**2. Results**

A total of 54 WNV-positive cases (6 in 2016, 48 in 2017) were included in the analysis. The age range of WNV-positive horses was 4 months to 18 years (median 5 years). The highest proportion of WNV-positive cases was seen in younger horses (Figure 1), especially those less than 5 years old (*n* = 30, 55%).

**Figure 1.** West Nile virus-positive cases in horses in South Africa by age in years, 2016–2017.

In total, 15% (8/54) of the WNV-positive cases were co-infected with another virus (four with MIDV and four with equine encephalosis virus, EEV). Of the 120 randomly selected WNV-negative controls, 14 (12%) were MIDV-positive, 15 (12%) were EEV-positive, and 91 (76%) tested negative for all viruses on the ZARV programme testing panel. During the telephonic follow-up, it was determined that most of the controls were not definitively diagnosed. However, various confirmed diagnoses included herpes virus infection, tickborne diseases (such as babesiosis, Karoo tick paralysis, and vestibular syndrome/facial paralysis from heavy auricular infestations), African horse sickness vaccine reactions, vertebral fracture, guttural pouch mycosis, brain tumours, and severe colic.

Neurological signs, with or without fever (rectal temperature >38.5 ◦C), were significantly more prevalent in WNV-positive cases than in WNV-negative controls (Table 2). Most of WNV-positive cases in 2016–2017 (48/54, 89%) displayed some neurological signs, of which 54% (26/48) had only neurological signs without fever. Approximately half of the WNV cases (28/54, 52%) had fever with or without neurological signs, fewer than the control group (76/120, 63%), although the difference was not significant (Table 3).

**Table 2.** Disease outcome of West Nile virus-positive cases and WNV-negative controls, by main syndrome, amongs<sup>t</sup> horses detected by passive surveillance of neurological and febrile cases in South Africa, 2016–2017.


**Table 3.** The most important clinical signs in the 54 West Nile virus-positive cases and 120 WNV-negative controls, amongs<sup>t</sup> horses detected by passive surveillance of neurological and febrile cases in South Africa, 2016–2017.


\* Two-tailed Fisher's exact test; significance set at *p* < 0.05.

All types of paralysis (hindleg, foreleg, paresis, and total paralysis) as well as ataxia and tremors/muscle fasciculations were significantly associated with WNV infection (Table 3). Laminitic stance/sensitivity in the feet was noted in 9% (*n* = 5) of the cases and only one of the controls (*p* = 0.023). Recumbency tended to be more frequent in cases than controls (*p* = 0.091). Icterus, anorexia, and hyperreactivity/hyperaesthesia were not significantly associated with WNV infection.

Fatality proportions were similar for both cases (39%) and controls (36%) (Table 2). In both groups, subjects with only neurological signs had the highest fatality proportions (cases 14/26, 54% vs. controls 27/44, 61%), while those with only fever had the fewest

fatalities. There was a tendency for WNV-positive cases with fever (with or without neurological signs) to be more likely to recover than those without fever (*p* = 0.057). Of the WNV cases that died, a larger proportion showed hindleg paralysis (9/21, 43%) and total paralysis (8/21, 38%) than foreleg paralysis (2/21, 10%) and tremors (2/21, 10%).

Of the 54 WNV-positive cases, 16 were euthanised due to a poor prognosis, with a median survival time of two days (range 0 to 469 days). Retained neurological signs after recovery were marginally more frequent in cases than in controls (Table 3), which were mainly related to ataxia or neurological instability. Three of these horses, an American Saddler and two Thoroughbred horses, were euthanised at 54–469 days after recovery, ranging in ages: 4 months, 6 years, and 18 years old. The 18-year-old Thoroughbred was also diagnosed with a cardiac tumour at euthanasia. In three other Thoroughbred horses (aged 2.5–3 years), performance was significantly affected by the retention of some degree of clinical signs resulting in early retirement from racing. One of them was sold as a pleasure hack even before racing; another was retired soon after attempting racing (following 8 months' recuperation from WNV infection), and the third had a very unsuccessful racing career.

The breeds mostly represented in WNV-positive cases for 2016–2017 were Thoroughbreds (*n* = 26, 48%), Warmbloods (*n* = 9, 17%), and Arabian horses (*n* = 7, 13%), with the mixed (*n* = 3, 6%) and local breeds (Boerperd and Nooitgedachter) (*n* = 2, 4%) having significantly fewer cases than the purebred horses (*p* = 0.009). Neuroinvasive proportions tended to be similar (90–100%) amongs<sup>t</sup> main breeds, but fatalities appeared to be fewer in Warmblood horses (22%); however, due to small numbers of individual breeds, statistical associations could not be made.

Spatial distribution of WNV cases (Figure 2) showed that most of the equine cases in 2016–2017 occurred in Gauteng (*n* = 19, 35%), KwaZulu-Natal (*n* = 14, 26%), and the Northern Cape (*n* = 11, 20%). The fewest cases were seen in the Western Cape (*n* = 5, 9%), Free State (*n* = 3, 6%), and North West provinces (*n* = 2, 4%), with no reported cases in Mpumalanga, Limpopo, and the Eastern Cape. The largest proportion of cases (*n* = 37, 69%) occurred at the 2nd and 3rd altitude quartiles (1057–1466 m).

**Figure 2.** Distribution of West Nile virus-positive cases (red) and WNV negative controls (green) in horses used in the study, by province, 2016–2017. Each marker may represent one or more cases at the same location.

In 2016, all six cases occurred during February–April. In 2017, the cases occurred during January–June, with a single case in early December in the winter rainfall region of the Western Cape. March was the month with the highest number of cases in both years (*n* = 28, 52%) and the greatest proportion of cases occurred during February–April (*n* = 46, 85%) (Figure 3).

**Figure 3.** Distribution of WNV-positive cases in horses in South Africa by month in which the initial signs were displayed, 2016–2017. The yellow bars show the 2016 cases and green bars show the 2017 cases.

Only 1/54 cases (2%) was reported to have been vaccinated against WNV in the 12 months before sample submission in 2016–2017, in comparison to 9/120 controls (8%). The WNV-positive case that had been vaccinated received an initial WNV vaccination 8 months and a booster vaccination 5 months prior to diagnosis, which were both administered by the owner. Of the 149 herd owners interviewed (cases and controls), only nine (6%) had vaccinated their horses in 2016–2017, whereas 30 (20%) had subsequently vaccinated in 2017–2018.

In the final multiple logistic regression model (Table 4), several variables were significantly associated with WNV infection. WNV cases were more likely to occur during March–April than any other time of the year (*p* = 0.007), and at an altitude of 1293-1466 m (*p* = 0.003). The odds of WNV diagnosis were the lowest in mixed and local breeds compared to intermediate and pure breeds, declined with increasing age (*p* = 0.041), and were lower in WNV-vaccinated horses (*p* = 0.047). Equine influenza virus (EIV) vaccination, although not significant (*p* = 0.145), was retained in the model as a confounder, as its removal resulted in substantial changes to the coefficients of the WNV vaccination and age variables. The Hosmer–Lemeshow test indicated the adequate fit of the final model (*p* = 0.094).


**Table 4.** Final logistic regression model of factors associated with WNV-infection in South African horses with acute febrile or neurological disease detected by the Zoonotic Arbo- and Respiratory Virus (ZARV) programme at the Centre for Viral Zoonoses (CVZ), 2016–2017.

> \* Reference level.
