Seroprevalence of Toxoplasma gondii infection in blood donors in mainland China: a systematic review and meta-analysis

Toxoplasma gondii transmitted from blood donors to receiving patients has become a concern as numerous articles about the epidemiology of T. gondii infection in blood donors from different provinces have been published in China. This study aimed to evaluate the seroprevalence of T. gondii infection in Chinese blood donors using a meta-analysis. A total of 40 eligible studies, published from 1986 to 2017 and covering 18 provinces and municipalities were included. Among a total of 49,784 Chinese blood donors, the overall IgG seroprevalence of T. gondii infection was 6.26% (95% CI: 4.62%–8.13%). The highest prevalence was in the Northeast of China and the lowest in Central China. The infection rate increased slowly over the years, but not significantly. A statistically significant correlation was found between the seroprevalence of T. gondii infection and the detection method and educational level (p < 0.01). There was no relationship between age, gender, occupation and blood type and seroprevalence of T. gondii (p > 0.05). The prevalence of antibodies to T. gondii in Chinese blood donors was lower than in other countries, but the risk of transfusion-transmitted toxoplasmosis still exits. More concise methods are still needed to evaluate the possibility of transfusion-transmitted toxoplasmosis from blood donors.


Introduction
Toxoplasmosis, a worldwide disease in humans and most warm-blooded animals, is caused by the opportunistic protozoan Toxoplasma gondii. One-third of the world's population is estimated to be infected by this parasite [25]. T. gondii infects humans mainly through oral, blood and congenital transmission [34]. T. gondii was discovered in 1908 [51] and first in 1955 in China, and the corresponding work was published in 1957 [51]. The first human case of toxoplasmosis in China was reported in 1964 [44]. Over the last few decades, epidemiological surveys have been conducted to monitor the prevalence of T. gondii in China. The prevalence was 5.17% (0.33%~11.79%) in the first national investigation between 1988 and 1992 [52], and then rose to about 7.9% between 2001 and 2004 [56], and 12.3% between 2006 and 2008 [43]. As an opportunistic pathogen, T. gondii rarely causes serious symptoms in healthy humans. However, the prevalence of T. gondii infection is rising and the number of clinical cases in immunocompromised patients [57] is increasing, such as transplant recipients, HIV-positive individuals, and cancer patients, as well as patients with congenital toxoplasmosis and psychosis. More attention should be given to toxoplasmosis as a serious public health problem.
It has been confirmed that T. gondii is a transfusion-transmissible pathogen [29]. In a meta-analysis, the prevalence of T. gondii in blood donors was estimated to be 33% worldwide [8]. In China, researchers paid more attention to T. gondii during the 1980s-1990s than at present. T. gondii screening in blood donors was piloted in many provinces. Based on these screening data, T. gondii infection deferral was added to Blood Donor Healthy Check Guidelines in 2001 [22]. The rule stipulates that toxoplasmosis recovered blood donors should be deferred for six months. Although T. gondii infection rates in Chinese blood donors have increased, rates are still relatively low compared with other countries [8]. In recent years, knowledge of T. gondii was low, and raw meat consumption and exposure to domestic pets has increased. Most blood donors are surprised by questions on T. gondii in questionnaires, and ask what toxoplasmosis is. So far, a comprehensive study on the prevalence of T. gondii has not been performed. Therefore, we conducted a national systematic review and meta-analysis to assess the prevalence of antibodies to T. gondii in Chinese blood donors.

Methods
This study is based on the preferred reporting items for systematic reviews and meta-analyses (PRISMA) checklist [23], used to search for and select studies, and to assess quality and extracted data. This was done by two researchers, independent of each other, to avoid bias.

Search strategy
Firstly, an inductive electronic search using keywords for all potential articles was performed; secondly, a deductive approach was used by searching for and retrieving articles from reference sections of identified publications as well as review articles related to blood donors or donations in China. We searched for epidemiological studies that were conducted before December 2017 in five English-language databases, including PubMed, Springer Link, Science Direct, Web of Science, and Wiley Online Library, and three Chinese databases: Wanfang, China National Knowledge Infrastructure (CNKI), and VIP (WeiPu). The keywords used to search the databases were Toxoplasma, Toxoplasma gondii, T. gondii, Toxoplasmosis, blood donation, blood donors, transfusion, Chinese and China. To maximize outputs, each keyword was searched individually or in combination. Result agreement and discrepancies between results were examined by a third researcher.

