Search criteria included peer-reviewed journal articles, conference proceedings, internal case definitions, and book chapters that investigated methods for the diagnosis of DA toxicosis in CSL and laboratory animals. Additionally, the effects of DA toxicosis in CSL, laboratory animals, and other marine species, as well as the clinical presentations of DA intoxication, were explored. A set time frame for the publication date was not included in the search criteria, and relevant published literature was included regardless of the publication date. Interrogated databases included Google Scholar and PubMed, as well as an internal list of DA-related publications and book chapters from TMMC. Keywords used to search online databases included “domoic acid,” “sea lion,” “zalophus,” “sea otter,” “cardiac,” “in utero,” “neurologic,” “fur seal,” “marine mammal.” and “ophthalmic”.

Inclusion criteria for antemortem diagnosis of DA toxicosis (acute or chronic) in CSLs were seizures, ataxia, head weaving, tremors, blindness, blepharospasm, reduced responsiveness, coma, and/or abnormal behavior that included disorientation, unusually human-oriented or inquisitive, aggressive behavior, or repetitive behaviors (i.e., frantic pacing, swimming in tight circles, rocking on the pen floor, and/or chewing on body parts or objects) (TMMC case definition). Relative eosinophilia (>0.68 × 10⁹/L), elevated hematocrit (median = 49.5%), and/or reduced serum cortisol concentration (<20 mg/dL) on admission blood work provided further indication of DA intoxication (TMMC case definition) (23).

Antemortem differentiation between acute and chronic DA intoxication was based on stranding history, clinical signs, and imaging when available. Acute DA intoxication was presumed to occur when a group of ≥5 sea lions with neurological signs stranded within 48 h and 80 km of each other and, if available, documentation of an active Pseudo-nitzschia bloom or elevated DA levels from monitoring programs such as the Southern California Coastal Ocean Observing System (SCCOOS) and the Central and Northern California Ocean Observing System (CeNCOOS). When available, post-mortem lesions in at least one of the sea lions consistent with acute DA toxicosis in the hippocampus provided further evidence for an acute DA intoxication event (4). Acute DA intoxication was usually inapparent on gross necropsy examination but was confirmed on histology by the presence of neuronal necrosis/sclerosis in the Cornu Ammonis of the hippocampus that may also affect neurons of the dentate gyrus, parahippocampal gyrus, amygdaloid body, olfactory, and pyriform lobe cortex (12, 24).

Inclusion criteria for chronic DA intoxication were defined based on stranding histories in which a single animal stranded in an abnormal location (delta, urban area, etc.), typically in poor body condition, and as an isolated event with neurological abnormalities (4). Antemortem magnetic resonance imaging (MRI) showing hippocampal atrophy provided further evidence of chronic DA intoxication when available. The post-mortem histological diagnosis of chronic DA intoxication was based on the often asymmetric atrophy of the hippocampal complex, characterized histologically by neuronal loss and gliosis (12, 24).

Therapy was aimed at controlling symptoms as previously described in the literature and the TMMC pharmacopeia (5). Phenobarbital (Golden Gate Pharmacy, Novato, CA 94949, USA) was administered for seizure control during rehabilitation and was started upon the initial diagnosis of DA toxicosis (4 mg/kg IM q12h for 2 days followed by 2 mg/kg q12h for 5 more days (IM or PO if eating). Lorazepam (Golden Gate Pharmacy, Novato, CA 94949, USA), midazolam (Almaject, Morristown, NJ 07690, USA), and/or diazepam (Hospira, Lake Forest, Il 60045, USA) were used in conjunction with phenobarbital for the treatment of active, break-through seizures (0.2 mg/kg IM PRN).

Adjunct therapy to help minimize inflammation and cellular necrosis included an antioxidant, alpha lipoic acid (10 mg/kg SQ q24h; Golden Gate Pharmacy, Novato, CA 94949, USA) (25) and an anti-inflammatory dose of dexamethasone (Bimeda-MTC Animal Health Inc., Cambridge, Ontario N3C 2W4, Canada) or prednisone (Novitium Pharma LLC, East Windsor, NJ 08520, USA), unless contraindicated due to co-morbidities such as corneal ulceration. Supportive subcutaneous fluids (lactated Ringer's solution, LRS, Vetivex, Dechra Veterinary Products, Overland Park, Kansas 66211, USA; Normosol-R, PlasmaLyte A, Abbott Animal Health, Abbott Park, Illinois 60064, USA) were administered to anorexic animals at 20–25 ml/kg/day for 3–5 days or until animals started eating.

