Evaluating the Impact of Modic Changes on Operative Treatment in the Cervical and Lumbar Spine: A Systematic Review and Meta-Analysis

Modic changes (MCs) are believed to be potential pain generators in the lumbar and cervical spine, but it is currently unclear if their presence affects postsurgical outcomes. We performed a systematic review in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. All studies evaluating cervical or lumbar spine postsurgical outcomes in patients with documented preoperative MCs were included. A total of 29 studies and 6013 patients with 2688 of those patients having preoperative MCs were included. Eight included studies evaluated cervical spine surgery, eleven evaluated lumbar discectomies, nine studied lumbar fusion surgery, and three assessed lumbar disc replacements. The presence of cervical MCs did not impact the clinical outcomes in the cervical spine procedures. Moreover, most studies found that MCs did not significantly impact the clinical outcomes following lumbar fusion, lumbar discectomy, or lumbar disc replacement. A meta-analysis of the relevant data found no significant association between MCs and VAS back pain or ODI following lumbar discectomy. Similarly, there was no association between MCs and JOA or neck pain following ACDF procedures. Patients with MC experienced statistically significant improvements following lumbar or cervical spine surgery. The postoperative improvements were similar to patients without MCs in the cervical and lumbar spine.


Introduction
Vertebral bone marrow edema is a recognized clinical entity that is believed to be associated with degenerative disc disease and back pain [1]. The aberrant signal changes identified with magnetic resonance imaging (MRI) were characterized into specific types by Modic et al., and were thus termed Modic changes (MC). Each distinct signaling phenotype is characterized by their unique signal intensity in MRI. Type I changes are characterized by hypointense T1-and hyperintense T2-weighted images, type II changes are recognized as hyperintense signaling on T1-and isointense to slightly hyperintense signaling on T2-weighted imaging, while type III changes are predominantly hypointense on the T1 and T2-weighted images [1]. Histological analysis has uncovered that the MRI signal intensity corresponds to anatomical changes. For example, type I MCs demonstrate bone marrow replacement with fibrovascular stroma, type II MCs are associated with fatty infiltration of the bone marrow, and type III MCs are characterized by sclerotic endplate changes [2,3]. Longitudinal studies have suggested that these changes exist on a continuum with transformations reported between all types, but most commonly with the progression of type I to type II [2,[4][5][6][7][8]. Degenerative disc disease is the most frequently identified disease accompanying MCs [1,9,10], but other injury patterns associated with MCs include autoimmune disease, low virulence bacterial infections, and microtrauma resulting in endplate abnormalities [11][12][13].

Study Eligibility
Studies were included if MCs, identified in preoperative MRI, were operatively treated in either the cervical or lumbar spine. Studies were excluded if (1) clinical follow-up time was not reported or it was less than one year; (2) a full text manuscript could not be obtained; (3) the article was not written in or translated to English; or (4) the article was a letter to editor or systematic review.
Two reviewers independently screened the identified articles and selected studies for full-text review after screening of the title and abstract. Articles were screened for inclusion based on the predetermined inclusion/exclusion criteria during the full-text review. A search of the references was performed of all articles meeting the inclusion criteria to identify any potentially missed manuscripts. For studies where article inclusion was unclear, a more senior author was consulted to resolve any discrepancies.

Data Collection Quality Assessment
The authors extracted all potentially relevant data from the identified studies with multiple self-designed tables. Data including the type and prevalence of Modic signal changes, clinical outcome measures, patient population, procedure type, and key study findings were reported based on the affected region of the spine. Studies were assessed for bias using the validated Newcastle-Ottawa Score with a high-quality score defined as a score greater than or equal to 6 [24].