Inclusion and exclusion criteria
Selected manuscripts needed to fulfill the following inclusion criteria: (i) cross-sectional study; (ii) locations within mainland China; (iii) targeted objectives were blood donors; (iv) serological diagnostic methods of IgG were used; (v) exact total and positive numbers were provided; and (vi) a sample size greater than 100. Studies were excluded if they did not fulfill all these criteria.

Data extraction
The desired data were recorded using a data extraction form which included title, year of publication, province, sample size, number of seropositive cases, and diagnostic methods by two reviewers (Wang and Lin) using the inclusion criteria. Data on risk factors such as gender, age, education level, occupation and blood groups were also extracted. Discrepancies were resolved by discussion between the two reviewers and by seeking the opinion of the third author (Pan), if necessary.

Meta-analysis
To avoid the confidence interval (CI) being out of the 0-1 range, and to prevent a study from having a large weighting when the proportion becomes too small or too large [2], we calculated seroprevalence estimates with variance stabilizing double arcsine transformation [2]. In addition, if the prevalence obtained from studies was not normally distributed, the prevalence needed to be transformed.
Point estimates and their 95% confidence intervals (CIs) for the prevalence rate of antibodies to toxoplasmosis were calculated for each study. The random effects model was adopted for overall and subgroup analysis if obvious heterogeneity existed, otherwise the fixed effects model was used. Furthermore, both models were adopted to test the difference of the two models for sensitivity analysis. Statistical heterogeneity was evaluated by the Cochran Chi-squared test (with p < 0.10 indicating statistically significant heterogeneity) and the statistic I 2 [6] (heterogeneity with I 2 of 0%-40% was considered not important, while I 2 of 30 to 60% was moderate heterogeneity, I 2 of 50%-90% was substantial heterogeneity, and I 2 of 75%-100% was considerable heterogeneity). A forest plot was used to provide a comprehensive overview of the included studies according to research year. 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 detection methods, regions (Northwest, Southwest, Northeast, South China, Central China, East China and North China), age, gender, occupation, blood group, and education background. Furthermore, meta-regression was used to investigate any significant difference between/among subgroups and the value of IgG seroprevalence. The publication bias was examined by funnel plots. In addition, the statistical significance was assessed by Egger's regression asymmetry test. For meta-analysis, we assumed that the included studies were a random sample from each study population. All analyses were carried out with R software version 3.4.1 (with the package ''meta'' [26] (version 4.8-4) for meta-analysis).

Baseline characteristics of blood donors in included studies based on geographic regions
A total of 3578 blood samples were found to have T. gondii IgG antibodies and the overall prevalence in blood donors was 6.26% (95% CI: 4.62%; 8.13%). The forest plot diagram of the current meta-analysis is presented in Fig. 2.

Seroprevalence of T. gondii infection in blood donors in relation to risk factors
The pooled seroprevalence for each subgroup was calculated using the random-effects model, if there was high heterogeneity. The pooled estimates by potential various risk factors associated with T. gondii infection in blood donors are presented in Table 3. The seroprevalence of T. gondii tested with the ELISA method was 7.30% (95% CI: 5.25%-9.67%), and 3.16% (95% CI: 1.87%-4.76%) with the IHA method, and the difference between the two methods was significant (p = 0.002). The seroprevalence of T. gondii infection for different educational levels was significantly different (p = 0.006), with 4.80% (95% CI: 3.44%-6.37%) in the university-level population, 6.58% (95% CI: 4.79%-8.63%) in the high school population, and 9.01% (95% CI: 6.89%-11.38%) in the middle school population. There was no relationship between age, gender, occupation or blood type and seroprevalence of T. gondii (p > 0.05).
Given the obvious difference between the two screening methods, we performed the analysis separately. Among the nine studies using the IHA method, sex data was provided in only three studies. We compared seroprevalence using different methods in men and women separately. In men, the seroprevalence using the ELISA method was 7.53% (4.81%; 10.78%), and using IHA 2.65% (0.52%; 6.23%). In women, the seroprevalence using the ELISA method was 6.78% (4.38%; 9.66%), and using IHA 2.11% (0.85%; 3.84%). The difference between the sexes was significant (p < 0.05).

