RESULTS

Changes in the sexually active dog population of the Anuradhapura Municipality from 2021 to 2025

Table 3 summarizes the results of the negative binomial regression analyses for nine demographic categories: total dog count; males; females; owned dogs; community dogs (free-roaming dogs cared for collectively by the community); owned males; owned females; community males; and community females.

Response | IRR_per_year | 95% CI | p-value | Percentage changes | Theta (Standard error) | Resid_dev_df
Total | 1.03 | 0.93–1.14 | 0.55488235 | 3.0 | 42.05 (27.18) | 1.67
Male | 0.903 | 0.816–0.998 | 0.03622964* | -9.7 | 46.42 (32.47) | 1.67
Female | 1.08 | 0.97–1.21 | 0.14899790 | 8.1 | 35.59 (23.07) | 1.68
Owned | 1.15 | 0.98–1.35 | 0.07018171 | 14.7 | 17.91 (11.54) | 1.68
Owned Male | 1.03 | 0.81–1.30 | 0.79099748 | 2.8 | 9.76 (6.57) | 1.68
Owned Female | 1.24 | 1.03–1.49 | 0.01069109* | 24.1 | 14.33 (9.20) | 1.69
Community Dog | 0.963 | 0.88–1.06 | 0.41294902 | -3.7 | 50.74 (33.51) | 1.66
Community Male | 0.953 | 0.81–1.12 | 0.53483703 | -4.7 | 17.76 (11.68) | 1.66
Community Female | 0.969 | 0.90–1.04 | 0.35403199 | -3.1 | 96.65 (69.54) | 1.67

After adjustment for the number of catching days, total dog counts (IRR = 1.03, 95% CI: 0.93–1.14, p = 0.55), owned dogs (IRR = 1.15, 95% CI: 0.98–1.35, p = 0.07), and females (IRR = 1.08, 95% CI: 0.97–1.21, p = 0.15) showed no significant temporal changes (Table 3). Although the point estimates indicated annual increases of 3.0% for total dogs and 8.1% for females, these trends were not statistically significant (p > 0.05).

In contrast, the number of male dogs declined significantly over time (IRR = 0.903, 95% CI: 0.816–0.998, p = 0.036), corresponding to an annual reduction of 9.7%.

Despite slight decreases in point estimates, no significant temporal trends were observed for community dogs (IRR = 0.963, 95% CI: 0.88–1.06, p = 0.413), community males (IRR = 0.953, 95% CI: 0.81–1.12, p = 0.535), or community females (IRR = 0.969, 95% CI: 0.90–1.04, p = 0.354).

Among the owned population, owned females exhibited a significant increasing trend, with an annual increase of 24.1% (IRR = 1.24, 95% CI: 1.03–1.49, p = 0.011). No significant trend was observed for owned males (IRR = 1.03, 95% CI: 0.81–1.30, p = 0.791).

Figures 9A–9I illustrate annual counts and effort-adjusted trends for the nine demographic groups from 2021 to 2025.

Figure 9A: Annual captures and effort-adjusted trends for the total dog population, 2021–2025.

Figure 9B: Annual captures and effort-adjusted trends for the total male dog population, 2021–2025.

Figure 9C: Annual captures and effort-adjusted trends for the total female dog population, 2021–2025.

Figure 9D: Annual captures and effort-adjusted trends for the community dog population, 2021–2025.

Figure 9E: Annual captures and effort-adjusted trends for the male community dog population, 2021–2025.

Figure 9F: Annual captures and effort-adjusted trends for the female community dog population, 2021–2025.

Figure 9G: Annual captures and effort-adjusted trends for the owned dog population, 2021–2025.

Figure 9H: Annual captures and effort-adjusted trends for the owned male dog population, 2021–2025.

Figure 9I: Annual captures and effort-adjusted trends for the owned female dog population, 2021–2025.

