A systematic review and meta-analysis of the seroprevalence of Toxoplasma gondii in cats in mainland China

Toxoplasmosis is caused by Toxoplasma gondii which can infect all warm-blooded animals. As the most common feline definitive host, cats play a vital role in the transmission of T. gondii. However, national estimates of the seroprevalence of T. gondii in cats in mainland China are lacking, and therefore a systematic review and meta-analysis were performed to provide insight into national environmental transmission levels and potential transmission to humans. Studies published up until July 1, 2016, on T. gondii seroprevalence in cats within mainland China were searched for in CNKI, WanFang, CBM, PubMed, Embase and through the reference lists of resulting articles. The seroprevalence with its 95% confidence interval (CI) for each individual study was presented, and then point estimates and their 95% confidence intervals (CIs) of pooled seroprevalence were calculated. Subgroup analyses were performed according to potential risk factors. A total of 38 eligible studies, published between 1995 to 2016, covering fifteen provinces and municipalities, and involving 7,285 cats, were included. The seroprevalence in cats per study ranged from 3.9 to 79.4% with a median of 20.3%. As substantial heterogeneity existed among studies, a random-effects model was used to estimate the pooled seroprevalence. The value of the point estimate seroprevalence was 24.5% (95% CI: 20.1–29.0). Seroprevalence in stray cats was significantly higher than in pet cats (OR = 3.00, 95% CI: 1.60–5.64). The seroprevalence increased significantly with cat age (P = 0.018) with 17.4% (95% CI: 7.6–27.2) in the group of ≤ 1 year old, 19.5% (95% CI: 12.7–26.3) in the group of ≤ 3 year-old and 31.6% (95% CI: 22.9–40.3) in the group of > 3 year-old. The seroprevalence of T. gondii in cats in mainland China was moderate and was associated with cat ownership and age. Due to the increasing prevalence of pet cats in China and the intimate relationship between these cats and humans, this might present a significant exposure risk, particularly for China’s large susceptible population. Therefore, further research is needed into the links between cat ownership and human T. gondii infection and how to reduce T. gondii exposure in humans via cat contacts and the environmental contamination with T. gondii oocysts by cats.


Background
Toxoplasmosis is caused by the obligate, intracellular protozoan Toxoplasma gondii, a widespread zoonotic parasite which can infect all warm-blooded animals [1], and is one of the most common zoonosis in the world [2]. Its wide distribution may be attributed to complex transmission patterns and parasite coevolution with multiple hosts [3]. Felids are the only definitive host and one infected cat can discharge millions of infective oocysts in faeces, although only over a few days after primary infection [4,5]. Intermediate hosts (such as humans, rodents and other animals) can be infected through ingestion of oocysts from the environment (food contaminated with oocysts or direct contact with oocysts excreted in cats faeces), consumption of undercooked meat containing T. gondii tissue cysts [6,7], or congenitally when parasites in a pregnant women infected with T. gondii for the first time spread to the foetus through the placenta often causing abortion, premature birth, stillbirth, malformation and/or neonatal congenital infection [8].
Although T. gondii infections of immunocompetent people are typically considered asymptomatic, infections in immunocompromised individuals, such as those with AIDS or organ transplant recipients, can result in severe consequences. For example, approximately 10% of AIDS patients in the USA and up to 30% in Europe are estimated to die from toxoplasmosis [9]. Moreover, positive correlations between previously assumed asymptomatic T. gondii infections with the incidences of schizophrenia [10], car accident [11], epilepsy [12] and suicide [13] in humans have now been reported. The seroprevalence of toxoplasmosis in psychiatric patients was once reported to be as high as 50% [11]. Globally, in 2010 T. gondii was estimated to have caused 10.28 million foodborne illnesses and 0.83 million Disability Adjusted Life Years (DALYs) [14]. These all highlight the global public health importance of this infection in human populations.
Toxoplasmosis remains a public health problem in mainland China, as there is an increasing number of AIDS patients with an estimate of 650,000 in 2005 increasing to 780, 000 in 2011 [15] and a huge number of women of childbearing age, estimated to be approximately 375.8 million in 2013 [16]. Cats play a major role in the transmission of T. gondii, pet cats may therefore be an important potential source of human toxoplasmosis due to their intimate association with humans, particularly if they are free-roaming and may themselves be exposed to environmental T. gondii parasites. The seroprevalence of T. gondii in pet cat owners (11.86%) is higher than in non-pet cat owners (7.38%) [17] or than in the general population (7.88%) surveyed in 2001-2004 [18], and the seroprevalence in some areas of China was as high as 34% [19]. With the rapid development of the Chinese economy and continuous improvement of living standards in China the number of families which have pet cats is increasing. For example, it was estimated that approximately 100 million cats were considered pets in 2010 in China [20], with a growth rate of 10% over the subsequent years [21].
To the authors' knowledge, there is no study which has addressed the overall seroprevalence of T. gondii infection in cats across mainland China nor the risk factors associated with these infections. Therefore, this systematic review and meta-analysis was performed to determine the seroprevalence of T. gondii in cats in mainland China over the last 20 years and to assess the potential risk factors related to T. gondii seroprevence in cats. The purpose was to provide an increased understanding to aid parasite control, as evidence grows of its importance for human health [22,23], particularly in China with such a large susceptible population.

