Since the 1960s, the reported incidence of canine and human rabies has increased in many countries in southern and eastern Africa [1, 2] even though detection rates and reporting systems have deteriorated [3] and effective human post-exposure prophylaxis and dog rabies vaccines have become available commercially [4]. Most human rabies deaths worldwide occur in Africa and Asia, with an estimated 24 500 human deaths per year in Africa, more than 100 times the number of cases officially recorded [5]. In Tanzania, canine rabies has been reported throughout the country and is considered endemic in the Iringa district [6], where 16 persons are reported to have died of rabies at the Iringa Regional Hospital between 1999-2004.

Rabies is an acute viral encephalitis transmitted by contact with saliva of an infected carrier on broken skin [5]. Throughout most of Africa and Asia, domestic dogs are the main reservoir of rabies [7, 8], well able to ensure persistence of the disease [9]. Dogs are also primarily responsible for the transmission of the virus to humans [6, 10–14]. Many studies on rabies dynamics and control in Africa have been focused on wildlife hosts such as bat-eared foxes [15], black-backed [16] and side-striped jackals [17], wild dogs and spotted hyenas [18]; or on domestic dog populations in rural or periurban areas such as in north-western Tanzania [11, 19–22], Zambia [23, 24], Zimbabwe [25, 26], Kenya [27] and Tunisia [28]. However, little is known about the dynamics of canine rabies in urban settings as only few studies have been carried out in a small number of areas. These have been limited to studies on dog demography in N’Djaména, Chad [29] and Maboloko, South Africa [30], on the incidence of rabies in domestic dogs again in N’Djaména [13], and the effectiveness and costs of urban dog mass-vaccinations in Tunis [31] and N’Djaména [32, 33]. Recently it was shown that combined parenteral dog mass-vaccination campaigns with human post-exposure prophylaxis (PEP) was more cost-effective than human PEP alone [34], and parenteral dog vaccination campaigns remain the most promising tool to reduce or even eliminate the incidence of canine rabies and human rabies exposures [11, 20].

Canine rabies has a surprisingly low basic reproductive rate (R₀) of 1.05 to 1.72 worldwide [20]. For rural Tanzania, the corresponding critical vaccination threshold (P_crit) based on R₀ of rabies is 20-40% of the targeted population [20]. Theoretical predictions of vaccination coverage necessary to prevent rabies outbreaks vary between 20-70% [20, 35]. Observed levels of vaccination coverage sufficient to control rabies depend on geographic situation and targeted population. In specific circumstances, a coverage as low as 37.5% can be sufficient [36], but in other, more interconnected populations, higher levels of coverage are necessary. However, the relatively low P_crit reported should be treated with some caution as herd immunity declines rapidly in the interval between vaccination campaigns due to mortality of vaccinated dogs, birth or immigration of susceptible dogs and loss of individual immunity. In order to sustain the reported P_crit vaccination coverage in the interval between campaigns, a higher initial coverage in the order of 70% is advisable [35]. In order to estimate realistic vaccination coverage that disrupts rabies transmission in the long-term, more information on achieved coverage of vaccination campaigns, the dog subpopulations to be targeted for vaccination, patterns of contact among dogs, and between dogs and humans is necessary [37]. As recognised by WHO [38, 39] dog demography studies are key to addressing many of these knowledge gaps. Even more so as rapid changes in human and dog demographics have important implications for the dynamics and control of rabies [20]. Rates of urbanization in Africa are amongst the highest in world [40], and urban communities often are characterised by highly mobile, rapidly increasing and closely-linked human and dog populations. Therefore, gaining better understanding of dog demography and ecology in these expanding urban communities remains a high priority.

Here, we report the findings of a dog-demography study carried out in four urban wards of Iringa Municipality (Tanzania), presenting results of (a) a cross-sectional household survey of 7993 households to assess the proportion of dog owning households as well as the demographics and number of owned dogs, (b) a detailed questionnaire study in a random subset of 360 households to assess dog life-histories and confinement of dogs, (c) a central-point and a house-to-house vaccination campaign to assess the achievable vaccination coverage by different intervention strategies, followed by d) a mark-recapture transect study to estimate the proportion of truly feral dogs in the population, making use of a survey on the durability of marking methods.

Several important findings emerged from this study that have direct relevance for the design of appropriate and effective dog rabies control strategies in urban Iringa. First, the dog population size was almost six times larger than assumed by town records; second, contrary to expectation, only a very small fraction of the observed population were feral dogs; third, survival of pups and sub adults was the major factor driving population growth; and fourth, the high birth and death rates imply a high turn-over rate and rapid decline in vaccination coverage following a single vaccination campaign.

Many of the demographic and ecological characteristics of the Iringa population were consistent with those recorded in other urban African dog populations, suggesting that these results are likely to have widespread relevance for the design of dog vaccination campaigns in urban communities throughout sub-Saharan Africa. The proportion of dog owning households found in the study (14%) was consistent with a larger study carried out in coastal and inland urban areas in Tanzania finding 7.1 to 15.1% of households keeping dogs [45] and similar to a study in a suburb of Lusaka, Zambia where 11% households owned dogs [23], but differed substantially from rural African communities, where dog ownership appears to be much more common and in the order of 16.4-23.9% of households in rural Tanzania [45] or even 53-81% of households in Machakos District, Kenya [46].

