Increasing Incidence of Zoonotic Visceral Leishmaniasis on Crete, Greece

To determine whether the incidence of canine leishmaniasis has increased on Crete, Greece, we fitted infection models to serodiagnostic records of 8,848 dog samples for 1990–2006. Models predicted that seroprevalence has increased 2.4% (95% confidence interval 1.61%–3.51%) per year and that incidence has increased 2.2- to 3.8-fold over this 17-year period.

in human dwellings and rural locations (hollows in olive trees and near rodent burrows) (8). Since 1991, 38 persons who came to hospitals in Heraklion were confirmed as having cases of ZVL. Today, canine infection is confirmed throughout the island (M. Antoniou, unpub. data), and seroprevalences (30%-40%) are some of the highest reported in Europe. These accounts anecdotally suggest that the incidence of ZVL has increased on Crete during the past decade, which, if so, would be relevant to the public and veterinary health sectors.

The Study
We statistically examined diagnostic records of 8,848 dogs sampled in the eastern Crete districts of Lasithi and Heraklion during 1990-2006. Data were supplied by the Faculty of Medicine, University of Crete (n = 1,205 dogs from 97 villages tested by using an indirect immunofluorescent antibody test [IFAT], cutoff titer 160, during 1999-2006, accompanied by demographic and geographic records); and by the Ministry of Agriculture Serology Laboratory, Heraklion (n = 7,643 dogs tested by using an IFAT, cutoff titer 200, during 1999-2005, but without accompanying records). Samples were collected by veterinarians in private practice from any dog initially brought to their clinic for any reason, or by government veterinarians for any dog encountered, irrespective of clinical condition, during prearranged visits to villages. Numbers of villages and dogs sampled in any year depended on available resources at the time.
These combined datasets were then compared with crude prevalence data for 1990-2005 (n = 7,643 dogs) calculated from the ministry records; no difference was detected in prevalence slopes (slope × data source interaction OR 1.06, 95% CI 0.995-1.30, p not significant; Figure). In univariable or multivariable logistic regression that controlled for dog age and clustering on villages (1999-2006, n = 1,205 dogs), no statistical confounding of the probability of a dog being seropositive was attributed to dog use (companion, guard, or hunting dog), sex, crude habitat type, or village altitude (p>0.215; see online Technical Appendix To assess the change in infection incidence, our principal aim, we used 3 standard epidemiologic models (9-11; online Technical Appendix) to calculate infection rates accounting for time, dog age, and potential loss of infection. The first method (model 1) used the cross-sectional age-prevalence data (IFAT cutoff titer 160), in which the proportion of seropositive dogs in each age class is fitted by varying the rates of infection and recovery. A second method (model 2) used these same data to describe the infection rate as it varied with both time and age until reaching the best fit. The third method (model 3) estimated the infection rate from longitudinal data of previously unexposed dogs <12 months of age that were followed up during 1 transmission season.
Results from the 3 models were consistent (Tables 1, 2) and showed a relative increase in the mean infection rate estimated to be 2.20-3.78-fold higher during 2005-2006 than during 1999-2000. The models differed in approach and age of dogs considered by necessity of the model, number of estimated parameters, or model reduction. Inclusion of a parameter describing loss of infection (Table 1,

Conclusions
The potential contribution of any improvements in diagnostic test sensitivity or vigilance to the increasing incidence of ZVL infection is unclear. The difference in cutoff titers between data sources minimally shifted the absolute prevalence values, but not relative prevalence slopes, with time ( Figure). Any loss of infection with age (Table  1, model 1) could result from nonmutually exclusive biologic processes including recovery from infection, death, or reduced past exposure (9,11). The latter possibility is unlikely on Crete because of the higher risk identified in young dogs in all biannual periods. Disproportionate numbers of deaths of seropositive dogs is not suggested by a decline in ZVL clinical signs in older dogs in this study (data not shown) or elsewhere (11,12). Loss of detectable Leishmania-specific antibody is the more likely explanation because the observed rates of serorecovery are not dissimilar to those (e.g., 0.062/month) estimated by cohort studies elsewhere (12).
The continual increase in canine seroprevalence during the latter part of the intervention (Figure)   likely reasons for the low efficacy of dog culling in other leishmaniasis-endemic regions have been described (13).
Officially, destruction of seropositive dogs is still required today unless the owner agrees to veterinary treatment of the dog or to keep the dog under sandfly-proof netting. However, there is no current policy on Crete to combat vectors. We conclude that the results of our study are consistent with a postwar reemergence and current increasing incidence of ZVL infection on Crete.