Search strategy

A search of CAB Abstracts (1910- present), Medline In-Process and Non-Indexed Citations and Ovid Medline (1946-present) and Global Health CABI was performed in August 2017 using the OVID interface to identify studies that measured impacts of canine rabies control programmes that conducted either canine surgical sterilisation and rabies vaccination or rabies vaccination only.

The searches used combined terms for dogs (dog, dogs, canine, canines, canis), rabies (rabies, rabid) and vaccination or sterilisation (vaccination, vaccine, vaccinate, immunisation/immunization, immunise/immunize, sterilis$, steriliz$, dog population management, animal birth control, neuter$, spay$, spey$, castrat$, ovariohysterectomy, gonadectomy).

Additional studies for inclusion were identified by hand searches of relevant articles, expert referral and a grey literature search of conference proceedings and funding body reports (S1 File contains the protocol with further details). Online proceedings of relevant conferences were searched using the pre-defined search terms. Any initiatives referenced in key papers were followed up, if not already included or excluded. A broad range of individuals/organisations (n = 23) who had worked in dog population management and/or rabies control internationally, either in academia, for a governmental or a non-governmental organisation, were approached via email to provide any unpublished reports or details of any programmes they knew of that might have relevant data. A reminder email was sent after four weeks to non-responders.

Study selection and criteria

Search results were exported into EndNote (Thomson Reuters, Philadelphia, PA, USA) and screened by a single author (AC) to determine if they met pre-determined inclusion and exclusion criteria (Table 1). Duplicates were removed using the EndNote function (n = 3094) and then manually during the screening process for those not discovered by EndNote (n = 466). If more than one study described the same programme then either the study with the most comprehensive presentation was used, or both were included if different impacts were reported. A second author (RD) screened 10% of the studies (after duplicates removed) for concordance. Any studies queried for inclusion were discussed with JS and RD.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Inclusion and exclusion criteria for study selection.

https://doi.org/10.1371/journal.pntd.0008497.t001

Data extraction and synthesis

Pre-determined qualitative and quantitative data were extracted from each included study using a standardised data extraction form designed for this study (S2 File). The main sections were study characteristics (e.g. country, setting, length of study, publication type), intervention details (including activities), outcomes (vaccination coverage and sterilisation coverage) and indicators of impact (as described in Table 1). Contextual factors and programme components that may have influenced the success, or otherwise, of the intervention were also identified and extracted.

Quality assessment

A quality appraisal was completed for each study, using a modified version of the appraisal tool for cross-sectional studies (AXIS) [28]. This was adapted to include only relevant questions and was reduced from 20 questions to 16 (S3 File). Methods for obtaining dog population size estimates, vaccination and/or sterilisation coverage and impacts measured were also evaluated, including potential biases associated with these methods. Risk of bias in individual studies e.g. potential confounders, and across studies was examined.

Reporting

This systematic review is reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [29]. The complete PRISMA checklist can be found in the supporting information (S4 File)

Summary of included studies

Included studies were published between 1988 and 2018 and were predominantly peer-reviewed publications (n = 13); two reports and one conference proceeding were also included (Table 2). The eight V-S studies represented six different programmes. Different aspects of the same programmes in Jaipur [23,33] and Colombo [31,36] were reported in the included studies. All six V-S programmes were carried out in Asia, whereas the V programmes (n = 8) were conducted in Asia, Africa and South America. The length of study ranged from 1.5 to 10 years with a median of 4.5 years.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Study characteristics.

https://doi.org/10.1371/journal.pntd.0008497.t002

Critical appraisal of the 16 studies identified that study design was not consistent and there were variations in outcomes and impacts measured and methods used, intervention details, study length and reporting. Study designs used could broadly be described as quasi-experimental, non-randomised, pre-post intervention or observational, repeated cross sectional. Impacts in the dog and/or human populations were measured or observed before and after an intervention had taken place, or during an ongoing intervention. Two studies [33,39] also compared an intervention area with a control area. Only four studies [23,31,34,39] determined the statistical significance of any of their findings.

