The literature was comprehensively screened and eligible studies were collected to examine the effects of CMV infection on the prognosis of IBD. The following computerized databases were screened: Springerlink (link.springer.com), PubMed (www.ncbi.nlm.nih.gov/pubmed), Web of Science (wok.mimas.ac.uk), Wiley Online Library (onlinelibrary.wiley.com), Chinese Biomedical Database (http://www.sinomed.ac.cn/zh/), China National Knowledge Infrastructure (www.cnki.net), Wanfang database (www.wanfangdata.com) and the VIP database (www.cqvip.com). Additional pertinent studies were obtained by quadratic search. The search included studies published until September 2014. A combination of keywords and free words was applied in the process of collecting literature with a highly efficient and sensitive searching strategy. The search involved a CMV keyword (‘cytomegalovirus’ or ‘cytomegaloviruses’ or ‘salivary gland viruses’ or ‘viruses, salivary gland’ or ‘herpesvirus 5, human’) and an IBM keyword (‘inflammatory bowel diseases’ or ‘inflammatory bowel disease’ or ‘bowel diseases, inflammatory’).

Eligible studies were selected according to the following inclusion criteria: i) Clinical cohort studies; ii) subjects in the enrolled studies had a confirmed diagnosis of IBD based on clinical, radiologic, endoscopic and histologic parameters (14); iii) the detection method used was polymerase chain reaction or enzyme-linked immunosorbent assay; iv) the outcome index included the progress of CMV-positive and CMV-negative patients with IBD, as well as the effects of glucocorticoid treatment, colectomy rate, prevalence of severe IBD and the disease extent; and v) the language of selected studies was restricted to Chinese and English. The corresponding exclusion criteria were: i) Studies containing summary and abstracts only; ii) studies with insufficient information; iii) duplicated publications; and iv) unpublished studies.

The data from eligible studies were extracted by two separated investigators. Relevant information was collected as follows: First author, publication year, study design, country of study and patient ethnicity, patient gender and age, effects of glucocorticoid treatment, disease duration, colectomy rate, severity degree of diseases and disease extent. Subsequently, the two investigators independently assessed the methodological quality of the included cohort trials using the Newcastle-Ottawa Scale (NOS) criteria (15). The NOS criteria associated with the selection of the cohort were as follows: Representativeness of the exposed cohort (NOS1); selection of the non-exposed cohort (NOS2); ascertainment of exposure (NOS3); and demonstration that an outcome of interest was not present at the beginning of the study (NOS4). In addition, NOS criteria involving the comparability of the cohorts were as follows: The study was selected and analyzed according to the most important factor (NOS5); and the study controlled other confounding factors (NOS6). Finally, the following NOS criteria involved the assessment of the outcome: Sufficiently long follow-up for outcome to occur (NOR8); and adequacy of cohort follow-up (NOR9). A third investigator was prepared to resolve any disagreement on the inclusion of a single study.

Stata 12.0 software (Stata Corp., College Station, TX, USA) was used for statistical analysis. The fixed effects model or random effects model was adopted for the calculation of the relative risk (RR) and standardized mean difference (SMD), along with the 95% confidence interval (95% CI). These statistical models aimed to evaluate the prognosis of patients with IBD in association with the effects of corticosteroid therapy, colectomy ratio, prevalence of severe IBD and disease extent. Z-test was employed to detect the significance of the pooled effect size (15,16). In addition, Cochran's Q-test (in which differences with P<0.05 were considered as statistically significant) and I² tests were employed to quantify heterogeneity among the parameters of the included trials (17). When P<0.05 or I²>50% were obtained, which indicated heterogeneous results, the random effects model was conducted; otherwise, the fixed effects model was implemented. In order to evaluate the influence of single study on the overall estimate, a sensitivity analysis was employed according to the Cochrane Handbook for Systematic Reviews of Interventions (http://handbook.cochrane.org/). Furthermore, for the purpose of ensuring the reliability of the results, publication bias was examined by funnel plots and Egger's linear regression test (with P<0.05 indicating a statistically significant difference) (18,19).