For Figures 9A–9I, the left axis shows annual counts as bar charts, whereas the right axis shows the number of catching days as a linear trend. Points represent observed rates, and lines represent fitted values from negative binomial regression models with log(number of days) included as an offset. Colors were kept consistent within categories. Although raw counts appeared to decrease across categories, significant effort-adjusted annual trends were detected only for male dogs and owned females (Table 3).

Changes in sex ratios in the sexually active dog population of the Anuradhapura Municipality from 2021 to 2025

Linear regression was used to evaluate changes in male:female ratios (Table 4). The overall male:female ratio declined significantly by 0.065 units per year (β = −0.065, p = 0.020), decreasing from 0.518 in 2021 to 0.240 in 2025. In contrast, no significant temporal trends were detected for owned dogs or community dogs (both p > 0.14). Mean ratios indicated a persistent female bias in the total, owned, and community populations throughout the study period.

Category | Mean ± Standard deviation | Range | β per year | 95% Confidence interval | T-test | p-value
Total | 0.0364 ± 0.111 | 0.0240–0.518 | –0.065 | –0.112 to –0.019 | –4.51 | 0.020*
Owned | 0.220 ± 0.072 | 0.131–0.302 | –0.034 | –0.091 to 0.022 | –1.93 | 0.149
Community | 0.410 ± 0.076 | 0.329–0.522 | –0.005 | –0.092 to 0.083 | –0.17 | 0.873

Figure 10 illustrates changes in male:female ratios for total, owned, and community dogs.

Figure 10 shows that the overall male:female ratio decreased significantly over time (β = −0.065, 95% CI: −0.112 to −0.019, p = 0.020), indicating an increasing female bias in total captures.

No evidence of linear temporal trends was detected for owned dogs (β = −0.034, 95% CI: −0.090 to 0.022, p = 0.149) or community dogs (β = −0.005, 95% CI: −0.092 to 0.083, p = 0.873).

The mean male:female ratios over the study period were 0.364 ± 0.111 for total dogs, 0.220 ± 0.072 for owned dogs, and 0.410 ± 0.076 for community dogs. All ratios remained below 1.0 throughout the study period, indicating a persistent female bias in the overall population, as well as in owned and community dogs. The significant decline in the total ratio reflected a shift from approximately 1 male per 2.1 females in 2021 to 1 male per 3.6 females in 2025.

All regression models satisfied the assumptions of linear regression, and no evidence of residual autocorrelation was detected.

Figure 10: Male:female ratios for total, owned, and community dogs from 2021 to 2025.

Population trends for different age groups

Age groups of owned and free-roaming dogs presented for vaccination and sterilization were determined primarily by dental aging, supplemented by other visual characteristics. Dogs were categorized into four age groups: <3 months, 3 months–1 year, 1–6 years, and >6 years.

Age-specific capture rates showed no significant annual trends after adjustment for the number of catching days (Table 5; all p > 0.18). Nevertheless, point estimates suggested annual declines in puppies <3 months (IRR = 0.753, corresponding to a 24.7% decrease per year) and senior dogs >6 years (IRR = 0.846, corresponding to a 15.4% decrease per year), whereas older puppies aged 3 months–1 year showed an estimated annual increase of 33.1% (IRR = 1.33). Adults of reproductive age (1–6 years) remained relatively stable, with an estimated annual decline of 6.24% (IRR = 0.938). All models demonstrated adequate fit and showed no evidence of overdispersion or zero-inflation.

Age group | IRR per year | 95% Confidence interval | % Change per year | p-value | Resid dev/ df
<3 months | 0.753 | 0.467–1.21 | –24.7 | 0.311 | 2.21
3 months- 1 year | 1.33 | 0.905–1.94 | +33.1 | 0.185 | 1.80
1 year- 6 months | 0.938 | 0.622–1.39 | –6.24 | 0.752 | 1.78
>6 years | 0.846 | 0.597–1.19 | –15.4 | 0.358 | 1.73

Figure 11: Age-specific trends of dogs in Anuradhapura, Sri Lanka, from 2021 to 2025.