Meta-Analysis
Due to significant data heterogeneity, variability in the clinical outcome measures assessed, and the different surgical procedures performed in the lumbar spine, only a limited meta-analysis could be performed. For meta-analysis inclusion, a minimum of four studies were required to report on the same patient reported outcome (PRO) with only four PROs meeting this criteria (Visual Analog Scale (VAS) Back, VAS Neck, Oswestry Disability Index (ODI), and Japanese Orthopedic Association (JOA) score). The weighted mean difference in values between patients with and without MCs was then assessed using forest plots and data heterogeneity was calculated as an I 2 value. The meta-analysis data were generated using R Studio Version 4.0.2 (Boston, MA, USA).

Results
The initial PubMed search identified 259 articles, after which the title and abstract screening identified 36 potentially relevant articles. Two were excluded due to an inadequate follow-up of less than one year [25,26], another two were excluded because of the risk of potentially identical patients [27,28], and one article's full-text could not be obtained [29]. Thirty-one articles were identified after a full-text review. Two studies were classified as a high risk of bias, so they were also excluded [30,31]. In total, 29 articles were included in the final analysis ( Figure 1). The characteristics of the studies included in the systematic review are presented in Tables

Cervical Spine
Eight retrospective studies described the correlations between MC and surgical outcomes (Table 3). No prospective studies evaluated MC in the cervical spine. In general, studies did not report a difference in the PROs based on the presence of MC.   Huang et al. [34] reviewed 116 cases of single-level anterior cervical discectomy and fusions (ACDF) for the presence of MC and their association with fusion rates and PROs. Patients with type II MC experienced significantly delayed early fusion rates at 3-and 6months, but the fusion rates were similar at one year. There were no significant differences between groups with regard to improvement in the VAS scores or JOA scores.
In a study by Li et al. [38] in 2022, 124 patients underwent single-level anterior cervical corpectomy and fusion for cervical myelopathy. They found that 41% of patients with MC had cage subsidence, defined as at least 1 mm, compared to only 15.9% of patients without MC (p = 0.003). Subsidence did not vary between the Modic subtypes. Patients with type I MC had a higher proportion of partial fusions, defined by incomplete bony remodeling (40% compared to 11% of controls). No cases of pseudarthrosis were observed and no other differences were identified with regard to the JOA, neck disability index (NDI), and VAS neck and arm pain scores.
Yang et al. [32] retrospectively reviewed patients who underwent one-level ACDFs or cervical disc arthroplasty. The presence of MCs did not result in different one-year VAS neck, NDI, or physical component summary (PCS-12) and mental component summary (MCS-12) scores from the 36-Item Short Form Health Survey (SF-36).
In a 2017 study by Li et al. [36], 72 patients were analyzed with type II MC and minimum five-year follow-up. Of those, 35 received cervical disc arthroplasty and 37 underwent ACDF for myelopathy or radiculopathy. All patients experienced postoperative improvement assessed by modified JOA (mJOA), NDI, and VAS, without differences between groups. The patients in the disc replacement group were noted to have improved range of motion at final follow-up.
In 2015, Li et al. [35] retrospectively compared 35 patients who had a single-level ACDF with type II MC at an adjacent level to 213 patients without type II MC. They observed no significant differences between groups with regard to the range of motion or disc height at adjacent levels, and no differences in improvement by the mJOA and NDI scores. The patients in the MC group had worse VAS neck pain at one-year follow-up (p < 0.05), although these differences resolved by 5-year follow-up.
A retrospective review of 117 patients with a single-level ACDF found that preoperative MCs at levels adjacent to surgery were associated with greater axial symptoms (p = 0.015) [37]. However, all patients demonstrated clinical improvement as assessed by the JOA without a significant difference between patients who did and did not have MCs.
In a separate 2015 study by Li et al. [39], 106 patients were retrospectively identified with one-level ACDFs for cervical spondylotic myelopathy. Preoperatively, 23 patients had type I MC, 28 had type II MC, and 44 had no MC. The patients in the type I MC group had significantly lower VAS neck pain at 24 months after surgery (1.5 ± 1.1) compared to patients without (MC 2.0 ± 1.5), although a specific p-value was not provided.
Baker et al. [33] retrospectively examined the records of 861 patients who underwent a one-to four-level ACDF for radiculopathy or myelopathy. Of those, 365 patients had MCs. No significant differences in postoperative NDI, VAS-neck, or VAS-arm were found between groups. However, after stratification by cervical level, they identified the presence of MC at C7-T1 predicted worse postoperative NDI (p < 0.001).
A limited number of studies collected PROs in a standardized manner, thus, a metaanalysis could only be performed for postoperative values. Four studies reported on neck pain and JOA following ACDF [34,35,38,39]. Random effects analysis of these studies demonstrated no significant association between MC and postoperative VAS neck scores (MD −0.17, 95% CI: −0.50-1.70) (Figure 2a) or the JOA scores (MD −0.07, 95% CI: −0.35-0.20) (Figure 2b). Among these studies, moderate heterogeneity in neck pain scores (I 2 = 52%) and low heterogeneity in JOA scores (I 2 = 0%) were observed. had significantly lower VAS neck pain at 24 months after surgery (1.5 ± 1.1) compared to patients without (MC 2.0 ± 1.5), although a specific p-value was not provided. Baker et al. [33] retrospectively examined the records of 861 patients who underwent a one-to four-level ACDF for radiculopathy or myelopathy. Of those, 365 patients had MCs. No significant differences in postoperative NDI, VAS-neck, or VAS-arm were found between groups. However, after stratification by cervical level, they identified the presence of MC at C7-T1 predicted worse postoperative NDI (p < 0.001).
A limited number of studies collected PROs in a standardized manner, thus, a metaanalysis could only be performed for postoperative values. Four studies reported on neck pain and JOA following ACDF [34,35,38,39].