As DA is retained in fetal amniotic and allantoic fluids, leading to prolonged persistence in circulation with subsequent progressive disease and fetal developmental abnormalities (8, 26), abortion was induced in pregnant females by administering dexamethasone at 0.25 mg/kg IM q24h for 3 days. If abortion did not occur by day 3, a single 500-microgram dose of prostaglandin F2alpha (Lutalyse®, Zoetis, Parsippany-Troy Hills, NJ, 07054, USA) was administered. Prophylactic ceftiofur crystalline free acid (EXCEDE, Zoetis, Parsippany, New Jersey 07054, USA) was administered IM (6.6 mg/kg) to reduce the risk of pyometra in debilitated animals.

A retrospective analysis of the clinical presentation, demographics, and survival to release of 2, 447 CSL diagnosed with DA intoxication and admitted to TMMC between 1998 and 2019 was performed. A subset of 92 CSL affected by DA and admitted to TMMC between July 2015 and August 2017 was analyzed to determine the performance of a quantitative behavioral assessment tool, referred to as the neuroscore (NS), in predicting whether a CSL affected by DA was a candidate for release. The subset of CSL was chosen because they all had at least three complete NS performed with the most up-to-date NS and criteria used consistently by experienced TMMC veterinarians. The NS criteria have been adapted over time based on the changing symptomatology of DA intoxication. Stranding data and clinical presentations, a review of the current diagnostic tests available to differentiate and diagnose acute and chronic DA intoxication, and the performance of the NS in helping to predict whether a CSL is fit for release, as well as its limitations, are reported.

All animals were stranded along the California coast between approximately 34°22′23′′ N and 40°0′4′′ N and were transported to TMMC for assessment and possible rehabilitation. Physical examinations were performed on all animals upon admission under manual or chemical restraint. Age class was estimated based on body mass, straight body length, tooth development, and the development of the sagittal crest in males (27). The state of nutrition was estimated based on a body condition scale of 1 to 5, with 1 being severely malnourished and 5 being over-conditioned (TMMC internal scale). Blood samples were collected via the caudal gluteal vein directly into vacutainer collection tubes, and the complete blood count and serum chemistry were analyzed as previously published (5, 28). Urine was collected when possible via cystocentesis or catheterization, and an abdominal ultrasound examination was performed on a female CSL to assess for pregnancy. An MRI was performed under general anesthesia using published anesthetic protocols (29). Sea lions were housed with one or more conspecifics of similar age, in either dry pens (if non-responsive initially) or pens with access to a saltwater pool and concrete deck space. They were offered frozen, thawed herring two to three times daily, and records indicating whether the animal was eating or not were maintained. Sea lions were diagnosed with domoic acid intoxication using the earlier described inclusion criteria.

A quantitative behavioral assessment tool (NS) aimed at quantifying the severity of DA-associated effects on stranded CSL was developed at TMMC and was performed on CSL diagnosed with DA intoxication as a standardized method to assess the clinical response to treatment. This assessment is similar to the method previously reported (5), although the assessment has been gradually adjusted over the years based on the changing symptomatology (acute vs. chronic) observed in CSL. The NS (Figure 1) was modeled after pain scoring systems in human neonates such as the premature infant pain profile and the neonatal infant pain scale (30, 31) and is a sum of multiple components assessing posture, mentation, the presence of abnormal movements, response to auditory and tactile stimuli, and locomotor skills. The sum of these components is the NS, where a lower NS is consistent with an animal with a higher probability of survival after release (as determined by a low likelihood of re-stranding). The cutoff value used at TMMC to determine whether a CSL was a candidate for release was a NS of 10 or below. The NS assessments began 2 days after an animal finished the previously described course of treatment (i.e., 2 days after the final dose of phenobarbital) to minimize potential residual sedative effects following phenobarbital administration. A single NS assessment was performed every other day by experienced TMMC clinical veterinarians until three scores were obtained, upon which the NS-trend as well as the overall clinical progression of the animal were taken into consideration to determine whether to release the animal.

Differences in the proportion of CSL stranding due to DA between age classes, sex, and outcome (released vs. died/euthanized) and in the proportion of acute vs. chronic cases were tested using Pearson's chi-squared test on the full dataset (CSL stranding between 1998 and 2019). Correlations between variables were tested using generalized linear models and likelihood ratio tests, with model specifics provided in Section 3.