Observed counts decreased sharply across all age groups in 2022, coinciding with a reduction in the number of catching days (Figure 11). After adjustment for effort, negative binomial models showed no significant temporal trends for any age group (Table 5). Predicted trends evaluated at the mean number of catching days indicated a non-significant increase of 33.1% among older puppies aged 3 months–1 year and declines of 24.7% and 15.4% among puppies <3 months and dogs >6 years, respectively. Adults of reproductive age (1–6 years) showed relatively little change. The wide CIs reflected the uncertainty associated with the limited number of annual observations (n = 5).

Influence of the buffer zone on reducing the population within the municipality

Buffer zones have been widely used in wildlife conservation and disease monitoring to create geographical barriers around vulnerable populations. In this study, the buffer zone refers specifically to the area outside the municipal boundary that was intensively covered to prevent the influx of unvaccinated and sexually intact dogs into the city through migration and deliberate dumping.

Figure 12A shows the zone-wise population patterns from 2021 to 2025. Counts appeared to stabilize in Zones 1 and 2 before intensified buffer zone coverage, whereas Zone 3 appeared to stabilize during 2024 and 2025, coinciding with increased buffer zone coverage.

Figure 12A: Observed zone-wise trends for the community dog population from 2021 to 2025.

However, total catch rates did not change significantly after 2023 in any zone (Table 6A). Zone 1 declined by 8.4% (ratio = 0.916, p = 0.461), Zone 2 increased by 6.1% (ratio = 1.061, p = 0.611), and Zone 3 declined by 4.1% (ratio = 0.959, p = 0.685), but all p-values exceeded 0.05.

Zone | Pre rate | Post rate | Ratio | % Change | p-value | 95% Confidence interval
1 | 33.5 | 30.7 | 0.916 | -8.4% | 0.461 | 0.724-1.16
2 | 28.9 | 30.7 | 1.061 | +6.1% | 0.611 | 0.844-1.33
3 | 33.8 | 32.4 | 0.959 | -4.1% | 0.685 | 0.782-1.17

Similarly, sex-stratified before/after analyses showed no significant post-2023 changes (Table 6B). Male counts remained stable across zones (all p > 0.67), as did female counts (all p > 0.53).

Zone | Sex | Pre rate | Post rate | Ratio | % Change | p-value | 95% Confidence interval
1 | Male | 9.95 | 10.00 | 0.964 | –3.6% | 0.835 | 0.682–1.36
2 | Male | 10.20 | 10.80 | 1.080 | +7.6% | 0.676 | 0.762–1.52
3 | Male | 8.75 | 8.59 | 0.978 | –2.2% | 0.890 | 0.715–1.34
1 | Female | 23.50 | 26.80 | 1.100 | +10.0% | 0.530 | 0.817–1.48
2 | Female | 28.30 | 26.60 | 0.924 | –7.6% | 0.601 | 0.685–1.25
3 | Female | 22.80 | 21.20 | 0.913 | –8.7% | 0.531 | 0.686–1.22

However, buffer spillover effects were sex- and zone-specific (Table 6C). Each additional dog sterilized from the buffer reduced Zone 1 male catch by 0.45% (IRR = 0.9955, 95% CI: 0.9913–0.9995, p = 0.038) and Zone 3 female catch by 0.57% (IRR = 0.9943, 95% CI: 0.9912–0.9975, p = 0.0003). No significant spillover effects were detected for Zone 2, Zone 1 females, or Zone 3 males.

Sex | Zone | Coefficient | Incidence rate ratio per buffer dog | p-value | 95% Confidence interval
Male | 1 | –0.00455 | 0.9955 | 0.038* | 0.9913–0.9995
Male | 2 | –0.00241 | 0.9976 | 0.273 | 0.9932–1.002
Male | 3 | –0.00217 | 0.9978 | 0.224 | 0.9943–1.001
Female | 1 | –0.00253 | 0.9975 | 0.124 | 0.9945–1.001
Female | 2 | –0.00168 | 0.9983 | 0.327 | 0.9950–1.002
Female | 3 | –0.00568 | 0.9943 | 0.0003*** | 0.9912–0.9975

Figure 12B: Sex-biased buffer spillover effects by zone.