Intervention details

Another confounding factor affecting impact measurements both within and across studies was the concurrent use of other components (Table 3). Education and community awareness were common components of both types of intervention [30–32,35,40–43]. V-S programmes were more likely to have provision for free or subsidised veterinary care [30,31,33,35]. In comparison V programmes were more likely to improve access to post-exposure prophylaxis (PEP) for people [39,41–43].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Comparison of intervention details.

https://doi.org/10.1371/journal.pntd.0008497.t003

Outcomes

Table 4 shows a summary of reported programme outcomes and methods used for measurement (Table 4).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Outcomes and methods for measurement.

https://doi.org/10.1371/journal.pntd.0008497.t004

There were differences between studies in terms of methods and reporting of outcomes. Dog population size was often used for subsequent calculations of vaccination coverage. Methods for estimating population size used were extrapolations from human:dog ratios [39–41,43]; government censuses [30,32]; direct dog counts [33,36]; household surveys [35,43]; and mark-resight techniques [34,41]. Subsequent application of different calculations or models were sometimes used. Inaccurate dog population size estimates were acknowledged as a limitation in many of the studies, particularly when secondary sources of data e.g. government censuses or extrapolations using human:dog ratios were used. When multiple methods were used within the same study the wide variations in estimates that could be obtained were highlighted [40,41,43].

Vaccination coverage was reported in the majority of the studies [30,32–35,37–41,43,44]. Again, methods for estimation varied, or were not always reported [37,44]. The consequences of using unreliable dog population estimates to estimate vaccination coverage were demonstrated by Valenzuela et al [43], who found that coverage could be 13.1–47.6%, depending on the population estimation method used.

Vaccination coverage was also not reported in a consistent way across the studies. It varied between an average over the total intervention area, smaller geographical areas such as region, area or village or by differing sub-populations e.g. adults, sub-adults, owned, stray. Some studies included dogs vaccinated outside of the intervention in their final estimate, e.g. coverage estimated in a pre-intervention initial survey [40] or vaccinations performed at other places e.g. private vets [38]. With the exception of Cleaveland et al. [39] who attempted to vaccinate new dogs each year, and Reece et al. [23,33] who vaccinated when they sterilised, it was often unclear if it was largely the same or new dogs that were vaccinated each year. There were also differences in whether vaccination coverage was reported annually or just at the start and end of the study period.

Wide variations in vaccination coverage (13.1–89%) were associated with reductions in dog and/or human rabies, in both types of programme. One V-S programme reported a vaccination coverage of 35%, but discussed that this was likely to be a low estimate as it only included dogs that were also sterilised, whereas male dogs were vaccinated but not sterilised [33]. However interruption of rabies transmission was also reported in a V programme with coverage estimated between 13.1 and 47.6% [43]. This is not surprising as the critical coverage needed to interrupt rabies transmission is estimated to be 20–40%, the 70% recommendation is to take a high population turnover into account, and ensure that coverage doesn’t fall below this critical percentage [2].

Sterilisation coverage was less commonly measured, and was reported as proportions of different sub-populations, which hindered comparison. These were as a percentage of sterilised females (65.7%) and males (5.8%) in the free roaming dog population [33]; percentage of free-roaming dogs >3m sterilised in each study area (61.8–86.5%) [34] and as a proportion of owned (64%) or unowned (>85%) dogs [35]. Other studies often reported numbers of sterilisations conducted only [30,31,36].

Impacts

Table 5 shows a summary of the results of the reported impacts measured in the included studies. Table 6 shows the methods used to measure each of these impacts.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Summary results from included studies.

https://doi.org/10.1371/journal.pntd.0008497.t005

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 6. Impacts measured and methods used.

https://doi.org/10.1371/journal.pntd.0008497.t006

Dog rabies cases (n = 11) were most commonly reported, followed by dog-bite incidents (n = 9) and human rabies cases (n = 9). Least commonly reported impacts were dog population demographics (n = 3), dog health and welfare (n = 3) and indicators of human behaviour change e.g. public attitudes/perception (n = 2).

Impacts could not be compared between studies due to contextual differences in study area, study design and length, differences in methods of measurement and the presence of other components of the intervention. Even within studies, comparing impacts measured at the start and end did not necessarily allow for a comprehensive understanding of intervention effectiveness. Initial changes sometimes reversed over time [30,32,41], reflecting the dynamic nature of both dog populations and control programmes.

