This study, conducted in an inner-city neighborhood of Vancouver, British Columbia, Canada, during June 2016–January 2017, compared the prevalence of L. interrogans in rat populations before and after a kill-trapping intervention. Each study site (12 total) comprised 3 contiguous city blocks and was designated as a control site or an intervention site (Figure 1). In control sites, no kill-trapping occurred; in intervention sites, kill-trapping occurred only in the central blocks, and the 2 adjacent blocks were designated as nonkill flanking blocks. We divided trapping in each intervention site into 3 time periods: before, during, and after the intervention (Figure 2). Before and after the intervention, rats were trapped, processed, and released. During processing, rats were marked with an ear tag, and morphometric information was recorded (Table 1). Urine was obtained from these rats and tested for L. interrogans by real-time PCR. During the intervention, we euthanized trapped rats; in control sites and flanking blocks, capture-release continued, and rats were not euthanized. The University of British Columbia’s Animal Care Committee (A14-0265) approved all procedures (Technical Appendix).

We used mixed-effects multivariable logistic regression to estimate the effect of the intervention on the odds that rats carried L. interrogans, while controlling for clustering by city block (4). The outcome was the L. interrogans PCR status (negative or positive) of individual rats. The predictor variable categorized rats by block and period of capture: 0, rats caught before the intervention; 1, rats caught after the intervention in control blocks; 2, rats caught after the intervention in nonkill flanking blocks; and 3, rats caught after the intervention in intervention blocks. Although we did not undertake the intervention in control sites, we considered the third 2-week trapping period independently from the other trapping periods in control sites to detect any temporal changes in L. interrogans prevalence not associated with the intervention. We excluded the 7 rats captured both before and after the intervention to avoid double-counting individual rats. For rats recaptured within the same period as their first capture (either before or after the intervention), we averaged weight and length across captures. We also excluded 1 rat missing data for covariates under consideration.

We used a hypothesis-testing model-building approach to estimate the effect of the intervention while controlling for covariates (Table 1). We kept covariates, selected on the basis of their potential to confound the relationship between the intervention and L. interrogans carriage, in the model if they changed the estimated relationship between the predictor and outcome variables by >10%. Because length and weight were collinear, we used the covariate with the largest effect on the relationship between the predictor and outcome. We dichotomized weight around its median because it was not linear with the log-odds of the outcome. For statistical analyses, we used RStudio (Boston, MA, USA).

Of the 438 rats trapped, we included 430 in the modeling process (Table 1). Sixty-four (14.9%; 95% CI 11.7%–18.7%) rats were PCR-positive for L. interrogans. Of 131 rats recaptured, 5 were L. interrogans positive at their first capture and recapture; no positive rats changed pathogen status within a trapping period.

Rats caught in intervention blocks after an intervention had 9.55 times the odds of carrying L. interrogans than did rats trapped before an intervention, while adjusting for weight and wound presence variables (Table 2). We found no significant changes in either flanking blocks or control blocks. In this model, 52.6% of the total model variance was due to the random effect of the block (11). Rerunning the final model including animals that were caught both before and after the intervention did not substantially affect the results (effect of the intervention in intervention blocks; adjusted odds ratio 8.88, 95% CI 1.68–68.08).

This study showed that kill-trapping was associated with increased odds that rats carried L. interrogans in the city blocks where kill-trapping occurred. We did not observe this effect in control blocks or nonkill flanking blocks.

Increased intraspecific transmission of L. interrogans resulting from kill-trapping is a plausible explanation for the observed effect. Previous research suggests that rat-to-rat transmission of L. interrogans is associated with social structures in rat colonies (4). Given that culling is ineffective at removing entire rat populations (7,8,12), kill-trapping may have disrupted social structures and promoted new interactions that facilitated transmission among remaining rats. For example, culling may have removed dominant rats (13), subsequently increasing aggressive interactions among the remaining rats as they established a new social hierarchy. The positive association between L. interrogans status and weight/wound presence (which are correlated with hierarchical dominance) supports this hypothesis because the bacteria may be transmitted through specific aggressive/dominance interactions (4).

We assessed only the effect of culling on a single ratborne pathogen. L. interrogans might be particularly susceptible to the effects of culling because of its dependence on rat social structures. Other vectorborne (e.g., fleaborne Rickettsia spp. [14]) or environmentally acquired (e.g., methicillin-resistant Staphylococcus aureus [15]) rat-associated pathogens might not be as easily influenced by culling. Future studies should determine the duration of effects induced by lethal control because effects on L. interrogans prevalence may wane with time. However, given that such methods are ineffective at removing entire rat populations and might therefore be used repeatedly as the population rebounds (7,8,12), short-term effects may be particularly important.

We demonstrated that rat culling has the potential to increase the odds for L. interrogans carriage among remaining rats and thus could potentially increase the risk for transmission to humans. Although public health risks resulting from such an increase postintervention might be offset by a decrease in the number of rats, we were unable to quantify the size of the rat population before and after intervention. Practical and ethical considerations make it difficult to empirically demonstrate a direct link between culling and increased pathogen transmission from rats to humans. Rather, after culling, the potential for a person to encounter a rat carrying L. interrogans increases if a person encounters a rat, suggesting that the risk for zoonotic transmission increases per rat contact.

The convergence of this study with previous literature documenting that reactive culling is often unsuccessful at removing rat populations (7,8,12) indicates that such methods are ineffective. Instead, ecologically based rodent management, which focuses on reducing resources available to rats (8), should be more widely applied to urban environments.

By integrating our results with other studies on the impacts of culling wild animals to control communicable diseases (5,6), we can conclude that killing animal reservoirs of human pathogens might have unintended consequences on the disease risks. This hypothesis underscores the importance of understanding the ecology of the targeted animal reservoir to design effective control programs.

Additional methods for a study of the effects of culling on Leptospira interrogans carriage by rats, Vancouver, British Columbia, Canada, June 2016–January 2017.