absorption from the three different sources were 21.9 ± 1.7%, 22.5 ± 1.7% and 27.4 ± 1.8% for cod bones, salmon bones, and control (CaCO₃), respectively (Figure 2A). The individual data is given in Table 3. There was a tendency towards an effect of treatment on Ca absorption after adjusting for vitamin D status and period (p = 0.072) as the difference in Ca absorption from control and cod bones was close to significant (p = 0.089). No effect of period was seen on mean Ca absorption after adjusting for treatment and vitamin D status (p = 0.33), although it increased from period I through III (Figure 2B). Mean absorption was 22.2 ± 1.7% in November, 23.4 ± 1.8% in January and 26.1 ± 1.8% in March.

The vitamin D status (25(OH)-D) of the subjects decreased significantly during the study period (P < 0.01) despite a daily supplementation for which reason it is included in the statistical analyses as a covariate. Vitamin D concentrations were 56.3 ± 4.2 nmol/l (November), 47.7 ± 4.3 nmol/l (January) and 42.5 ± 4.2 nmol/l (March), and posthoc analyses showed that vitamin D concentration in November was significantly higher compared to both January (p = 0.025) and March (p < 0.01), whereas the vitamin D status in January and March did not differ (p = 0.20) (Figure 2C).

The levels of parathyroid hormone (PTH) was analysed in the start of the study. The results were within normal range (data not shown).

The average daily energy intake was 14.9 ± 2.1 MJ. The age, body mass index, estimated energy intake, estimated calcium intake and chemical analyses of calcium content of breakfast are given in Table 1.