Lumbar-Discectomy/Microdiscectomy
There were a total of 11 articles discussing the outcomes after discectomy or microdiscectomy with three articles comparing those procedures to lumbar fusion ( Table 4). Regardless of the presence of MC, patients experienced significant improvement following surgery. However, the studies delivered mixed results as to whether MCs modified the magnitude of benefit obtained from surgery.

Lumbar-Discectomy/Microdiscectomy
There were a total of 11 articles discussing the outcomes after discectomy or microdiscectomy with three articles comparing those procedures to lumbar fusion ( Table 4). Regardless of the presence of MC, patients experienced significant improvement following surgery. However, the studies delivered mixed results as to whether MCs modified the magnitude of benefit obtained from surgery.    El Barzouhi et al. [41] conducted a multicenter, randomized trial of 283 patients with sciatica. They compared microdiscectomy to conservative care. Of the 283 patients, 41 percent of patients had vertebral endplate signal changes (VESCs). No significant differences in the amount of recovery or disabling back pain were identified between patients with and without VESCs at one year follow-up.
Gautschi et al. [56] prospectively identified 338 patients with lumbar degenerative disc disease. Of the 338 patients, 175 of them had a microdiscectomy. Similar outcomes for VAS back and leg pain, ODI, Roland-Morris Disability Index, Timed Up and Go test, EuroQol-5D (EQ-5D), and PCS-12 were identified between patients in the MC and no-MC cohorts.
Kumarasamy et al. [21] prospectively followed 309 patients undergoing microdiscectomy, 86 of whom had MC. The patients in the MC group had worse postoperative numeric rating scale (NRS) back pain (1.6 to 1.1, p = 0.001), although the overall improvement in both groups was 4.3 points, indicating that there was no difference in the magnitude of overall improvement. Both groups experienced a significant improvement in disability, but the group without MC experienced statistically significant greater improvement, although this did not meet the MCID threshold. Patient satisfaction, as evaluated by the MacNab criteria, was noted to be significantly worse for patients with MCs.
Sørlie et al. [45] prospectively evaluated 178 patients undergoing microdiscectomy and found that 36 patients had type I MCs. Multivariate analysis demonstrated no statistical differences between groups for VAS back and leg pain, ODI, and EQ-5D. All patients had significant improvement after surgery, although patients who smoked and had type I MCs had less improvement.
Rahme et al. [54] retrospectively identified patients treated with lumbar microdiscectomy. They included 41 patients in the final analysis and identified 19 patients with preoperative MC at the level of the operation. At median follow-up of 41 months, 32 patients had MCs (since four of the 22 patients without MCs converted to type I MCs and another nine converted to type II MCs). Additionally, 60% of the type I MCs converted to type II MCs. The study concluded that the majority of patients after a lumbar discectomy will convert to type II MCs, but MCs do not results in worse outcomes, recurrent symptoms, or decreased satisfaction.
Ohtori et al. [53] prospectively identified 23 patients with type I MC undergoing lumbar discectomy and matched them to 22 patients without MC. All patients exhibited improvement in VAS back, JOA, and ODI with no significant differences between the groups at 12 or 24 months after surgery.
MacLean et al. [23] retrospectively analyzed 129 patients undergoing discectomy for lumbar radiculopathy with 77 patients having MCs. They had complete outcome data for 96 of these patients and concluded each group experienced statistically and clinically significant improvement as assessed by the PCS-12, ODI, and VAS leg pain scores. Most subgroups met the MCID and there were no significant differences in clinical improvement based on the presence or absence of MCs.