Methods
The study was conducted according to the PRISMA guideline (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) [24]. The PRISMA checklist was used to ensure inclusion of relevant information in the analysis (see Additional file 1).

Search strategy
A literature search was conducted for publications from January 1, 1995 to July 1, 2016. We aimed to include all published studies in English or Chinese on seroprevalence of T. gondii in cats across mainland China. We identified published studies within the following five bibliographic databases (three in Chinese and two in English): "toxoplasma gondii" and "cat" in Chinese ("gongxingchong/or gongxingti", and "mao", respectively) were used as search terms in the Chinese databases (China National Knowledge Infrastructure (CNKI), WanFang and The Chinese Biomedical Literature Database (CBM)), and "toxoplasma" and "china" and "cats" were MeSH terms in the PubMed online and "toxoplasma" and "china" and "cats" were emtree term-exploded in Embase. We also visually scanned all reference lists from relevant studies in an attempt to locate additional studies that may not have been identified by searching the electronic databases. We did not contact authors of original studies for additional information. No attempt was made to retrieve unpublished studies. Full text articles were downloaded or obtained through library resources.

Inclusion and exclusion criteria
A total of 53 full texts were read for eligibility screening (Fig. 1). Selected manuscripts needed to fulfil the following inclusion criteria: (i) cross-sectional study; (ii) locations within mainland China; (iii) targeted objectives included cats; (iv) serological diagnostic methods were used; (v) exact total and positive numbers were provided; and (vi) a sample size greater than 25 (for statistic calculations [25]). Studies were excluded if they did not fulfil all of these criteria.

Quality of the studies
We evaluated risk of bias among the included studies using a quality assessment checklist. The following items were examined and given a score based on a simple scale system (2 for "yes", 0 for "no", or 1 for "unsure").
Was the research question/objective clearly described and stated? Was the sampling method described in detail? Was the period of study clearly stated? Was the serological test method clearly pointed out? Were the subjects categorized into different subgroups?

Data extraction
For data extraction, the detailed characteristics of each study were extracted using a pre-designed data-collection excel form. Information was recorded as follows: study characteristics (the first author, year of publication, year of study, location); study methodology (survey method in detail, sampling method and the serological test used); characteristics of cats (pet or stray, gender, age category, survey season and region); sample size; the number of the positives and/or seroprevalence of T. gondii; score of each study.

Data analysis
While the inverse variance method is widely used and works for prevalence proportions around 0.5, two problems arise when the proportions get closer to the limits of the 0 and 1 range. The first of these problems is that the confidence interval (CI) does not preclude confidence limits outside the 0-1 range; the second problem is that a study gets a large weighting when the proportion becomes too small or too big [26]. Therefore, we here calculated seroprevalence estimates with the variance stabilising double arcsine transformation by the following formula: t = arcsin (sqrt (r/(n + 1))) + arcsin (sqrt ((r + 1)/(n + 1))), where t = transformed seroprevalence, r = positive numbers and n = sample size; se(t) = sqrt(1/(n + 0.5)), where se = standard error and the back transformation to a proportion is done using: p = (sin(t/2)) 2 [26].

Pooling and heterogeneity analyses
The seroprevalence and its 95% CI for each study were first calculated, and then point estimates and their 95%  CIs of pooled seroprevalence of all included studies were analyzed. Forest plots were used to express the results of each study and the heterogeneity among studies. Summary of seroprevalence estimates were obtained using fixed-effects or random-effects meta-analyses which were determined by the I 2 statistic (inverse variance index), which describes the percentage of variation between studies that is due to heterogeneity rather than chance. I 2 does not inherently depend upon the number of studies considered, with values of 25, 50 and 75% corresponding to low, moderate, and high degrees of heterogeneity, respectively [27]. Potential sources of heterogeneity were investigated further by arranging groups of studies according to potentially relevant characteristics. In this study, subgroup analysis was stratified by group (i.e. stray or pet), gender (male or female), age ( Meta-regression was used to investigate any significant difference between/among subgroups and the value of an odds ratio was calculated.

