This exploratory analysis identified that areas of relative excess of FIV and FeLV exist in the US. Both the choropleth maps of PMR and the spatial scan test for evidence of high risk clusters identified similar areas of relative excess of one infection over the other. Since it is assumed that both infections share similar risk factors, it would be expected that the occurrence of both infections relative to each other would be more or less uniform throughout the US. However, our spatial analyses show that higher numbers of FIV infections were reported in the southern and eastern US compared to FeLV infections. In contrast, reported FeLV infections were observed to be higher in the western and north-central US compared to FIV infections. These results suggest that the relative excesses of one infection over the other may be the result of different factors affecting these geographical areas. The distinct pattern in the geographical variation of the PMR can be explained in a number of ways relating to the agent, environment and host factors. For example, the dominant viral strain might vary over the study area. Furthermore, vaccination management, level of veterinary care, and thus the age and survival times of cats, may differ from place to place.
Factors that play a role in promoting aggression and bites are known to be most important in the transmission of infection from one cat to another for both FIV and FeLV. These known risk factors include feline population type (pet, stray and feral), cat density, sex, age, neutering status, and access to outdoors
[6, 7, 11]. Previous studies indicated that FeLV infection is age dependent and primarily acquired by “friendly” cats through prolonged close contact between virus shedders and susceptible cats involving mutual grooming, sharing of food and water dishes, and use of common litter areas
. However, other studies have indicated adulthood, outdoor lifestyle, neutering status, and fighting to be associated with FeLV as well
[11, 13, 18]. Thus, it is difficult to discern whether these known risk factors, being unique to one infection or the other, could lead to such geographical variability, and results suggest the existence of an unknown spatial risk factor. Further, previous studies have found differences in sero-prevalence across the US despite controlling for these factors
Identification and segregation have been the most important tools in the control of both infections
. Although a FIV vaccine was introduced in 2002 in the US, its efficacy remains controversial; whereas vaccination has been attributed as a factor associated with the decreasing prevalence of FeLV
. It is possible that the prevalence of vaccination may influence the infection patterns observed in this study. The decision to vaccinate a pet would be dependent on owner compliance and related to their socio-economic status, and these factors would vary geographically.
Previous studies have found that approximately 50% and 80% of FeLV infected cats in multi-cat households are likely to die in the two and three years following diagnosis, respectively
[37, 38]. On the other hand, clinical signs in most FIV infected cats are reflective of secondary diseases, and FIV is not thought to cause severe clinical illness in naturally infected cats until advanced age. In fact, with proper care FIV infected cats can live for many years
. Therefore, one would expect to find more FIV than FeLV survivors when sampling from, on average, older populations. Further, cats testing positive for FeLV are likely to be much younger than those testing for FIV, which also implies that most older cats that are FIV positive are more likely to be pets, and therefore may belong to people of higher socioeconomic status than cats that are young, FeLV positive, and more likely to be owned by shelters or catteries.
Different viral clades or strains of FIV are known to predominate in different geographical regions and could reflect the patterns observed in this study. Although clade-specific information was not available for this study, clade A viruses are common in the western US, whereas clade B viruses predominate in the eastern US
. However, the association between viral clades and pathogenicity is unclear
It is important that limitations be considered when interpreting results from this study. The observed variability in infection could be reflective of diagnostic submissions specifically to IDEXX laboratories. This could lead to admission risk bias, a form of selection bias, as is common with registry or hospital based studies, particularly if preference of diagnostic lab by sample submitters in an area is related to the true prevalence of either FIV or FeLV.
Further, sero-prevalence of co-infections with FIV and FeLV ranging from 0.3% to 1.6% have been reported in North American cats
[11, 13, 19, 42]. However, estimation of the PMR assumes both the infections to be independent of each other. Not accounting for coinfections would lead to biased estimates of the PMR. However, as the proportion of coinfections increases, the PMR converges to 1; this means the bias is towards the null. Thus, the PMR estimate in this study is rather conservative, i.e. less extreme. Similarly, the result of the spatial scan test is believed to be conservative, i.e. any significant results are truly significant.
The scan test used in this study implements circular shaped windows to detect clusters which may pose a problem when the outcome of interest is aggregated in a non-circular fashion. The scan test may, for example, detect a larger circular high risk cluster by including surrounding regions of low risk
. Though other non-circular scan tests have been proposed in the literature, none allow for continuously distributed spatial observations such as PMRs.
For this study, an exploratory approach was applied to compare two similar infections and explore the areas of relative excess rather than derive risk estimates for each area primarily because the underlying population (total number of tested cats in each administrative region) was not known. Such an approach has been reported in the veterinary literature to compare relative excess of one disease to the other
. An advantage of these study designs (e.g. case-case study) is that factors may be identified as more important for one disease than the other.
The evidence of distinct clusters of infection necessitates the need to investigate overall spatial dependence in the occurrence of cases (clustering), and if these are identified, to adjust for their presence when evaluating the association of putative risk factors to these infections. Ignoring clustering may result in biased standard errors and thus can compromise risk factor studies