Search Strategy and Included Studies
The systematic database search identified 591 potential publications, of which 515 remained after removing duplicates. Two reviewers then screened for relevant studies based on titles or abstracts and an additional 205 articles were removed. The final 11 studies were assessed for eligibility. Four studies were excluded due to the non-CCRT treatment regimen (n = 1), no correct complete response (CR) or partial response (PR) grouping (n = 1), and no extractable IVIM data (n = 2). Finally, seven articles containing 237 patients were included in the qualitative evaluation[7, 26, 27, 29-32]. Four studies were included in the quantitative assessment, as the patients belonged to the same research center (n = 2) and the study had no original data (n = 1). The screening and selection process is outlined in a PRISMA flow chart (Fig. 1).
Study Design
All included studies were prospective studies. All studies, except one, involved radiology, obstetrics and gynecology, ultrasound, and oncology departments. The study design is summarized in Table 1.
Patient Population
The study consisted of 237 patients. The number of patients receiving chemoradiation ranged from 17 to 63. All patients had primary squamous cell carcinoma (n = 168), except for two studies that did not mention the cancer type. The FIGO staging ranged from Ib to IVb (Ib [n = 4], II [n = 108], III [n = 81], and IV [n = 44]). The average age of the patients ranged from 47 to 69 years. Three studies [27, 30, 31] described metastasis. All studies involved CCRT, including a combination of external beam radiotherapy (EBRT), intracavitary brachytherapy (ICR), and chemotherapy. Patient characteristics are described in Table 2.
MRI Technological Parameters
All the studies included a 3. 0T MRI scanner. None of the studies used anti-spasmodics. All studies, except one, used conventional T1 and T2-weighted sequences. The total acquisition time of the IVIM sequence ranged from 3.1 to 10.0 min in the single-shot echo-planar imaging (SS EPI). All articles used the three-parameter models (D, D*, f) with a fixed S0. Details of the technical characteristics of MRI examinations are provided in Supplementary Table 1, Additional file 1.
MRI IVIM Analysis
All studies, except two, had two reviewers for IVIM analysis and both reviewers reached an agreement through an inter-observer. Details of the IVIM analysis are shown in Table 3.
Follow-up for Therapeutic Efficacy
All patients completed their follow-up period. Only three studies performed an MRI examination along with pathological histology within 1 - 3 months after treatment[26, 27, 31]. The remaining studies were followed up using MRI. These studies[7] used the European Organization for Research and Treatment of Cancer (EORTC) criteria, and the rest applied RECIST. The descriptions of CR groups and non-complete response groups - PR, stable disease, or progressive disease - are described in RECIST (ver. 1.1). The efficacy results at the last follow-up are shown in Supplementary Table 2, Additional file 1.
Subgroup Meta-analyses
Changes in IVIM parameters over time in patients in the CR groups and non-CR groups are presented in Supplemental Table 3. The meta-analysis assessed data available from four longitudinal single-arm studies that reported changes in ADC, D, D*, and f during and after CCRT for cervical cancer are show in Table 4 and Supplementary Figures 1–4, Additional file 1.
In the total pooled effect sizes, f values showed more incremental changes in the CR groups [MD = 6.24 (95% CI: 4.12 to 8.36; p = 0.000)] and in the non-CR groups [MD = 5.84 (95% CI: 2.92 to 8.75; p = 0.000)], but the I2 (94.1%, 91.7%) displayed significant heterogeneity among studies. Similarly, although the numbers are small, the trend indicated an increase in ADC in the CR groups [MD = 0.51 (95% CI: 0.39 to 0.63; p = 0.000); I2 = 89.0%] and the non-CR groups [MD = 0.38 (95% CI: 0.15 to 0.62; p = 0.001); I2 = 96.2%], and in the D value in the CR groups [MD = 0.18 (95% CI: 0.16 to 0.21; p = 0.000); I2 = 99.4%] and in the non-CR groups [MD = 0.16 (95% CI: 0.13 to 0.20; p = 0.000); I2 = 98.1%], respectively.
Early Therapeutic Response
In the subgroup meta-analysis, ADC and D values showed a significant increase during the 2nd and 4th week after CCRT in the CR groups and the non-CR groups. Although the incremental change was not as large, there were statistically significant incremental changes in f values both in the CR groups [MD = 0.8.42 (95% CI: 2.49 to 14.35; p = 0.005); I2 = 91.7%] and the non-CR groups [MD = 3.26 (95% CI: 1.13 to 5.40; p = 0.003); I2 = 0.0%] during the 3rd week of treatment. Supplementary Figures 1 –4, Additional file 1, show details of IVIM parameters assessing the early treatment response.
Our study showed that all indicators increased following the 1st week of treatment, but the increase in the CR group was greater than that in the non-CR group (Supplemental Table 3, Additional file 1).
Therapeutic Response Before and After Chemotherapy
We found that the f values demonstrated a statistically significant improvement in evaluating effectiveness before and after therapy in the CR groups [MD = 6.76 (95% CI: 1.78 to 11.74; p = 0.000); I2 = 91.0%] and the non-CR groups [MD = 7.61 (95% CI: 3.71 to 11.52; p = 0.000); I2 = 77.8%], respectively. Supplemental Figures 1 – 4, Additional file 1, show the values of ADC and D pre- and post-treatment. The D* values indicated no statistically significant increase or decrease both at pre- and post-treatment.
Begg’s publication bias tests for ADC values in the non-CR group showed p = 0.024, indicating publication bias from heterogeneous sources. Whereas, Begg’s publication bias tests for ADC, D, D*, and f values were quantitatively evaluated to be p > 0.05, indicating non-publication bias from heterogeneous sources (Table 4).
Risk of Bias
Supplemental Table 4, Additional file 1, and Figure 2 show the risk of bias analyses. We applied QUIPS for evaluation and obtained the following results. Only one study[30] mentioned the failure of follow-up and its causes. Three studies[27, 30, 31] described other prognostic factors. None of the included studies made an effort to modify probable confounding factors.