Bias and sensitivity tests
The across-study bias (publication bias) was examined by funnel plots. In addition, the statistical significance was assessed by the Egger's regression asymmetry test [28] and Begg rank correlation method [29]. The Duval & Tweedie non-parametric 'fill and trim' linear random method was used to test and adjust for publication bias [30]. To test the robustness of a pooled estimate, we evaluated the effect of each study on the pooled seroprevalence by excluding single studies sequentially (i.e. estimated based on 37 studies each time). A study was deemed to have no influence if the pooled estimate without it (i.e. n = 37) was within the 95% confidence limits of the overall seroprevalence (n = 38) [31]. Extracted data were entered into Microsoft Office Excel 2007 and Stata 12.0 was used in all statistical analyses.

Search results and eligible studies
We retrieved 856 published studies through five databases and the reference lists of relevant studies (Fig. 1). A total of 775 records were excluded through an initial screening of the titles and/or abstracts. A further 28 records were excluded when taking duplication into account. The remaining 53 full-text articles were assessed, of which 15 records were further excluded according with our inclusion criterion. A total of 38 studies  were included in this meta-analysis. Table 1 shows the characteristics of the final 38 studies eligible for inclusion, which covered 15 provinces and municipalities. The years of the studies performed and published ranged from 1991 to 2015 and from 1995 to 2016, respectively. The total number of cats was 7,285, with a range of 27 to 589 per study. Serological assays used in eligible studies retrieved only involved four tests including Enzyme Linked Immunosorbent Assay (ELISA, n = 22), Indirect Hemagglutination Test (IHA, n = 10), Modified Agglutination Test (MAT, n = 4), and Test Paper (n = 2). The evaluated scores indicating the quality of selected studies were from 6 to 10.

Pooling and heterogeneity analyses
The seroprevalence estimates of T. gondii in cats are shown in a forest plot (Fig. 2). Toxoplasma gondii seroprevalence of each study varied from 3.9 to 79.4% (median 20.3%) with substantial heterogeneity among studies (χ 2 = 1,192.78, P < 0.001; I 2 = 96.9%, 95% CI: 96.1-97.7). The pooled overall seroprevalence was The pooled estimates by potential risk factors are presented in Table 2. In subgroup analyses, because there was a significantly high level of heterogeneity among studies within most subgroups, all estimates of the pooled seroprevalence for each subgroup were calculated using the random-effects model. Of the 38 studies, 33 provided information on the groups of cats investigated (12 about stray cats and 28 about pet cats), and the pooled seroprevalence was significantly higher in stray cats (40.9%) than in pet cats (16.7%) (P < 0.001; OR = 3.00; 95% CI: 1.60-5.64) (Fig. 3). In the eight studies which presented data from both stray and pet cats, the seroprevelance was also significantly higher in stray cats (35.9%) than in pet cats (13.0%) (P < 0.001; OR = 4.87; 95% CI: 2.10-11.30). The seroprevalence varied from 17.4 to 31.6% among three age groups (Fig. 4) and the difference was also significant (P < 0.05), as seen in Table 2. A total of 13 studies provided estimates about gender, but no significant difference was observed between male and female cats (P = 0.743; OR = 1.07; 95% CI: 0.67-1.65). Similarly, no significant difference was observed among survey seasons (P = 0.911) as detailed in Table 2. On the basis of geographical regions, the lowest seroprevalence (17.4%) was in Western China and the highest (32.3%) was in Central China but with no significant difference among regions (P = 0.469). When stratified according to the main serological test used, no significant difference was found among the three main methods (P = 0.109), as see in Table 2. Two studies applied a different method (i.e. Test paper) to screen T. gondii antigens in cats' serum. After excluding these two studies, the pooled seroprevalence and its 95% CI were 24.9% (95% CI: 20.3-29.0), which was closely aligned with the previous estimates.

Bias
The funnel plots showed no publication bias (see Fig. 5), which was also confirmed from Egger's test (the bias coefficients b = 2.49; 95% CI: -7.25-9.23; t = 1.06, P = 0.294). No theoretical missing study was filled by the Duval and Tweedie non-parametric method (see Additional file 2).