Lurie et al. [49] retrospectively reviewed 307 patients treated with discectomy for radicular pain and identified 81 patients with MC. The patients with type I MC improved significantly less in ODI compared to the type II or no MC patients. However, patients with type I MC did experience an average improvement in ODI of 26.4 points following discectomy.
Udby et al. [44] retrospectively reported on 620 patients undergoing lumbar discectomy for radiculopathy with 290 patients having MC. At two year follow-up, they concluded that both groups had significant improvement by ODI, EQ-5D, VAS back and leg pain, with no significant differences between groups.
Cao et al. [48] retrospectively reviewed 91 patients with lumbar disc herniation and MC who had a combination of low back and radicular leg pain. A total of 47 patients were treated with discectomy, while 44 underwent a posterior lumbar interbody fusion (PLIF). All patients had significant VAS back pain improvement; however, the patients in the PLIF group experienced greater VAS back improvement. Both groups had significant VAS leg pain improvement and there was no postoperative difference based on the type of surgery.
Xu et al. [50] retrospectively reviewed 276 patients undergoing percutaneous endoscopic lumbar discectomy for radiculopathy and identified 94 patients with MC. They found no differences in VAS leg pain through final follow-up of 30 months. At three months follow-up, all patients had similar improvement in ODI and VAS back pain, however, the patients with MCs were found to have worse trending ODI and VAS back pain at final follow-up.
Ulrich et al. [42] retrospectively identified 205 patients undergoing lumbar discectomy or lumbar fusion. They found no differences in the outcomes between patients with MCs versus those without MCs, regardless of the procedure type when assessing the Spinal Stenosis Measure, EQ-5D, and NRS back pain scores at the 36 month post-surgical visit. Over 70% of patients reached the postoperative MCID at 36 month follow-up. Thus, the study concluded that MCs do not significantly affect the postoperative clinical outcomes.
Five studies reported values for ODI [21,23,44,45,50], and four studies reported VAS back pain [21,44,45,50]. A meta-analysis was performed among these studies and identified no significant associations between MC and either postoperative VAS back pain or ODI ( Figure 3). However, significant data heterogeneity was present for VAS and ODI (I 2 = 88% and I 2 = 96%, respectively), indicating that additional high quality studies are required to confirm the finding that MCs do not affect the improvement in VAS back or ODI scores postoperatively.
at final follow-up.
Ulrich et al. [42] retrospectively identified 205 patients undergoing lumbar discectomy or lumbar fusion. They found no differences in the outcomes between patients with MCs versus those without MCs, regardless of the procedure type when assessing the Spinal Stenosis Measure, EQ-5D, and NRS back pain scores at the 36 month post-surgical visit. Over 70% of patients reached the postoperative MCID at 36 month follow-up. Thus, the study concluded that MCs do not significantly affect the postoperative clinical outcomes.
Five studies reported values for ODI [21,23,44,45,50], and four studies reported VAS back pain [21,44,45,50]. A meta-analysis was performed among these studies and identified no significant associations between MC and either postoperative VAS back pain or ODI (Figure 3). However, significant data heterogeneity was present for VAS and ODI (I 2 = 88% and I 2 = 96%, respectively), indicating that additional high quality studies are required to confirm the finding that MCs do not affect the improvement in VAS back or ODI scores postoperatively.