Publication bias and sensitivity tests
Funnel plot and Egger's test were both used to examine publication bias. As shown in Figure 5, the funnel plot indicates no publication bias, which was also confirmed from Egger's test (t = 0.11, p = 0.91). A sensitivity analysis was conducted for the pooled results by converting the pooled model (from the random effects model to the fixed effects model). The results demonstrated no large differences in proportions and 95% CIs before and after pooling, indicating stability in the pooled results.

Discussion
Although T. gondii infection in China has been studied for 60 years [24] and many papers have investigated the prevalence of T. gondii in different populations, including blood donors, no systematic review on T. gondii in blood donors was carried out. In this study, we searched databases and identified a total of 40 relevant articles which contained eligible data on the seroprevalence of T. gondii infection in 49,784 blood donors across mainland China. To our knowledge, this is the first report to evaluate the national level of T. gondii seroprevalence in blood donors, which could be of great importance to public health surveillance and associated control policies. The overall seroprevalence of T. gondii infection in blood donors in mainland China from 1986 to 2017 was 6.26% (95% CI: 4.62%-8.13%). Our study showed a low seroprevalence of T. gondii infection in blood donors in mainland China compared to the average seroprevalence of 33% (95% CI: 28%-39%) worldwide [8]. Compared to the prevalence recorded in other Asia countries, the prevalence of T. gondii in China was the lowest [8].
T. gondii is widely distributed, especially in warm, moist and low altitude regions, and at temperate to tropical temperatures oocysts remain infectious for up to 1.5 years [20]. In this study, it was interesting to note that the cold northeast regions of China, at high attitude, had the highest prevalence, and the warm and low attitude regions in the south of China had the lowest prevalence. The result was consistent with Pan's review [24]. This may be related to economic development levels and sanitary conditions. Another reason was that the investigations available for the north of China were few; in some regions, only one province reported the prevalence of T. gondii infection.
In our research, there were two factors, namely screening methods and education levels, which were associated with T. gondii seroprevalence. An experiment that compared IHA, ELISA and another screening method with each other found that there was no significant difference in sensitivity and specificity between IHA and ELISA [50]. However, we found that testing methods may be one of the main sources of heterogeneity in this meta-analysis. In addition, it may be caused by the sample size and/or other confounding factors. Worldwide, T. gondii infection is associated with gender, age, contact with animals and raw meat consumption [8], but not with blood group. However, in Iran, the difference between men and  women was not statistically significant [19]. In our study, lower educational level blood donors had higher seroprevalence, which may be related to the living conditions and chances of contact with animals.
There remain some limitations in our study. First, the articles in this current study were mostly published ten to twenty years ago; only six papers were published in the last five years. This shows that less attention has been focused on T. gondii  infection in blood donors in recent years. Second, most of the studies focused on seroprevalence and the methods were varied; only two of them selectively detected DNA positivity for T. gondii [7,59]. In addition, there was no investigation of the patients who received DNA-positive blood. Therefore, it could not be determined whether there was a possibility of transfusion-transmitted toxoplasmosis. Third, in terms of risk factors, only the characteristics of blood donors were analyzed in most of studies Risk factors like contact with animals (cats or dogs) or eating uncooked or raw meat and vegetables were not investigated in the articles. Fourth, to make the data comparable and to minimize the heterogeneity, we only adopted IgG as an indicator to analyze the prevalence of T. gondii infection. In diagnosis of T. gondii infection, the most useful indicator of active infection may be IgG and/or IgM. In this analysis of 40 studies, IgG and/or IgM was used in only 19 studies, IgM was used in two additional studies [5,46]; the data about IgG and IgM need to be explored further.
In conclusion, the prevalence of antibodies to T. gondii in Chinese blood donors was lower than in other countries. However, the risk of transfusion-transmitted toxoplasmosis still exits. Only in one report, two patients were confirmed to have transfusion-transmitted toxoplasmosis from May 1986 to 1989 in China [36]. Although direct identification methods of T. gondii infection have been developed, routine use is not available in blood banks. In most Chinese blood centers, blood is processed by leukocyte filtration, which reduces the number of mandatory intracellular pathogens including T. gondii. The risk of T. gondii infection was also greatly reduced. Based on this, it is not necessary to screen T. gondii in blood donations in China.