Our subject was an experienced open-water ultra swimmer (53 years old, 110.5 kg body mass, 1.76 m body height, and a BMI of 35.7 kg/m2). The experiment was approved by the ethical committee of the Kanton, St. Gallen, Switzerland. The swimmer has a broad experience in open-water ultra-swimming. He completed in 2009 the 26.4-km swim in Lake Zurich within 11 h and 1 min. In 2010, he crossed the English Channel in a team relay. In July 2011, he crossed the Great Belt between Germany and Denmark, swimming from Germany to Denmark and back to Germany with a total distance of 50 km in water of approximately 16°C within 19 h and 15 min. In 2010 and 2011, he completed in total fourteen 24-h swims in pools where he achieved distances up to 53 km. For this specific cold water swim, he started preparing in January 2012 with open-water swims. The first swim was in January 21 in water of 1.7°C where he achieved 1 km in 32 min. With increasing temperature, he could increase the length of the swim distance, leading to a swim time of 1 km for 28 min in water of 4.2°C and 33 min in water of 4.7°C. In addition to the open-water swims, he trained in a heated (25°C) indoor pool where he completed training units between 4 and 10 km at a mean speed of approximately 2.8 km/h.
Measurements and calculations
Before the start of the swim, anthropometric characteristics such as body mass, body height, the circumferences of the limbs and the thicknesses of skinfolds at the pectoral, mid-axilla, triceps, subscapular, abdominal, suprailiac, front thigh, and medial calf sites were measured. The circumferences of the limbs as well as all skinfold thicknesses were measured on the right side of the body. Based on these data, body mass index, percent body fat, fat mass, and skeletal muscle mass, using anthropometric methods, were calculated. Body mass was measured using a commercial balance (Beurer BF 15, Beurer, Ulm, Germany) with a precision of 0.1 kg. Body height was determined using a stadiometer with a precision of 1.0 cm. The circumferences of the limbs were measured using a nonelastic tape measure (KaWe CE, Kirchner & Wilhelm GmbH + Co. KG, Asperg, Germany) with a precision of 0.1 cm. The circumference of the upper arm was measured at mid-arm; the circumference of the thigh was taken at mid-thigh, and the circumference of the calf was measured at maximum girth. All skinfold data were obtained using a skinfold caliper (GPM-Hautfaltenmessgerät, Siber & Hegner, Zurich, Switzerland) and recorded to the nearest 0.2 mm. The skinfold measurements were taken once for all skinfold sites. The anatomical sites for the measurements of skinfold thicknesses were pectoral (anterior axillary line), mid-axilla (vertical), triceps (in the middle of the upper arm), subscapular (at the angulus inferior scapulae), abdominal (vertical, right to the navel), suprailiac (at the anterior axillary line), front thigh (mid-thigh), and medial calf (maximum girth). The investigator identified the correct anatomical site using an orientation with finger- and handbreadth from prominent anatomical sites, such as a prominent protuberance or insertion of a tendon. The procedure was performed three times, and the mean of the three measurements was used for the analyses. The available time for taking the skinfold measurements was standardized to ensure reliability. According to Becque et al. , readings were performed 4 s after applying the caliper. One trained investigator took all the skinfold measurements, as inter-tester variability is a major source of imprecision in skinfold measurements. Intra- and inter-investigator agreement was assessed from 27 male runners prior to an ultra-marathon, based on measurements taken by two experienced primary care physicians . Intra-class correlation (ICC) within the two investigators was excellent for all anatomical measurement sites and for various summary measurements of skinfold thicknesses. Agreement tended to be higher within than between investigators, but it still reached excellent reliability (ICC > 0.9) for the summary measurements of skinfold thicknesses. ICC for investigator 1 versus investigator 1 and for investigator 2 versus investigator 2 for the single skinfold thicknesses were between 0.98 and 0.99, respectively. For the sum of seven and eight skinfolds, respectively, ICC was 0.99. For the sum of eight skinfolds for investigator 1, bias (i.e., average difference between investigator 1 and investigator 2) was −0.515 mm, and standard deviation of the average difference was 1.492 mm; 95% limits of agreement were between −3.439 and 2.409 mm. Percent body fat was estimated using the anthropometric formula according to Ball et al.  for males with percent bodyfat = 0.465 + 0.180 × (Σ 7SF)−0.0002406 × (Σ 7SF)2 + 0.0661 × (age), where Σ7 SF is the sum of seven skinfold thickness of the pectoralis, axilla, triceps, subscapular, abdomen, suprailiac, and thigh in millimeters and age in years. The predicted residual sum of squares (PRESS) r
was high (0.90), and the PRESS standard error of estimates (SEE) was excellent (2.2% at the mean) for the equation when applied to a sample of 160 men. Fat mass was estimated using the equations from Stewart and Hannan  for male athletes: Fat mass(g) = 331.5 × (abdominal skin − fold thickness) + 356.2 × (thigh skin − fold thickness) + 111.9 × (body mass) – 9, 108. The coefficient of determination was 0.82, and the standard error of the estimate was 1,843 g, which is equivalent to 2.4% for a typical athlete in the sample. Skeletal muscle mass (SMM) was estimated using the formula of Lee et al. with SMM = Ht × (0.00744 × CAG2 + 0.00088 × CTG2 + 0.00441 × CCG2) + 2.4 × sex – 0.048 × age + race + 7.8 where Ht = height, CAG = skinfold-corrected upper arm girth, CTG = skinfold-corrected thigh girth, CCG = skinfold-corrected calf girth, sex = 1 for male; age is in years, and race = 0 for white men and 1 for black men. This equation was validated using magnetic resonance imagining (MRI) to determine the skeletal muscle mass. There was a high correlation between the predicted skeletal muscle mass and the MRI-measured skeletal muscle mass (r
= 0.83, P < 0.0001, SEE = 2.9 kg). The correlation between the measured and the predicted SMM difference and the measured SMM was significant (r
= 0.90, P = 0.009). In order to compare skinfold thicknesses with the reported data from Keatinge et al. , we measured skinfold thickness over the biceps, at the lower corner of the scapula, at the costal margin below the midpoint of the clavicle, and the abdomen 50 mm below and lateral to the umbilicus.
Temperature was continuously measured as body core temperature in the rectum and as body surface temperature at the left forearm and the right calf using Endotherm® (http://www.endotherm.ch). These thermoelectric probes measure temperatures from −40°C to 85°C with a resolution of 0.0625°C and a precision of 0.1°C. The probes were programmed to take one measurement every 12 s (five measurements per minute) and were applied 10 min before the start of the swim in a heated room of 20°C located around 100 m apart from the start of the swim. The probe in the rectum was inserted using a protective container provided by the manufacturer. At the forearm and the calf, the probe was fixed on a neoprene belt of 3 mm in thickness and 8 cm in breadth at the forearm and 12 cm at the calf, and then fixed with a plastic tape. Energy intake during the swim was estimated using the information on the product supplied by the support crew. Energy expenditure was estimated using a stepwise calculation using body mass, mean velocity, and time spent during performance . The completed distance was continuously recorded on the Seewolf using the global positioning system. Water temperature was measured continuously using the onboard thermometer of the Seewolf. Air temperature was provided by the local weather station (http://www.wetter-lindau.de).