When scaled to annual sterilization totals, buffer effects were biologically meaningful for the two significant groups (Figure 12C). At the mean annual buffer sterilization rate of 60 dogs, predicted reductions were 23.7% for Zone 1 males (95% CI: 2.9%–40.8%) and 29.2% for Zone 3 females (95% CI: 14.0%–41.1%). At 100 buffer dogs sterilized per year, predicted reductions increased to 36.3% for Zone 1 males (95% CI: 4.9%–59.1%) and 43.4% for Zone 3 females (95% CI: 22.1%–60.5%).

No evidence was found for a delayed cascade effect. Covering the buffer zone during the previous program did not predict the total count in Zone 1 (IRR = 1.38, p = 0.102), and the gap length between buffer and Zone 1 programs was also not significant (p = 0.234). Results were non-significant for both males (buffer presence p = 0.112) and females (buffer presence p = 0.186). Model Akaike information criterion values were higher for lag models (419.8) than for contemporaneous models (398.3), indicating no evidence of delayed displacement.

Percentage of owned dogs vaccinated and sterilized by October 2025

A questionnaire survey conducted at vaccination and sterilization centers within and outside the municipality from late September to mid-October 2025 was used to estimate vaccination and sterilization coverage among owned dogs. The survey collected 1123 responses from 1687 pet owners who visited the clinics, including 737 from 41 locations within the municipality and 386 from 8 locations outside the municipal boundary. The response rate was 66.57%. Of the total respondents, 912 (approximately 81%) owned at least one dog, representing 1679 owned dogs. Table 7 summarizes vaccination and sterilization coverage achieved by the end of 2025 within and outside the municipality.

Figure 12C: Predicted impact of buffer sterilizations on adjacent zones.

Coverage as a percentage | Within the municipality | Outside the municipality
Percentage of sterilized dogs | 55.17928287 | 50.96296296
Sterilized dogs as a percentage of adult animals* | 60.41439477 | 61.42857143
Sterilized female dogs percentage of all females | 73.11608961 | 73.86759582
Sterilized female dogs as a percentage of adult females | 78.04347826 | 86.53061224
Sterilized male dogs as a percentage of all males | 37.64478764 | 30.31914894
Sterilized male dogs as a percentage of adult males | 42.66958425 | 41.9047619
Percentage of dogs vaccinated at least once | 94.22310757 | 83.11111111
Percentage of dogs vaccinated annually | 74.20318725 | 68.88888889
Dog population density (dogs per dog-owning household) | 1.701694915 | 2.096273292

By October 2025, 74.2% of owned dogs within the municipality had received annual rabies vaccination for at least three consecutive years. Owned dogs being vaccinated for the first time and those vaccinated irregularly across years were recorded as separate categories. Annual vaccination coverage outside the municipal boundary was 68.9%, close to the 70% target set by the World Health Organization for establishing herd immunity against rabies [35]. When dogs vaccinated against rabies at least once were considered, vaccination coverage increased to 94.2% within the municipality and 83.1% outside the municipal boundary.

The percentage of sterilized adult owned dogs was 60.4% within the municipality and 61.4% outside the municipality. Sterilization coverage among adult females was 78.0% within the municipal boundary and 86.5% outside it. Adult male sterilization coverage was approximately 42% both within and outside the municipality.

Influence of the number of dogs per household on regular ARV and sterilization

Crude sterilization coverage ranged from 0.0% to 100.0% by location. Negative binomial regression indicated that centers with higher mean numbers of dogs per household significantly vaccinated and sterilized fewer dogs overall (Table 8A).

