High rates of spin bias in systematic reviews regarding anterior cervical discectomy and fusion vs. posterior cervical foraminotomy: a systematic review Authors: Daniel Rusu, Jadesola Olurin, Aidan Lindgren, Niels Mohty, Henry Avetisian, Marc A. Abdou, Jeffrey C. Wang, Raymond J. Hah, Ram K. Alluri Published: 2026 ABSTRACT BackgroundSurgical management for cervical degenerative disc disease includes anterior cervical discectomy and fusion (ACDF—the gold standard) and posterior cervical foraminotomy (PCF); however, current literature remains conflicted over the superiority of one procedure. Systematic reviews allow for comprehensive synthesis of literature but are susceptible to spin bias that may skew results and affect clinical decision making. This study aimed to evaluate the presence and number of spin bias in systematic reviews and meta-analyses comparing ACDF and PCF. To evaluate the prevalence, type, and predictors of spin bias in systematic reviews and meta-analyses comparing anterior cervical discectomy and fusion (ACDF) and PCF.MethodsA systematic search was conducted on PubMed, Web of Science, Embase, and Scopus for systematic reviews and meta-analyses comparing ACDF and PCF in June 2025. Articles were included if they were systematic reviews with or without meta-analysis, discussed treatment with surgical intervention, compared outcomes between ACDF and PCF, were available in English, and had the full text available. Any articles that did not follow these criteria were excluded. Included articles were assessed for the 9 most severe types of spin bias established by Yavchitz et al. and graded on methodological quality using A Measurement Tool to Assess Systematic Reviews, version 2 (AMSTAR-2).ResultsSpin bias was found in 11 of 12 (92%) included articles, with number of different types of bias ranging from 0 to 2. Type 9 was the most common type of spin in 5 of 12 (42%) articles. Type 5 spin was found in 4 (33%) articles, Type 3 in 3 (25%) articles, Type 1 in 2 (16%) articles, and Types 2 and 4 in 1 (8%) article each. AMSTAR-2 quality assessments rated 7 of 12 (58%) articles as “critically low”, and 5 of 12 (42%) as “low”; none received a “high” quality rating.ConclusionsSpin bias is highly prevalent (92%) in systematic reviews and meta-analyses comparing ACDF and PCF, with a maximum of 2 types of spin per article. All studies received either “critically low” or “low” quality ratings according to AMSTAR-2 criteria. Notably, all studies rated as “critically low” quality had at least 1 type of spin bias. Efforts and strategies to mitigate spin bias in orthopaedic literature are imperative to collectively enhance published literature. Highlight box Key findings • Spin bias was present in 92% of systematic reviews and meta-analyses comparing anterior cervical discectomy and fusion (ACDF) and posterior cervical foraminotomy (PCF). • The most common forms of spin involved favorable conclusions despite high risk of bias or methodological limitations in included primary studies. • All included reviews were rated as either “low” or “critically low” quality using A Measurement Tool to Assess Systematic Reviews, version 2 (AMSTAR-2) criteria. What is known and what is new? • ACDF and PCF are widely used surgical options for cervical radiculopathy, and systematic reviews are frequently relied upon to guide clinical decision making. However, prior studies across orthopaedic subspecialties have demonstrated that spin bias can distort study conclusions. • This is the first study to systematically evaluate spin bias specifically in systematic reviews and meta-analyses comparing ACDF and PCF. The prevalence of spin in this literature is among the highest reported in orthopaedic research and is consistently associated with poor methodological quality. What is the implication, and what should change now? • Spine surgeons and clinicians should critically appraise systematic reviews comparing ACDF and PCF rather than relying solely on abstracts or conclusions. • Authors, reviewers, and journals should adopt stricter adherence to reporting and quality assessment tools (PRISMA, AMSTAR-2) and increase awareness of spin bias to improve the reliability of evidence guiding surgical decision making. Introduction Cervical spondylosis can result in the development of neck pain and neurological symptoms including radiculopathy or myelopathy. Studies have shown that 80–90% of the population has degenerative cervical spondylosis on magnetic resonance imaging (MRI) by the age of 50 years old (1). Initial treatment is conservative, followed by surgical management for refractory symptoms. Options for operative management in cases presenting with radiculopathy may include anterior cervical