Lumbar Fusion
Nine studies evaluating lumbar fusion surgery were identified. One study evaluated oblique lateral interbody fusions (OLIF) [43], two evaluated transforaminal lumbar interbody fusions (TLIFs) [23,40], two included multiple fusion techniques [51,56], two evaluated PLIF [42,47], and two evaluated posterolateral fusions [48,58]. In general, patients had substantial improvements in the clinical outcomes and fusion rates regardless of the presence of MCs, with the exception of one study, which found decreased fusion rates and worse outcomes in patients with type III MCs [58].
Chung et al. [43] retrospectively analyzed 86 patients with 125 operated levels by one-or two-level OLIF and identified MC at 72 of these levels. They found no association between the MC and fusion rate or cage subsidence at 28 months after surgery.
Jiao et al. [40] retrospectively identified 89 patients, 51 with MC, who underwent single-level TLIFs. The authors found no difference in the fusion rate or clinical outcomes based on MCs, but type I MCs were associated with significantly higher rates of cage subsidence (40% to 15.8%). All other outcome measures were similar between groups including disc height, segmental lordosis, and lumbar lordosis.
MacLean et al. [23] retrospectively reviewed 44 patients who underwent TLIF for radiculopathy and instability. They concluded that patients experienced significant improvement by PCS-12, ODI, and VAS leg pain regardless of the presence of MCs.
Two studies reported on multiple fusion techniques: TLIF, PLIF, PLF, extreme lateral interbody fusion, circumferential fusion, and anterior lumbar interbody fusion [51,56]. Between these studies, 121-disc levels were included in their analysis with 73 levels having MCs. No outcome measures were significantly affected by MCs including ODI, VAS back and leg, or SF-36.
Of the two studies reporting on decompression with PLIF, Ohtori et al. [47] prospectively evaluated 33 patients with MC while Ulrich et al. [42] retrospectively identified 57 patients undergoing lumbar fusion with 41 having MCs and 16 having no MCs. All patients in the study by Ohtori et al. [49] had solid arthrodesis at 24 months after surgery with average union at nine months. Neither study found differences in the VAS nor ODI based on the presence of MCs.

Other Lumbar Surgeries
We identified four articles in our search that did not fit into the above categories. Three of these studies reported on disc arthroplasty while the other study reported on vertebroplasty. Masala et al. [52] prospectively identified 218 patients with type I MC who underwent vertebroplasty. Of those, 98% showed improvement in VAS and ODI, 1% showed no improvement, and 1% died for unrelated reasons. Three prospective studies evaluated lumbar disc arthroplasty [46,55,57]. They all reported improved ODI in patients with MC compared to those without. Gornet et al. [46] reported these findings for type II MC, Blondel et al. [55] reported these findings for type I MC, and Hellum et al. [57] reported these findings for both type I and II MCs.