We tested whether it was possible to use different variables to predict the outcome (release vs. death) after stranding due to DA on a subset of 92 CSL strandings between 2015 and 2017. Candidate variables included the state of nutrition, straight body length, presence of circulating eosinophilia, diagnosed co-morbidities at admission, and whether the animal began eating while in rehabilitation, in conjunction with the NS results. We report which combination of variables was the best predictors of releasability to aid rehabilitation facilities in determining whether a DA-affected CSL is fit for release (as determined by the low likelihood of re-stranding). Generalized linear models with logit link functions and binomial error distributions were constructed for different combinations of candidate variables, and their performance was compared using 10-fold cross-validation error, AIC, and AUC (area under the curve) (32). Cross-validation was used to explicitly test the predictive performance of a model. (33) For the 10-fold cross-validation test, the dataset is randomly divided into 10 parts. For each of those 10 parts in turn, a model is fitted using the 9 parts (training data) not included in the current part, after which the fitted model is used to predict the outcome of the data in the current part (test data). This approach repeatedly simulates a situation where a model is constructed using observed data and is then used to predict the outcome of future observations that have not been used to construct the model. We compared mean eosinophil counts of DA-affected CSL between the dataset used in this study and previously reported results (23) using a two-tailed t-test. As those results were reported separately for acute and chronic cases, we first calculated the overall mean and standard error, considering sample size weighting.

Statistical results were reported in terms of the level of statistical support for a tested correlation or difference between models, as opposed to a binary “significant or not significant” classification based on an arbitrary P-value cutoff of 0.05. “No support” is used when P-values are high (~0.1 and above), “weak support” when P-values are close to 0.05 (~0.01–0.1), “moderate support” when P-values are small (~0.001–0.01), and “strong support” when P-values are very small (~<0.001). This reporting method emphasizes the fact that statistical significance is determined on a continuum and follows current reporting recommendations (34).

Analyses, data manipulation, and plotting were performed in R, using packages lme4, ggplot2, pROC, and dplyr (35–39).

A wide variety of tests have been evaluated to investigate and characterize the effects of DA toxicosis in CSL as alternatives to the detection of DA itself in body fluids or tissues, but few have been evaluated for sensitivity or specificity or have been applied to the clinical setting on a regular basis. Changes in circulating numbers of eosinophils, serum apolipoprotein E, and cerebrospinal fluid proteins have been detected in groups of sea lions with DA toxicosis compared to unaffected animals, and DA antibodies have been detected in sea lions with chronic toxicosis (23, 40–42). Antemortem MRI evaluation of hippocampal size (18, 43) and non-invasive diagnostic methods have also been explored. Behavioral studies in CSL affected by DA found a direct correlation between long-term spatial memory and right-sided hippocampal structure (45). Measurement of habituation rates to an auditory stimulus and the presence of persistent abnormal behaviors such as head weaving and muscle fasciculations can also help diagnose DA toxicosis (21, 46). As neurons in the olfactory bulb may also undergo DA-induced necrosis, olfactory testing has also been evaluated for the diagnosis of DA intoxication (47). A summary of these tests, their results, and potential limitations are listed by publication date in Table 1.

Despite the variety of antemortem tests that have shown promising results as diagnostic tools to help characterize the effects of DA intoxication, not all have been applied clinically on a larger scale, either due to logistical reasons, test and reagent development, and/or availability, lack of resources, or low specificity. In addition to diagnostic tools, inexpensive, practical, and minimally invasive methods to aid clinicians in determining whether a CSL affected by DA is clinically fit for release are still needed.

Various clinical presentations previously described in the literature were observed in CSL with DA intoxication, including, but not limited to, head weaving, ataxia, seizures, and coma that varied in severity depending on the animal. Cardiomyopathy, blindness, and the presence of eosinophilia (diagnosed as an absolute eosinophil count >0.68 × 10⁹/L) have also been well-documented in the literature (4, 12, 14, 23). Historically, eosinophilia was reported in both acute and chronically intoxicated CSL in 2009, with a median of 0.925 × 10⁹/L (23). Similarly, we found this trend to be consistent in our full dataset of stranded CSL between 1998 and 2019. Data differentiating between acutely intoxicated CSL and chronically intoxicated CSL were available on 1,640 out of the 2,448 CSL within the full dataset. Of these 1,640 CSLs, eosinophil data were available on 642, where 63.91% (n = 540) had eosinophilia on admit blood work; 70.6% (n = 293) of acutely intoxicated CSL and 61.7% (n = 140) of chronically intoxicated CSL had an eosinophilia. We found a slightly higher mean eosinophil count of 1.07 × 10⁹/L (acute and chronically intoxicated CSL combined) compared to previously published data. We found no statistical support for a difference in previously published eosinophil counts (mean = 1.01 × 10⁹/L) and our current study (mean = 1.07 × 10⁹/L), t-test, t = 1.471, df = 670, p-value = 0.142.

As DA-producing blooms are most common in spring and early summer, it is not uncommon for late pregnancy CSL to strand with DA intoxication (8). To date, clinical treatment has been aimed at saving the adult rather than the fetus, as fetal damage due to in utero exposure is suspected, making the offspring's prognosis poor (8, 11). The number and outcomes of all CSLs diagnosed as pregnant, blind, having eosinophilia on admission blood work, or with cardiomyopathy between 1998 and 2019 are reported in Table 2.