Outcome | Predictor | β | Standard error | 95% CI | IRR | 95% CI IRR | p-value
Vaccinated dogs | No. of dogs per household | –0.237 | 0.069 | –0.372 to –0.102 | 0.789 | 0.689–0.903 | <0.001*
Sterilized dogs | No. of dogs per household | –0.306 | 0.149 | –0.598 to –0.014 | 0.786 | 0.550–0.986 | 0.040*

Each additional dog per household predicted a 21.1% reduction in total vaccinated dogs (IRR = 0.789, 95% CI: 0.689–0.903, p < 0.001). A similar pattern was observed for sterilization, where each additional dog per household was associated with a 21.4% reduction in the probability of sterilization (IRR = 0.786, 95% CI: 0.550–0.986, p = 0.040), as shown in Figure 13A.

Figure 13A: Association between mean dogs per household and location-level vaccination and sterilization rates.

The aggregate result suggested that locations with larger dog-owning households had poorer outcomes, supporting the hypothesis that higher numbers of dogs per household may reduce annual vaccination and sterilization uptake. However, GLMM using dog-level data from 865 of the 912 dog-owning households, after excluding rows with mismatched data, showed a contrasting pattern. Between-location variance was substantial for both outcomes. For vaccination, MOR = 1.69 and ICC = 0.07, indicating that the median odds of vaccination differed by 69% between high- and low-coverage locations for households with similar numbers of pets. Heterogeneity was greater for sterilization, with MOR = 2.43 and ICC = 0.17, indicating a 2.4-fold median difference in odds between locations. This between-location variance exceeded the household-level effect of dog number, indicating that coverage differences were driven more by location context than by household size.

Figure 13B: Aggregate-level versus household-level inference.

While the aggregate model-predicted an approximate one-quarter reduction for each additional dog per household, individual-level responses showed strong variation among locations, including an almost four-fold gap between the locations with the best and worst sterilization performance.

Adjusted sterilization probabilities ranged from 20.1% (95% CI: 8.8%–39.7%) in Location 7 to 88.6% (95% CI: 71.8%–96.0%) in Location 27. Supplementary Table 1 provides vaccination and sterilization percentages and ranks for each location.

Sterilization showed a wider spread and larger CIs than vaccination, consistent with the higher MOR. Several locations with small sample sizes, such as Location 4 (n = 2 households), had wide CIs spanning 19.6%–83.7%, indicating that their ranks were unreliable despite extreme point estimates. The graphical patterns showed greater between-location heterogeneity for sterilization than for vaccination, as indicated by the broader distribution of points, and a lower mean rate for sterilization than for vaccination, as indicated by the dotted reference line.

Six locations met the priority criteria: adjusted sterilization coverage <35% and an upper 95% CI <50% (Table 8B).

Reported cases of TVT and skin treatments in Anuradhapura Municipality from 2022 to 2025

A total of 286 TVT cases and 2409 skin treatment cases were recorded in the Anuradhapura Municipality across 14 programs between 2022 and 2025, covering 83 mobile clinic days. Table 9A summarizes the annual counts, sampling days, and observed rates of TVT and skin treatment cases from 2022 to 2025. Data for the base year 2021 were unavailable.

Figure 13C: Caterpillar plots of adjusted vaccination and sterilization probabilities by location.

Location no. | No. of households | Adjusted sterilization Confidence interval
2 | 12 | 28.2% (15.3–46)
7 | 8 | 20.1% (8.8–39.7)
15 | 24 | 33.4% (20.4–49.6)
16 | 27 | 32% (20.2–46.6)
42 | 43 | 29.8% (20.8–40.6)
49 | 52 | 23.5% (17.5–30.9)

Figure 13D: High-priority locations for sterilization based on adjusted sterilization probability.

