In total, 18 pregnant, privately owned bitches of different breeds were enrolled. Two bitches (breed: Cavalier King Charles) were excluded from analyses, due to a vaginal tumour and insufficient owner compliance, respectively. The participation of the bitches in this study was declared by informed consent of each owner. The breeds were as follows (n = 16): Beagle, Pomeranian, Pyrenean Mountain Dog, Golden Retriever, two Old English Bulldogs, Boston Terrier, Labrador Retriever, Hovawart, four Cavalier King Charles, Landseer, Dachshund and Staffordshire Bullterrier. The body weight and age ranged from 3 to 64 kg and from 2 to 6 years, respectively. Primiparous (n = 9) as well as multiparous bitches (n = 7) were included in this study. Pregnancy of the bitches was confirmed by ultrasonography between days 20 and 28 after ovulation (n = 9) or when this information was not available after first mating date (n = 7). The estimated day of ovulation was defined as the day when peripheral plasma progesterone level reached 4 ng/ml (Hase et al. 2000). A previously validated temperature logger (DST micro-T®, Star Oddi, Gardabaer, Iceland; Maeder et al. 2012) was inserted into the vagina of each of the bitches on day 56–61 after estimated ovulation or first mating date. The temperature loggers weighed 3.3 g and had a diameter of 8.3 mm and a length of 25.4 mm. The housing material consisted of implantable, biocompatible ceramic material. To prevent movement in the vagina and potential displacement, temperature loggers were attached to a progesterone-free modified Controlled Internal Drug Release device (CIDR-blank, Pfizer, New York, NY, USA) for ewes. The Y-shaped dimensions of the CIDR-blank were 6.0 × 3.5 cm with a diameter of 1.0 cm. A transponder (Backhome Bio Tec-Transponder®, Virbac, Virbac Corporation, Fort Worth, TX, USA) was placed under the silicon lining of the CIDR-blank. The transponder had a length of 1.4 cm and a diameter of 0.2 cm. Using a reading device (Minimax®, Virbac, Virbac Corporation, Fort Worth, TX, USA), the presence and position of the logger could be confirmed. The temperature loggers were programmed (Mercury Application Software®, version 2.05, Star Oddi, Gardabaer, Iceland) to measure temperatures at 10-min intervals. The loggers were inserted via a vaginal speculum. For smaller breed dogs with a body weight below 20 kg, a smaller speculum (Model 63 956, length 11 cm, diameter 1.3 cm, Storz, Tuttlingen, Germany) was used. Larger breed dogs had a larger speculum (Model 63956, length 15 cm, diameter 2 cm, Storz, Tuttlingen, Germany) inserted. The temperature logger was pushed through the speculum using a sterile cotton swab and placed approximately 14–18 cm deep in the vaginal cavity depending on the size of dog. A reapplication of the logger was necessary in case of logger loss due to moistness and softening of the vaginal tissue. All procedures were performed at the dogs' home to reduce stress of transport and unfamiliar surroundings. The dogs were kept in their familiar environment with no restrictions to their individual daily exercise routine. The temperature logger was expelled spontaneously from the vagina before delivery of the first puppy. This logger expulsion was defined as time of onset of parturition. Temperature data were downloaded after parturition (DST communication box, Star Oddi, Gardabaer, Iceland). Rectal temperature measurements were recorded with a digital thermometer (VT 1831®, Microlife USA, Incl, Clearwater, FL, USA) by the owners at their convenience.
Data from the temperature loggers were entered and downloaded (Mercury Application Software®, version 2.05), respectively into Excel spreadsheets (ms office 2003, Microsoft Corporation, Redmond, WA, USA), analysed using spss® for Windows (Version 18.0, SPSS Inc., Munich, Germany) and MedCalc software (version 10.1.3.0, Medcalc, Mariakerke, Belgium).
For further analysis, hourly means were calculated for every bitch independently. The temperature recordings at 10-min intervals that were lower than 36.0°C were considered as artefacts due to logger loss and were not included in the calculation of the hourly mean. To describe temperature before parturition, vaginal temperature was averaged in 24-h intervals (i.e. 1–24, 24–48, 49–72, 73–96 and 97–120 h) and 12-h intervals (i.e. 1–12, 13–24, 25–36, 37–48, 49–60, 61–72 and 73–84 h) preceding the time the bitch lost the logger. The average temperature from these intervals was compared using repeated-measures anova. The post hoc test of Bonferroni was performed.
The difference between vaginal temperature at a particular time of day and vaginal temperature measured 24, 36 and 48 h before was calculated for each of the hourly averages during the last 120-h before delivery. The diagnostic value of a decrease in vaginal temperature to predict parturition within the next 24, 36 and 48 h was tested using receiver-operating characteristics (ROC) analysis as described previously for dairy cows (Burfeind et al. 2011). The continuous variable was the difference in vaginal temperature (24, 36 or 48 h), and the classification variable was the occurrence of delivery within 24, 36 or 48 h. Due to continuous vaginal temperature measurement, calculated differences were available for 24 h/day. Therefore, 24, 36 and 48 positive events, defined as the occurrence of delivery within 24, 36 and 48 h respectively, existed per bitch. Sensitivity was defined as the proportion of positive events (occurrence of delivery within 24, 36 and 48 h) correctly predicted by the test (decrease in vaginal temperature). Specificity was defined as the proportion of negative events (absence of delivery within 24, 36 or 48 h) correctly diagnosed as negative by the test (no decrease in vaginal temperature).