Between 1998 and 2019, TMMC rescued a total of 11, 737 CSL, 21% (n = 2, 447) of which were diagnosed with DA toxicosis (Figure 2). Strong statistical support for an increase in CSL stranding due to DA intoxication over time was found (generalized linear model with log link function and Poisson error distribution; ChiSq = 41.7, p-value < 0.001), where, on average, the number of DA-associated strandings increased by 2.1% each year, which may in part be due to an increase in the total CSL population over time (49). Of the DA-associated strandings, the proportions diagnosed with acute and chronic syndromes were 63.5% (n = 1,042) and 36.5% (n = 598), respectively, with strong statistical support for an annual increase of 7.4% of the average proportion of chronic DA stranding (ChiSq = 42.3, df = 1, p-value < 0.0001). We were unable to categorize a total of 807 cases due to the lack of use of the classification of acute or chronic DA intoxication at admission by the clinician.

Between 1998 and 2019, 44% (n = 1,074) of stranded CSL with DA intoxication (acute or chronic) were released, while 56% (n = 1,366) died or were euthanized. Of the animals that died or were euthanized, 42.4% (n = 579) died naturally, and 57.6% (n = 787) were euthanized. A total of 149 animals with DA intoxication re-stranded (6.1%) after being released, a rate that was significantly higher than the re-strand rate due to causes of stranding excluding DA intoxication during this same period (3.7%; ChiSq = 20.7, df = 1, p-value < 0.0001). This rate considers each re-strand as a separate event, even if the same animal re-strands multiple times.

Strong statistical support was found for age and sex effects (ChiSq = 3,698, df = 4, -value <0.001; ChiSq = 881, df = 1, p-value < 0.001, respectively). Most sea lions with DA toxicosis were adult females, followed by juvenile males, subadult males, subadult females, yearling females, and adult males (Table 2). The average rehabilitation times (number of days between admission and disposition date) were 24.9 days (range = 2–169) for animals that were released (excluding the animals that were placed in permanent human care), 8.3 days (range = 0–98) for those that died or were euthanized, with a combined (survived or died) average of 15.9 days for all outcomes. Demographic statistics are summarized in Table 3.

Records from a subset of 92 DA-intoxicated CSL that were stranded between July 2015 and August 2017 were reviewed to assess the predictive accuracy of the NS in determining whether a DA-affected CSL was fit for release. Other potential candidate variables that aid in determining releasability and components of the NS that were most important in determining whether a DA-intoxicated CSL was fit for release were also assessed. Table 4 lists all models tested along with their results. Of note, the best predictors of releasability were the third NS, whether the animal started eating while in rehabilitation, the difference in NS over time with a decreasing trend of NS values, and the state of nutrition at admission. Utilizing these four variables, the model correctly predicted whether a CSL was released or not 85% (AUC = 0.90) of the time. The presence of eosinophilia, co-morbidities, and straight body length had no significant correlation with the outcome of the animal and thus were not included in the models. Of the 92 animals in the subset, 46.7% died or were euthanized (n = 43), and 53.2% were released (n = 49). Nine animals that died or were euthanized had full histopathology collected on necropsy, and eight of the nine animals had brain and/or cardiac lesions consistent with domoic acid toxicity, ranging from unilateral to bilateral hippocampal atrophy, inflammation, gliosis, and/or neuronal degeneration; myocardial degeneration, necrosis, or edema. The animal that did not have lesions consistent with domoic acid toxicosis was diagnosed with meningitis (22). The remaining animals that died or were euthanized either did not have histopathology collected or did not have full histopathology collected to include the brain or heart.

We also assessed which components of the NS were most important in determining fitness for release. In a situation where all combinations of the different NS subsections were possible (classification tree modeling), auditory approach response, tactile response, and mentation were the most important. When using only one subsection at a time in a logistic regression model, the subsections that correlated most strongly with outcome were auditory approach response, tactile response, and locomotion.

As the survivability of released CSL with DA intoxication is unknown and outside the scope of this study, the re-strand rate was used as an assessment of the effectiveness of the NS in aiding clinicians in determining whether an intoxicated CSL was fit for release. At TMMC, a NS of 10 for the third/final NS performed has been used since 2006 as a cutoff to differentiate between a releasable and non-releasable animal. Prior to the implementation of the NS in the release assessment (1998–2006), the average re-strand rate for DA-intoxicated CSL was 8.2%. In the time frame after the consistent use of the NS during the release assessment (2007–2019), the average re-strand rate was 6%. The differences between the re-strand rates were not statistically supported; however, the consistent implementation of the NS during the release assessment was associated with an improved average re-strand rate for released DA-intoxicated CSL.