Sensitivity tests
The sensitivity tests indicated that all single-studyomitted estimates lay within the 95% CI of the respective overall seroprevalence (see Additional file 2). This suggested that the pooled seroprevalence was not substantially influenced by any single study. The stability of such results validated the rationality and reliability of our analyses.  [70] and much lower than in Ethiopia (87.72%) [71] and Estonia (60.8%) [72], it was much higher than in the neighbouring country Japan (5.4% in pet cats) [73,74]. Our study shows a moderate seroprevalence of T. gondii in cats in mainland China when compared to the average seroprevalence of 30-40% worldwide [1]. In our research two factors (stray or domestic, and cat age) were significantly associated with T. gondii seroprevalence.
There was high heterogeneity in seroprevalence levels in cats across mainland China among the eligible studies, but no significant publication bias was found at our cutoff level of 0.05 with either Egger's test, or Duval-Tweedie's method. This high heterogeneity index is suggestive of potential variations, which could be due to real characteristics of cats surveyed, geographical regions, surveyed seasons or due to study effects such as diagnostic methods. To trace the source of heterogeneity, cats were first divided into two subgroups, stray cats or pet cats. In stray cats the pooled seroprevalence of T. gondii infection was significantly higher than in pet cats. This is consistent with studies reported in Spain [70], Tehran [75] and Brazil [76]. This higher seroprevalence in stray cats may be associated with their hunting and diet habits, as a stray cat lives outdoors, hunts and potentially feeds on oocyst contaminated scraps and garbage and/or Toxoplasma-infected wild birds and rodents, with more risk of ingestion of the parasite. Although the seroprevalence in pet cats is lower than in strays, nearly 1 in 5 pet cats has been exposed to T. gondii and the number of pet cats is rapidly increasing in China, strongly associated with the rapid Fig. 3 Forest plot of the seroprevalence estimates of T. gondii in cats by stray or pet cats social change of the country [21]. Some practices such as feeding pet cats raw meat may increase the chance of exposure to T. gondii and transmission from them [20].
The seroprevalence in cats increased with cat age, ranging from 17.4% in cats ≤ 1 years of age in comparison with 31.6% in cats > 3 years of age. This agrees with a study in which a significantly higher seroprevalence of T. gondii is observed in an adult cat group compared with the juveniles [77,78]. This is likely to be explained by the positive association between an increase in age, with an increased risk of exposure to T. gondii oocysts over time, and a long lived immune response to this. No significant difference was observed between sexes, again supporting a study which showed that sex was not considered a determining factor for infection with T. gondii in cats [49]. This indicates that there is little or no difference between the cat sexes in both infection risk behaviour and/or immunological susceptibility.
Toxoplasma gondii is widely distributed, especially in warm, moist and low altitude regions [79], and at temperate to tropical temperatures oocysts remain infectious for up to 1.5 years [80]. Thus, it would be predicted that infections in cats may differ among regions or seasons in relation to climate [81]. Indeed, after the data were stratified based on geographical regions, cats in Central China including Hunan, Hubei and Henan provinces, characterized by a subtropical monsoon climate and suitable for the survival and sporulation of oocysts in  the wild, had a higher seroprevalence than in other regions although this was not significant in the overall analyses. This is partly due to a low number of studies from the central region. In terms of seasons in which surveys were conducted, the highest pooled seroprevalence in cats was in Spring and the lowest in Summer, but this was also not significant, again likely due to small sample sizes within the studies (i.e. 176 to 335 cats per subgroup).
Although the serological methods to identify T. gondii infection differed among studies, ELISA, IHA and MAT were the most common and there were no significant differences among these methods in the reported seroprevalences. In testing seroprevalence of T. gondii in cats with ELISA, IHA and LAT (Latex Agglutination Test), the results from these three kits were similar [74]. By using MAT and ELISA in detecting T. gondii in cats, no significant difference was seen between the two methods [82]. All three diagnostic methods were also compared for the routine screening of T. gondii infections and were shown to have good compliance with each other [83]. All of these findings, including our meta-regression analysis here and meta-analyses on the adjusted seroprevalence with both sensitivity and specificity of each test (see Additional file 3; and original data, see Additional file 4), suggest that testing method was unlikely to be a significant source of heterogeneity in this analysis.
There are two main limitations in our meta-analysis. First, as the numbers of eligible studies in subgroups are small, the estimates and the predictive values of the risk factors should be assessed accordingly. Secondly, no information about cats' environment, such as rural or urban areas, has been described, thus making it impossible to assess the effect of this potentially important factor with regard to implementing control. However, this is the first study, to our knowledge, to estimate the overall seroprevalence of T. gondii in cats in mainland China, leading the way for future research in areas and cat groups which might be informative for future control interventions if required. In addition, there was no information on potentially important issues such as: (i) are pet cats allowed to go outside? and (ii) the effect of rural versus urban areas on T. gondii seroprevalence levels. Future studies incorporating the potential differences between urban and rural areas are required if we are to reduce overall infection levels in China.