discectomy and fusion (ACDF) and posterior cervical foraminotomy (PCF). ACDF is considered the gold standard surgical intervention for cervical spondylosis with intractable or progressive radiculopathy or myelopathy. First described in 1958, this procedure aims to lessen pain while improving stability of the cervical spine by limiting motion or through removal of compressive lesions (2). Complications may include adjacent segment disease, postoperative dysphagia, postoperative hematoma, and recurrent laryngeal nerve palsy (3-5). Introduced in 1944, PCF is another commonly performed intervention (6). PCF avoids anterior-related complications and prevents fusion-related complications such as hardware failure, adjacent segment disease and pseudoarthrosis (7,8). Possible complications of PCF include higher infection risk compared to anterior procedures, iatrogenic instability, and higher revision rates at the index level due to inadequate decompression (9-11). Current literature remains conflicted regarding the superiority of one operation over the other. Proponents of ACDF highlight lower reoperation rates and higher rates of patient satisfaction (9,10). Supporters of PCF note shorter operations, lower costs, greater reduction of neck and arm pain, and preserved range of motion (11-15). Physicians frequently use systematic reviews, which allow for a comprehensive synthesis of current literature, as a factor in determining treatment options for patients. However, due to time constraints, physicians may only read the abstracts of studies and make clinical decisions based on the results and conclusions presented, without those corroborating findings with the full manuscript (16). As a result, physicians are left susceptible to misleading conclusions. For example, spin bias, interpreted as intentional or nonintentional misrepresentation of study findings, risks obscuring the true findings of systematic reviews and meta-analyses (17). Spin may lead to the implementation of interventions not truly supported by data and thus, could hinder the practice of evidence-based medicine. An examination of the literature reveals a concerning prevalence of spin across multiple orthopaedic treatment options and subspecialties. Spin prevalence in articles focused on spine, sports medicine, arthroplasty, and trauma ranged from 34.2% to 93.1% (18-28). The primary objective of this study is to review the prevalence of spin in systematic reviews comparing ACDF and PCF, and we hypothesize that spin bias will be present in at least 30% of systematic reviews. We present this article in accordance with the PRISMA reporting checklist (available at https://jss.amegroups.com/article/view/10.21037/jss-2026-1-0021/rc). Methods Search process A search was conducted on the following four databases—PubMed, Web of Science, Embase, and Scopus on June 25, 2025. The following search term was used to find systematic reviews regarding ACDF and PCF: ((systematic review OR meta-analysis) AND (anterior cervical discectomy and fusion OR ACDF OR anterior cervical fusion OR anterior cervical discectomy) AND ((minimally invasive OR cervical OR posterior) AND foraminotomy)). The results from all four databases were imported into Covidence, a tool for managing systematic reviews. Duplicates were automatically removed by Covidence. This review was not registered. Inclusion criteria and initial screening Articles were included if they were systematic reviews with or without meta-analysis, discussed treatment with surgical intervention, compared outcomes between ACDF and PCF, were available in English, and had the full text available. Any articles that did not follow these criteria were excluded. Initially, two independent reviewers screened the titles and abstracts of articles using Covidence. Minor disagreements were settled through discussion between reviewers. Included articles were then screened by the same two independent reviewers using the full manuscript for a final list of included studies. This final list was exported to Google Sheets for data collection. Article review Prior to data collection, two reviewers were trained on how to evaluate the presence of spin bias by a third author, who has experience in identifying spin within articles, using example articles. Two independent reviewers screened included articles for the nine most severe types of spin, previously defined by Yavchitz et al. (17). The nine most severe types of spin were chosen for this analysis in order to focus on spin with the greatest potential to guide surgeon decision making while being the least subjective. This methodology was adopted from and validated in prior medical literature examining spin bias (21-23). For each article, the