Discussion
Radiographic evidence of intervertebral disc degeneration is reported to be present in greater than 90% of individuals by the time they reach 50 years of age [59]. Most patients with disc degeneration remain asymptomatic, but within the asymptomatic population, 21% of patients older than 60 have spinal stenosis, while 36% have a herniated intervertebral disc [60]. Therefore, identifying potential factors that may result in back or neck pain is prudent [61]. Modic changes are one potential etiology for symptomatic back and neck pain, thus an evaluation of their potential improvement after surgery is indicated. Although the heterogeneity in the literature precluded our team from performing a meta-analysis of all of the included studies, the majority of studies assessing the surgical outcomes after cervical and lumbar spine surgery have not detected clinically important differences between patients with and without Modic changes. Despite heterogeneity, our metaanalysis similarly identified no differences between postoperative VAS back pain or ODI in patients with MC undergoing lumbar discectomy. There was less data heterogeneity in patients undergoing ACDF, and our meta-analysis found no differences in postoperative VAS neck pain or JOA based on the presence of MCs.
Surgeons perform over 100,000 ACDFs each year [62] and MCs are present at rates surpassing 40% in patients greater than 50 years of age who undergo cervical spine MRI [63]. The results of our pooled analysis found that MCs were present in a similar 36.5% of patients who required cervical spine surgery. Thus, understanding the potential indicators of surgical success (e.g., presence or absence of MCs) is important for risk stratification and guiding the patient's expectations on their potential postoperative clinical improvement. When evaluating the clinical outcomes, our pooled data suggest that MCs are not predictive of postoperative improvement in patients undergoing cervical spine surgery. While one study reported MCs at the level C7-T1 were associated with worse outcomes, this group comprised 1.2% of their overall patient population and was likely severely underpowered to detect a true difference [33]. In contrast, Li et al. [39] reported that patients with type I MC were more likely to have reduced back pain at 2.5 year follow-up. Given that the available data are based on retrospective studies, with approximately half the patients coming from one study [33], it is likely that additional well-designed studies are needed to definitively claim that MCs have no significant clinical effect on cervical spine postsurgical outcomes. However, the evaluated studies do pose a low risk of bias, and the current literature suggests that MCs do not significantly impact the clinical or surgical success of cervical spine surgery [34,38].
Seven studies evaluated the microdiscectomy or discectomy outcomes in relation to MC. While discectomy and microdiscectomy overall appear efficacious as a treatment for axial back pain reduction and PRO improvement for patients with MCs, those with type I MC may experience reduced improvements following discectomy. Only one study evaluating microdiscectomy identified worse improvement in patients with MCs, although this was not clinically significant [21]. Two other retrospective studies identified a relationship between MC and worse PRO improvement (ODI and VAS back) following discectomy. Although these studies identified discectomy as an effective treatment option for patients with MC, patients with type I MC experienced less improvement at long-term follow-up [49,50]. Type I MC likely represents inflammatory changes, which may be linked to exacerbations of low back pain. Despite the propensity of MC to transform to other types, they have also been shown to remain constant over time [64]. Although discectomy may alleviate discogenic pain, it may have little effect on the concomitant changes to the vertebral endplate. This may be one potential reason why patients with MC-I do not exhibit as robust of an improvement when compared to other patients. While the results of our meta-analysis indicated no differences in back pain or disability, significant heterogeneity and the limited number of studies prohibit strong conclusions, and more research is needed, especially comparing the improvement across MC types.
We identified eight retrospective studies and one prospective study reporting MC in lumbar fusion surgery. These studies demonstrated that MC do not influence patientreported outcomes or fusion rates following lumbar fusion surgery regardless of approach. Jiao et al. [40] identified that patients with type I MC experienced greater improvement in PROs following TLIF. Kwon et al. [58] reported lower fusion rates and worse PROs in type III MC. They suggested that given the sclerotic nature of type III MC, additional fusion such as pedicle screw fixation may be required for patients with type III changes [58]. Although this sclerotic bone may interfere with fusion, resorption of the sclerotic bone has been observed over time [65]. Care should be taken when selecting patients for surgery with MC-III because this population remains understudied due to its low prevalence. While lumbar disc replacement was evaluated in three prospective analyses and all three demonstrated excellent results that are not affected by the presence of MC, further research is needed to analyze the impacts of disc arthroplasty on endplate disease.

