Agreement between Repeated Subjective Visual Evaluations of Body Condition of St. Lawrence Beluga Whale (Delphinapterus leucas) Carcasses Using a Visual Analog Scale Open Access

Several populations of beluga whales (Delphinapterus leucas) live in Arctic waters. An isolated and distinct population of beluga is found in the St. Lawrence Estuary, Quebec (de March and Postma 2003). This population was designated endangered in 1983 by the Committee on the Status of Endangered Wildlife in Canada.

As part of a conservation effort to protect this population, causes of mortality have been monitored for the past 35 yr (Lair et al. 2016). Over this period, the pathologists performing the postmortem examination have subjectively evaluated the body condition of the animals. Body condition is an indication of energy storage at the time of death. It gives insight on the chronicity of the illness and allows the examiner to appreciate the significance of animal health parameters associated with some postmortem findings, such as parasitic infections or dental attrition.

The objective of this study was to evaluate the level of agreement between three prosectors on the body condition of beluga carcasses on the basis of repeated visual assessments. Visual evaluation of body condition in belugas relies on three criteria: the general shape of the animal and the visual appearance of the cervical or nuchal region and the epaxial regions (Bradford et al. 2008, 2012; Gryzbek 2013; Joblon et al. 2014). Assessment of the trunk and visibility of the ribs is not practical in belugas, in contrast with other cetacean species (Joblon et al. 2014). Beluga with a round shape and convex cervical and epaxial regions are evaluated as being in excellent body condition, whereas animals with a concave cervical and a more angular shape, as seen with a loss of muscle mass in the epaxial regions, are deemed emaciated.

Standardized photograph sets of beluga carcasses (n=62) were evaluated by three veterinarians involved in the beluga postmortem examinations (S.L., A.S.-B., and S.L.). Each set comprised three full-body photographs: a lateral view, a caudocranial view, and a craniocaudal high-angle shot. The evaluations of body condition were performed using a visual analog scale (VAS), a tool designed to transform subjective evaluations into figures that is most useful when studying a continuum, such as body condition. The VAS was coded following the method proposed by Reips and Funke (2008) and generated and recorded a numeric score from 0 to 999 (Reips and Funke 2008). A training session was organized to familiarize the three observers with the VAS tool. The two most extreme sets of photographs were positioned at each end of the VAS and served for comparison during the scoring process. Each observer performed this evaluation in duplicate with a 14-d interval between both evaluation sessions. Intraobserver and interobserver agreements were assessed using Bland-Altman diagrams and Lin's concordance correlation coefficient (CCC), precision (variability of the scoring), and accuracy (systematic bias). Lin's CCC, precision, and accuracy were qualified as being almost perfect, substantial, moderate, or poor when >0.99, 0.95–0.99, 0.9–0.95, or <0.9, respectively (McBride 2005). Statistical analyses were performed with R version 3.2.3 software (R Development Core Team 2012).

Scores ranged from 0 to 999 with a mean of 593. Intraobserver, interobserver, and total agreements were poor (all CCC<0.9; Table 1). Similarly, intraobserver, interobserver, and total precision were poor (all values <0.9; Table 1). Intraobserver and total accuracies were moderate (0.94 and 0.94, respectively; Table 1). The Bland-Altman diagrams (Fig. 1) showed the differences (y-axis) in relation to the means (x-axis) between both scores of each observer and between averaged scores from different observers. Perfectly correlated variables would have had a difference of zero, regardless of their mean.

The choice of using a VAS rather than a categorical evaluation tool was based on the better possibility of identifying subtle variations with a continuous scale during evaluations (Funke and Reips 2012). A VAS also allows the use of a wider variety of statistical tests (Aitken 1969). The agreement analysis indicated that the repeated scores were correlated, even if the intraobserver, interobserver, and total agreements were qualified as poor. This indicated substantial variability in repeated scoring of individual beluga carcasses. The moderate accuracy suggested that, although the subjective evaluation lacked precision, the variations in scoring were random and did not strongly bias the mean.

The thresholds used to qualify the CCC were designed to assess the agreement between two laboratory methods. It is a very stringent scale that might not be adapted to the current evaluation of subjective scoring of body condition. The graphic presentation of the results thus provided complimentary information.

The Bland-Altman plots showed a similar spindly distribution. The very low and very high mean scores were located closely to the zero-difference line, whereas the variation of the difference was much higher for animals with an intermediate mean score, showing that carcasses with obviously concave cervical and epaxial regions were easily identified as cachectic by the observers and that a high score was given to carcasses with a rounded shape. The discrepancy for carcasses with scores between the two extremes was a high, however. Although this discrepancy could have been a side effect of the use of a VAS, we believe that it truly reflected the inherent variability of subjective visual evaluation. These observations were consistent with the literature. McBain (2001) states that the detection of weight loss is difficult in cetaceans unless it is very severe. It is also difficult for inexperienced observers to see.

Photographs have been used in western gray whales (Eschrichtius robustus), North Atlantic right whales (Eubalaena glacialis), and short-beaked common dolphins (Delphinus delphis) for visual evaluation of body condition (Pettis et al. 2004; Bradford et al. 2008, 2012; Joblon et al. 2014). Although Joblon et al. (2014) reported moderate interobserver agreement, these methods have not been fully validated, and their precision and accuracy remain unpublished.

In beluga carcasses, visual evaluation using a visual analog scale did not consistently capture the continuum of body condition. Nevertheless, different observers agreed on their evaluations for cachectic animals and those in very good body condition, suggesting that trained observers can only provide basic information on body condition and that other methods should be considered when finer details are needed. Different methods could be evaluated and validated in beluga whales, for example, scoring using width and length determined using aerial photographs (Perryman and Lynn 2002; Miller et al. 2012), a scale for postnuchal depression (Gryzbek 2013), body mass index equations from the postmortem weight-length relationship and length–maximum circumference equations (Hart et al. 2013).

We thank the staff of the Réseau québécois d'urgences pour les mammifères marins, Pierre Béland, Carl Guimond, and Richard Plante, for the collection and transport of carcasses, and the veterinarians and veterinary students involved in postmortem examinations, including Daniel Martineau, Sylvain De Guise, Christiane Girard, Igor Mikaelian, and André Dallaire. We thank Lena Measures, Robert Michaud, and Pierre Béland for their long involvement and significant contribution to this program. We thank Julie Arsenault and Christian Bédard for their advice on our work. This project was supported by the Canadian Wildlife Health Cooperative, Fisheries and Oceans Canada, and Parks Canada.