The average observed dog:human ratio of 1:10 – 1:25 was also consistent with estimates from urban areas in coastal and inland Tanzania (1:14.4-1:27.2) [45], Maboloko, Bophuthatswana (1:11) [30], in the periurban Kikambuani, Kenya (1:15) [46], in suburban Zimbabwe (1:16) [47], in N’Djaména, the capital of Chad (1:21.5) [29] and in Mutendere, Zambia (1:45) [23]. These contrast with rural areas, where dog:human ratios tend to be much lower (1:6.7 in Palabana, Zambia [23] and 1:8 in Machakos District [46]). However, despite generally low dog:human ratios, dog densities in urban areas are generally much higher than in rural areas. For example, a ten times higher density was found in periurban Kenya (Kikambuani, 110 dogs km^-2) compared to rural areas in Machakos district [46]. The densities recorded in this study (334 dogs km^-2) are broadly consistent with these findings. In countries outside Africa, higher urban dog densities have been reported ranging from 534-936 dogs km^-2 in Mexico [48] and up to 3000 dogs km^-2 in Sri Lanka [24]. This suggests that people remain tolerant of dogs even with increasing numbers and that dog numbers are probably controlled by the availability of human derived food and thus human numbers rather than by the availability of other resources.

The overall Iringa dog population was much larger than veterinary officials in Iringa assumed. The official number of 1240 dogs reported in Iringa Municipality was less than the number of dogs we counted in four out of the 14 urban wards and six times less than our extrapolation of 7619 dogs for the municipality, (based on an extrapolation from the dog:human ratio). This underestimation of the actual dog population has implications for the estimate of vaccination coverage realised in governmental vaccination campaigns and calls for the introduction of improved methods of collecting census data of the dog population.

The approximated proportion of feral dogs in town accounted for less than 1% of the entire dog population, which was surprisingly low given that Iringa had been selected as a study site on accounts of a perceived large feral dog population. These results are consistent with estimates of 1.1 – 10.6% feral dogs in N’Djaména, Chad, [32] but contrast with the 61% ownerless dogs reported in Bangalore, India [49]. This stark difference is likely to arise from the cultural role and perception of dogs, and attitudes towards killing or neglecting unwanted dogs. In Africa, dogs are generally kept with a purpose such as guarding, herding or hunting and have little cultural or religious value [50]. This is in contrast to India, where dogs have a special status in society and ownerless dogs are not only tolerated but often provided with some care from the community [51]. In Iringa, the low feral dog population suggests that food resources are scarce and do not support the development of a viable feral population. A key result of this study is that, in urban Africa, rabies can be controlled by targeting vaccination campaigns at the owned dog population as the low proportion of feral dogs has a negligible effect on the vaccination coverage realised.

The dog population growth rate (10% per annum) exceeded that of the human population (1.5% per annum) [41], and was slightly higher than that recorded in Zimbabwe [26] and Kenya [46] and a lot higher than the population growth of domestic dogs reported from rural areas of Tanzania [20]. The main determinants of population growth were pup and sub-adult survival, and to a lesser degree, early fecundity, whereas a longer lifespan seemed to have less influence. The findings suggest that the dog population is able to grow rapidly despite high sub-adult mortality, as a result of large litter sizes and relatively short generation times. It should be noted that the Leslie Matrix model used in this study is limited in that it takes neither migration nor density effects into account and a cohort study would be necessary to assess immigration and emigration frequencies. Evidence from elsewhere in Africa (N’Djaména, Chad) indicates that emigration of pups may be an important factor in population dynamics (Mindekem personal communication). However, in the Iringa study area, only a small part of pups were sold or given away outside of their birth-ward.

Overall the dog population was young with 52% of the dogs younger than one year. The mean life expectancy of the dogs in the study-area was 2.76y for both sexes after survival of the first year. This corresponds closely to findings of average life expectancies of 2.8y [46] in Machakos and 1.9 y in dogs in the Serengeti [52], but is certainly much lower than in North American dogs with a life expectancy of 4.5y [53]. The 72% mortality found within the first year exceeds the mortality reported in Machakos District [27] and the reported 57% first year mortality in N’Djaména [29].

Adult females produced 0.6 litters per year (or one litter in 1.7y), with similar intervals between litters (1.6y) to those observed in dogs in communal lands and urban areas in Zimbabwe [26, 47]. The average litter size was 5.5 pups, which was slightly larger than the average 4.6 and 4.8 pups per litter recorded in Zimbabwe and the average litter size of 4.7 pups in rural Tanzania [20]. Females started breeding relatively early, with the youngest breeding female 10 months old, and reproduction was recorded up to 11 years. None of the animals encountered during the study was spayed or castrated and the lack of any apparent control of reproduction resulted in high fecundity which more than compensated for the low survival of pups and juveniles and resulted in high population growth.

With a vaccination coverage of almost 78% of owned dogs, the majority of the owned dog population was relatively easily accessible for free vaccination campaigns. This corresponds to similarly high vaccination coverage reached in urban N’Djaména at central point vaccination campaigns [32].

The observed high mortality and even higher fecundity indicate a high turn-over rate in the population, which in turn leads to a rapid decline in the immunisation coverage of the dog population in the interval between vaccination campaigns. In the light of the relatively low critical vaccination threshold for rabies elimination in some areas of Tanzania [20] and the cost effectiveness of dog vaccination campaigns over post-exposure treatment [34] our results call for stringent, regularly repeated mass-vaccination campaigns targeting the owned dog population. Such repeated vaccination campaigns in comparable settings as presented in our study, should have good potential to control rabies as the owned dog population is highly accessible to vaccination and as the size of the feral dog population is negligible.