Initially, a total of 195 studies (28 in Chinese and 167 in English) were identified from the literature screening, including 193 from electronic search and 2 from manual search. Subsequently, 142 articles were excluded due to not meeting the inclusion criteria following review of the title and abstract, and 46 full-text articles remained. Ultimately, 7 studies (6 in English and 1 in Chinese) satisfied the inclusion criteria after elimination of 36 off-topic studies and 3 studies with insufficient information. The enrolled studies (1,12,13,20–23) were published between 2001 and 2013, and provided complete clinical data for 374 patients with IBD. Among these 7 studies, 5 studies included Asian patients and 2 studies included Caucasian patients. One was conducted in China, 3 were conducted in Japan, 1 was conducted in India, 1 was conducted in France and 1 was conducted in Italy. Furthermore, the 7 studies contained 374 patients with IBD; of these, 104 cases were CMV-positive and 270 cases were CMV-negative. The baseline characteristics and NOS scores for the 7 eligible studies are summarized in Table I and Fig. 1, respectively.

Of the studies included in the meta-analysis, 3 reported the disease duration of CMV-positive and CMV-negative patients (12,20,23). Heterogeneity was observed (I², 95.8%; P_h<0.001) and the random effect model was applied to compare the observations of the studies. According to the meta-analysis results, the disease duration of CMV-positive patients was significantly reduced when compared with that of CMV-negative patients (SMD, −0.81; 95% CI, −1.19 to −0.43; P<0.001), and the forest plot of this analysis is shown in Fig. 2A. The results show that CMV infection may be associated with the disease duration of IBD.

A total of 4 studies reported the efficacy of corticosteroid therapy in both CMV-positive and CMV-negative patients (1,20,22,23). No heterogeneity was observed among the 4 studies (I², 32.1%; Ph=0.220), and thus the fixed effect model was applied. As shown in Fig. 2B, the results of the present meta-analysis demonstrated that corticosteroid therapy had a markedly improved efficacy in CMV-positive patients compared with their efficacy in CMV-negative patients (RR, 1.24; 95% CI, 1.02–1.49; P=0.029). The results show that CMV infection may be associated with the corticosteroid therapy efficacy of IBD.

Out of the 7 eligible studies, 4 studies reported the colectomy rate in both CMV-positive and CMV-negative patients (1,20,22,23). No heterogeneity was observed (I², 32.8%; P_h=0.216) and the fixed effect model was applied to examine the association of colectomy rate with CMV infection. As shown in Fig. 2C, the findings of the present study suggested that a significantly higher rate of CMV-positive patients received colectomy, when compared with CMV-negative patients (RR, 2.13; 95% CI, 1.03–4.40; P=0.042). The results show that CMV infection may be associated with the colectomy rate in IBD patients.

In total, 3 studies reported the incidence of severe IBD in both CMV-positive and CMV-negative patients (1,20,23). No heterogeneity was observed (I², 45.8%; P_h=0.158) and the fixed effect model was used to perform meta-analysis. The results identified that the incidence of severe IBD in CMV-positive patients was significantly higher compared with that in CMV-negative patients (RR, 1.32; 95% CI, 1.04–1.67; P=0.022; Fig. 2D). The results show that the CMV infection may be associated with the incidence of severe IBD.

The association of the area to which IBD extended with CMV infection was also investigated. In total 5 of the eligible studies reported the presence of pancolitis (1,12,20,22,23) and 5 studies reported left-sided IBD (1,12,13,21,22) in both CMV-positive and CMV-negative patients. No heterogeneity was observed (I²<0.01%; P_h=0.602) and the fixed effect model was applied for meta-analysis. Considering the IBD onset area, CMV-positive patients were found to present significantly higher susceptibility to pancolitis compared with the CMV-negative patients (RR, 1.31; 95% CI, 1.01–1.72; P=0.045; Fig. 2E). By contrast, no significant difference was identified in the susceptibility of CMV-positive or CMV-negative patients to the incidence of left-sided IBD (RR, 0.97; 95% CI, 0.72–1.30; P=0.828; Fig. 2F). The results show that CMV infection may be associated with the disease location of IBD.

The results of sensitivity analysis demonstrated that no single study was able to impact the overall estimate of the SMD value of disease duration in the CMV-positive or -negative patients (Fig. 3A). In addition, the RR value of efficacy of corticosteroid therapy, colectomy rate, the incidence of severe IBD and disease location were also not affected by a single study (Fig. 3B-F). Furthermore, publication bias was investigated using funnel plots and Egger's linear regression. As shown in Fig. 4, no evident asymmetry was observed in the shapes of the funnel plots, which appeared to be symmetric, while the Egger's regression test suggested the absence of publication bias. These results suggested that no significant publication bias was detected in the present meta-analysis (P>0.05).