Conclusions
Since the last review on this topic in 2016, several prospective and large retrospective studies have been published with a low risk of bias [19]. After summarizing all of the available data on surgical patients with MCs, quantitative and qualitative analyses suggest that patients with MCs have similar clinical improvements in PROs (ODI and VAS back/neck) and similar fusion rates when compared to patients without MCs. However, additional high-quality studies are needed to further elucidate changes in the fusion status, especially among patients with type III MC. Furthermore, long-term prospective studies evaluating the outcomes of patients with type I MC undergoing discectomy are merited given that some studies indicate that they have worse clinical improvement after surgery. Institutional Review Board Statement: Institutional review board approval was not obtained as this research did not involve human subjects.

Informed Consent Statement: Not applicable.
Data Availability Statement: All data is contained within the article or appendix.

Conflicts of Interest:
The authors declare no conflict of interest. Provide an explicit statement of questions being addressed with reference to participants, interventions, comparisons, outcomes, and study design (PICOS).

Protocol and registration 5
Indicate if a review protocol exists, if and where it can be accessed (e.g., Web address), and, if available, provide registration information including registration number.

Eligibility criteria 6
Specify study characteristics (e.g., PICOS, length of follow-up) and report characteristics (e.g., years considered, language, publication status) used as criteria for eligibility, giving rationale.

2,3
Information sources 7 Describe all information sources (e.g., databases with dates of coverage, contact with study authors to identify additional studies) in the search and date last searched.
2 Search 8 Present full electronic search strategy for at least one database, including any limits used, such that it could be repeated. 2

Study selection 9
State the process for selecting studies (i.e., screening, eligibility, included in systematic review, and, if applicable, included in the meta-analysis).

Section/Topic Number Checklist Item Reported on Page Number
Data collection process 10 Describe method of data extraction from reports (e.g., piloted forms, independently, in duplicate) and any processes for obtaining and confirming data from investigators. 3

Data items 11
List and define all variables for which data were sought (e.g., PICOS, funding sources) and any assumptions and simplifications made. 3

Risk of bias in individual studies 12
Describe methods used for assessing risk of bias of individual studies (including specification of whether this was done at the study or outcome level), and how this information is to be used in any data synthesis. 3

Summary measures 13
State the principal summary measures (e.g., risk ratio, difference in means). 3

Synthesis of results 14
Describe the methods of handling data and combining results of studies, if done, including measures of consistency (e.g., I 2 ) for each meta-analysis. 3

Risk of bias across studies 15
Specify any assessment of risk of bias that may affect the cumulative evidence (e.g., publication bias, selective reporting within studies). 3

Additional analyses 16
Describe methods of additional analyses (e.g., sensitivity or subgroup analyses, meta-regression), if done, indicating which were pre-specified.

Study selection 17
Give numbers of studies screened, assessed for eligibility, and included in the review, with reasons for exclusions at each stage, ideally with a flow diagram. 3

Study characteristics 18
For each study, present characteristics for which data were extracted (e.g., study size, PICOS, follow-up period) and provide the citations.

3-16
Risk of bias within studies 19 Present data on risk of bias of each study and, if available, any outcome level assessment (see item 12).
3, Tables 1  and 2 Results of individual studies 20 For all outcomes considered (benefits or harms), present, for each study: (a) simple summary data for each intervention group (b) effect estimates and confidence intervals, ideally with a forest plot.

3-16, tables
Synthesis of results 21 Present results of each meta-analysis done, including confidence intervals and measures of consistency.

9, 15 (Figures 2 and 3)
Risk of bias across studies 22 Present results of any assessment of risk of bias across studies (see Item 15).

Summary of evidence 24
Summarize the main findings including the strength of evidence for each main outcome; consider their relevance to key groups (e.g., health care providers, users, and policy makers).

16, 17
Limitations 25 Discuss limitations at study and outcome level (e.g., risk of bias), and at review-level (e.g., incomplete retrieval of identified research, reporting bias). Describe sources of funding for the systematic review and other support (e.g., supply of data); role of funders for the systematic review.