Dog-bites, dog rabies and human rabies cases

There was conflicting evidence regarding the impact of vaccination with or without sterilisation on dog bite incidence, with increases (n = 2), decreases (n = 3) and no consistent trend (n = 4) all reported. Cleaveland et al. [39] found a significant decline in bites in a V programme. Reece et al. [23] proposed that sterilisation had an additional impact on dog-bite incidence as a result of its effect on dog bites that were due to maternal aggression. The different components of the interventions may address different motivations for dog bites. Reported changes also need to be evaluated in context, e.g. improved awareness may lead to increased reporting of dog-bites, rather than an actual increase in dog-bites. Hasler et al. [31] found a decreasing trend in dog bites in the household survey, whereas hospital records reported an increase, combining both sources of data demonstrated an actual increase in bite reporting. Focus groups and surveys suggested that this was due to better awareness of PEP and people being more likely to seek treatment.

Recorded numbers of dog and human rabies cases were often low in the included studies, even before the intervention. It was acknowledged in some papers that there were likely to be additional undiagnosed cases that were not reported [42], and references to weak surveillance systems [44]. Capacity for surveillance varied between studies, and this affected whether reported rabies cases were suspected or confirmed.

In at least one study [32] a change in the PEP programme was a large part of the intervention, so this was likely to have led to a reduction in human deaths, regardless of the canine component of the intervention. In another study [31] a well regulated PEP system was already in place before the intervention which meant the number of human rabies deaths was low even before the programme was implemented. This study saw no change in human rabies deaths. The remaining eight studies saw a decline in human rabies cases. Five reported a drop to zero cases by the end of the study [32,33,38,42,43]; a further two dropped to zero during the study but had cases towards the end [30,41].

Dog rabies cases showed a decline in all studies in which this outcome was measured. As with human cases, additional information such as surveillance capacity, is beneficial to assist in interpretation of these results as seen. For example, in Mpolya et al [41] a decrease in cases was reported, but also an increasing trend in samples submitted for testing. Activities to improve surveillance were not always noted in the publications, which is an important exclusion if efforts were made.

Dog population demographics

Within the V-S studies there were conflicting results regarding effects on population size. However studies which reported a decrease used more robust methods for measurement [33,34,36]. Only V-S studies made repeated measurements of population size and were using it as an indicator of impact. The V studies that measured population size were doing so to assist in planning the intervention or for calculating vaccination coverage, and so only did so at one point within the study, or used different methods at different time points in the study.

Dog demography results from V-S studies reported no evident trend in proportion of females pregnant when sterilised (n = 1) [23] or age structure of population (n = 1) [34]. The percentage of lactating females decreased in the one study in which this was measured [36]. Dog health and welfare changes were not commonly measured. Hasler et al. [31] reported a higher welfare for dogs during the intervention, measured through a welfare assessment (n = 1), however this was in comparison to an intervention using culling rather than vaccination only, and was not looking at long term free-roaming dog welfare after the intervention. Other impacts measured were fewer dogs that had died due to disease or disappeared during the last year (n = 1) [35], and significantly higher body condition score plus absence of a visible skin condition (n = 1) [36]. Impacts related to dog population demographics, or the health and welfare of free roaming dogs were not reported in any of the V programmes.

Human behaviour change

Indicators of human behaviour change were not commonly measured (n = 2) and measures used were not consistent between the studies. Impacts reported were a positive difference in social acceptance scores between non-dog owners at the start and end of the study, a decrease in the perception of dog related problems [31], and a large shift towards ownership, with a concurrent reduction in unowned and community dogs [35].

Role of sterilisation

Although the included studies demonstrate examples of successful V-S programmes, evidence was not available to support an additional impact of the sterilisation component on dog or human rabies cases, time taken to see a reduction in these, or sustainability of results. For the related impacts associated with the theoretical benefits, there was generally evidence of positive effects on dog bites [23], population reduction [33,34], population turnover [35,36], increased health and welfare [36], and perception of and behaviour towards free-roaming dogs [31,35].

The V-S approach is often used in an attempt to achieve multiple objectives linked to managing free-roaming dog populations. These may include reducing human-dog conflict and improving the lives of free-roaming dogs. However, it is important to make a distinction between conducting sterilisation in order to achieve these wider objectives, and sterilisation for the purposes of rabies control. The effectiveness of V only programmes in reducing dog and human rabies cases has been demonstrated, both in studies included in this review, as well as additional evidence from various settings [2,45–48]. Therefore, programmes with rabies control objectives should focus on ensuring adequate vaccination, and consider sterilisation as a separate activity, with its own impacts to be assessed. This may help clarify any perceptions about a necessity for sterilisation in situations where rabies control is the priority.