Year | Program | Municipal council days | TVT cases | Skin cases | TVT/7 days | Skin cases/ 7 days
2022 | 5 | 25 | 77 | 571 | 21.6 | 159.9
2023 | 3 | 20 | 73 | 635 | 25.6 | 222.3
2024 | 3 | 18 | 69 | 585 | 26.8 | 227.5
2025 | 3 | 16 | 67 | 618 | 29.3 | 270.4

After adjustment for the number of clinic days per program, TVT cases increased by 9.8% per year; however, this trend was not statistically significant (IRR = 1.098, 95% CI: 0.893–1.355, p = 0.373). In contrast, skin treatments increased significantly by 18.8% per year (IRR = 1.188, 95% CI: 1.067–1.324, p = 0.001). Table 9B summarizes the negative binomial regression results.

Outcome | Incidence rate ratio per year | % Change per year | 95% Confidence interval | p-value | Dispersion
TVT | 1.098 | 9.8 | 0.893–1.355 | 0.373 | 4.69
Skin treatment | 1.188 | 18.8 | 1.067–1.324 | 0.001* | 10.55

Figure 14 shows observed annual rates per 7 clinic days with model-predicted trends and 95% CIs. The predicted TVT rate remained relatively stable across years, whereas the skin treatment rate showed a pronounced upward trajectory.

Figure 14: Annual rates of canine TVT and skin treatments per 7 mobile clinic days, 2022–2025. Points represent observed annual rates calculated as total cases divided by total clinic days × 7. Lines and shaded areas show predicted trends and 95% confidence intervals from negative binomial regression models with log(clinic days) as an offset. Skin treatments increased significantly by 18.8% per year (p = 0.001), whereas TVT increased by 9.8% per year, although the trend was not significant (p = 0.373).

Other indicators of welfare

Deaths due to postoperative complications remained low, with only 60 canine deaths reported during the entire program, all related to aspiration. This accounted for <1% of all surgeries performed, indicating a high level of perioperative welfare. Postoperative edema following surgical castration was reported to the Municipal Veterinary Office for a limited number of male dogs; however, the exact number was unavailable to the authors.

Regional incidence of dog bites, human ARV, and rabies-related human deaths

The national health data system does not maintain records of dog bites, human ARV, or human rabies deaths at the local government level, preventing a comprehensive assessment of the program’s contribution to reducing rabies incidence. However, according to unpublished Ministry of Health data, seven rabies-related human deaths were reported from Anuradhapura District between 2020 and 2023: three in 2020, one in 2021, one in 2022, and two in 2023. No rabies-related human deaths were reported from the district in 2024 or 2025. District-specific statistics for human post-exposure prophylaxis could not be obtained.

Analysis of monetary benefits and costs

Each program treated an average of 35 dogs at a daily operational cost of 52,600 LKR (164 USD). This included drugs and surgical supplies (2000 LKR), daily wages for six assistants (1000 LKR each), meals and accommodation for six assistants and two veterinarians (three meals per day at 400 LKR each and 15,000 LKR for accommodation), and transport costs (20,000 LKR). The calculation excluded veterinary service charges because government veterinarians conducting similar programs would likely receive fixed monthly salaries rather than daily or project-based payments. Therefore, the average per dog cost for vaccination and sterilization was approximately 1500 LKR (4.68 USD). The actual per dog cost may be slightly lower because one to ten cat surgeries were also performed each day.

With the Sri Lankan human population currently estimated at 23.35 million, the national dog population, based on a 1:8 dog-to-human ratio, would be approximately 2.92 million. Considering an annual expenditure of 700 million LKR on human post-exposure prophylaxis alone, the per dog burden of human post-exposure prophylaxis, under a hypothetical scenario in which no dogs were vaccinated or sterilized and dogs were responsible for all bites, would be approximately 240 LKR (0.75 USD).

The Global Alliance for Rabies Control estimates the annual cost of rabies in Sri Lanka at 21,122,199 USD (6,759,103,680 LKR). Accordingly, the per dog cost of rabies would be approximately 2315 LKR (7.23 USD), which exceeds the per dog cost of the intervention. Based on these estimates, the benefit-cost ratio of the intervention is approximately 1.5.