Furthermore, the impact of sterilisation and vaccination on the related impacts is unclear and needs further investigation. Impacts on dog population demographics, dog health and welfare, and human behaviour change were less commonly measured than those measuring changes in dog-bites and dog and human rabies cases, and only in the V-S studies. This is unsurprising given that rabies control was an objective in all the papers, whereas V-S programmes were more likely to have additional objectives as discussed above. A better understanding of how vaccination alone affects these related impacts would be beneficial, both for evaluating any additional effects of sterilisation, and to provide additional insights for planning and implementing V programmes.

Dog population demography is an important factor in rabies control in terms of achieving and maintaining vaccination coverage. Evidence of how dog population demography may change in response to vaccination was not available in the included studies. Studies that have been conducted in relation to rabies vaccination tend to be cross sectional and are often conducted prior to an intervention to help inform planning [49,50]. Longitudinal studies are less common but can provide important insights into demographic changes occurring over time and factors that may regulate these changes [13,51]. A clear effect of vaccination on dog population demographics has not been found, although in one study from Tanzania, within vaccination villages survival was higher in vaccinated dogs than in unvaccinated dogs [52].

Human behaviour change is also important as human mediated movement of dogs is often identified as a source of rabies outbreaks [11,53,54] and participation in interventions is often the key to the success of rabies control programmes [55–57]. Positive changes in perception and ownership of free-roaming dogs after V-S programmes were reported [31,35]. However, human behaviour change is complex, and may be less to do with the sterilisation programme itself and more to do with factors such as presence in the community or trust in the organisation. In the Kwa Zulu-Natal programme, included as a vaccination only study [42], it was reported that sterilisation had initially been used but was discontinued due to expense and a lack of results seen in dog population demographics (data not reported). However a subsequent positive change in community attitudes towards veterinary services was attributed to their presence in communities for the sterilisation project, and led to an increase in vaccination numbers [58].

It was not possible to make an assessment of the effects of sterilisation on sustainability of impacts achieved by rabies control interventions. Included studies varied in duration, follow up, and time-point within the programme for which data was recorded. Within both V and V-S interventions, incursions of rabies were sometimes seen in later years [30,41]. Further studies have also described rabies outbreaks in areas that had previously seen interruption of transmission [40,54]. This is also linked to implementation and ability to scale up an intervention. V-S interventions are likely to be conducted on relatively small spatial scales due to longer implementation time, and expanding them geographically is likely to be challenging [59]. Geographical co-ordination is key as reintroduction of rabies from non-vaccinated areas is a challenge in rabies control [60].

Sustainability is also linked to economics, however a comparison of cost-effectiveness was outside the scope of this review. Mass vaccination has been demonstrated to be a cost-effective intervention [61], and sterilisation is far more resource intensive. Any benefits of sterilisation would have to be sufficient to justify the considerable additional costs. This has not been supported from modelling of the two strategies [14].

Limitations

There are challenges with using systematic review methodology to evaluate complex interventions such as canine rabies control programmes. Difficulties in identifying and synthesising all relevant data are due to the lack of a standardised definition of the intervention, use of defined exclusion and inclusion criteria in study selection [62], and the importance of context and implementation on intervention impact [63].

The review demonstrates a large amount of variation in the implementation of the sterilisation component of V-S interventions. For example, if sterilisation of owned dogs was optional and reliant on owners or caregivers bringing dogs to a site [30] or if accessible dogs were caught and released [33,34]; or if there was a focus on females (and prepubescent males in Jaipur) [31,33]. Whilst these details may vary due to contextual factors and local adaptation may be needed, they are important for enabling an understanding of how impacts are achieved. There was also variation in reporting of the sterilisation component. In a number of the included studies, and outside of peer-reviewed publications, numbers of dogs sterilised is often the only measure of intervention effort and used as a representation for success. This may be due to widely held assumptions that intervention effort (i.e. number of dogs sterilised) is matched by effectiveness.

Improved characterisation of the sterilisation component including reporting of sterilisation coverage in terms of sex-specific proportions of defined sub-populations of dogs (i.e. owned, unowned, free-roaming or confined) would be beneficial in future research to allow an understanding of how impacts may be achieved. In contrast, whilst vaccination programmes may also vary in intervention details e.g. fixed point or house-to-house, they all have a common aim of achieving over 70% rabies vaccination coverage annually. The 70% target is sometimes applied to sterilisation too but there is no basis for this [27]. It may come from catch-neuter-vaccinate-release (CNVR) programmes in which vaccination and sterilisation coverage are the same.

Differences in contextual factors hindered meaningful comparison between the included studies. Geographical factors, local dog population dynamics, attitudes towards the dog population and baseline rabies prevalence are all likely to pose different challenges and affect the impact of control programmes. This is a challenge in all public health interventions, and establishing which components associated with success are a direct result of the intervention, and which are due to the context is important [64]. Different contexts may have different mechanisms at work in terms of how impacts are achieved. Identification of these mechanisms in future studies e.g. if sterilisation leads to a reduction of dog population turnover or reduction in abandonment of dogs, would enable a more comprehensive understanding of not just if sterilisation contributes to rabies control, but also why.

The perception of a programme in terms of ‘success’ may also vary across different locations and cultures, and is, to some extent, dependent on what the perceived problems were with free-roaming dogs at the outset of the intervention. In some settings, rabies control is not the priority for communities or policymakers with regard to free-roaming dogs. In a recent study from Chennai, only 15% of people interviewed cited rabies as a concern regarding free-roaming dogs [65]. In such settings there may not be a desire for vaccination-only campaigns, and the need to understand the benefit of sterilisation for DPM is more relevant. Sterilisation may also be able to act as an entry point for vaccination programmes in these settings [17].

In addition to variability in implementation of the sterilisation component, there were variations in what, if any, methods of population or rabies control had been used prior to the study period, as well as other components used during the study period. Interventions in which additional methods of population control, e.g. culling, were reported were excluded. However, additional components such as education initiatives, community awareness programmes and access to other veterinary care were present in the majority of included studies and had the potential to affect the outcomes measured. This highlights a major limitation with this study, which is that it is an oversimplification to classify interventions as V-S or V only. As well as the presence of other components, two of the studies, which were included as ‘vaccination only’, did refer to sterilisation programmes conducted in the study areas previously [44], or at some point in the intervention [42]. Furthermore, even if interventions were only using vaccination and sterilisation, it is difficult to separate out the specific effects of each component.

Few of the V-S studies used study designs that would allow causation to be determined, or attempted to partition the effects of the sterilisation component on rabies control impacts. This is likely due to many of the studies describing an ongoing programme, rather than being designed to answer a specific research question. In addition, variation in indicators measured, and methods used for measurement, made it difficult to make comparisons between studies regarding reported effects. An estimation of dog population size in the intervention area is particularly important. A range of different methods were used in the included studies, many of which have previously been found to have low validity [66]. The use of standardised indicators for relevant impacts would aid in synthesising evidence across interventions and settings.

Searches were carried out in English, and some studies may have been missed because of this, although several non-English language publications were included in the search results as they had English abstracts. The databases used for the searches were chosen because CAB Abstracts has been shown to have the widest coverage of veterinary literature and Medline is recommended if there is a biomedical aspect to the question [67]. Global Health is the only bibliographic database dedicated to Public Health, and benefits from good coverage of international literature.

Attempts were made to minimise publication bias by including grey literature in the review. We anticipated that many organisations may not necessarily be publishing their data. Despite the large number of organisations working in this field, few data were obtained by this route. Additional studies that were identified often either did not have sufficient methodological detail or did not include repeated measures of impacts. Reports from organisations often measured intervention effort only (i.e. vaccination or sterilisation numbers) or data that were collected remained unanalysed.

Further work

In order to further investigate the question under review, prospective studies which compare the outcomes and impacts used in this review in an area, or comparable areas, using the two types of intervention would be needed. In V-S programmes, attempts should be made to partition the objectives and impacts of the different components, and the mechanisms believed to be at work should be identified. This will encourage deeper thinking about how best to implement each component and how to measure its impact. A clear description of contextual factors and intervention details is also important, as understanding the interaction between all of these factors is key in evaluating complex interventions [68].

Conclusion

It is not possible to assess the impact of sterilisation, in addition to vaccination, to control canine-mediated rabies, based on the literature found in this systematic review. Prospective studies comparing outcomes and impact of the two types of intervention are needed if comparative evidence of effectiveness is to be obtained. In the absence of such evidence, a distinction should be made between the use of sterilisation for wider dog population management objectives rather than for rabies control objectives. There are many other social and ecological arguments for the use of sterilisation, and the impacts of sterilisation programmes warrants further investigation for these reasons. However, from the perspective of rabies control, unless evidence emerges that demonstrates an additional impact of sterilisation, the priority should be on implementation of mass vaccination.