reviewers recorded the presence or absence of each of these types of spin, year of publication, publishing journal, journal impact factor, citations per year, total citations, adherence to PRISMA guidelines within manuscript, performance of a meta-analysis, and sources of funding. Additionally, A Measurement Tool to Assess Systematic Reviews, version 2 (AMSTAR-2), was used to rate the quality of included articles from a scale of critically low to high (29). Data collection was conducted in duplicate using Google Sheets with each reviewer having a separate sheet to perform data collection independently. After completion, any disagreements in data collection were resolved through discussion between the two reviewers for a consolidated analysis. Statistical analysis Poisson regression modeling, given the data’s skewed nature, was utilized to assess for associations between number of types of spin bias and AMSTAR-2 rating, total number of citations, citations per year, impact factor of the publishing journal, reported funding, presence of PRISMA in the manuscript, and performance of a meta-analysis. Overdispersion in data was not found, incidence rate ratios (IRRs), and 95% confidence intervals were reported. This model was performed using IBM SPSS v30.0.0.0 (IBM corporation, Armonk, NY, USA). Then, a Firth’s penalized likelihood regression model was used to see if the presence or absence of spin could be predicted using AMSTAR-2 rating, total number of citations, citations per year, impact factor of the publishing journal, reported funding, presence of PRISMA in the manuscript, and performance of a meta-analysis. Odds ratios (ORs) and 95% confidence intervals were reported. A penalized regression was utilized instead of multivariable logistic regression due to variables exhibiting quasi-complete separation—AMSTAR-2 rating, reported funding, presence of PRISMA in the manuscript, and performance of a meta-analysis. This model was performed using RStudio V2023.06.1+524 (Posit Software, Boston, MA, USA). Statistically significant differences were defined by α=0.05. Interrater reliability was assessed using Cohen’s kappa statistic on Google Sheets. Results The initial search on June 25, 2025 resulted in 89 articles across the four databases. Of these, 44 duplicates were removed. After initial screening, 21 articles were advanced to full text screening. Of these, 12 articles met inclusion criteria (10,12,14,30-38). At full text screening, studies were excluded for wrong intervention (7,39-43), wrong study design (44), the article not being available in English (45), and the full text not readily available (46) (Figure 1). Study selection process. Article characteristics Included studies were published between 2009 and 2024 across 10 different journals. Journal impact factors ranged from 1.4 to 3.5 with an average of 2.63. Ten of 12 (83%) articles performed meta-analyses and 11 of 12 (92%) conformed to PRISMA guidelines. Five (42%) of the included studies reported receiving funding. The number of times each included article was cited by another article ranged from 0 to 74, with an average of 26.75. Citations per year ranged from 0 to 9.29, with an average of 4.68. Spin bias Spin bias was found in 11 of 12 (92%) of the included articles, with number of different spin types present in each article ranging from 0 to 2 (Figure 2). Five articles (42%) contained two different types of spin bias and six articles (50%) contained one type of spin bias. Type 9 was the most commonly observed spin type found in 5 of 12 (42%) articles. Type 5 spin was found in 4 (33%) articles, Type 3 in 3 (25%) articles, Type 1 in 2 (16%) articles, and Types 2 and 4 in 1 (8%) article each. Spin types 6, 7, and 8 were not found in any of the included articles (Table 1). Seven of the 11 articles with spin bias favored PCF, posterior endoscopic cervical foraminotomy (PECF), or minimally invasive-posterior cervical foraminotomy (MI-PCF), while 4 did not clearly favor either PCF or ACDF. The only article without spin bias favored MI-PCF, reporting shorter length of stay and decreased average cost, but comparable operative time, complication and reoperation rates, and patient-reported outcomes (38). Interrater reliability exhibited substantial agreement (kappa statistic =0.630). Total types of spin by article. Table 1Nine most severe types of spin per Yavchitz et al.Types of spin Articles with spin, n (%)1. Conclusion contains recommendations for clinical practice not supported by the findings2 (16.67)2. Title claims or suggests a beneficial effect of the experimental intervention not supported by the findings1 (8.33)3. Selective reporting of or overemphasis on efficacy outcomes or analysis favoring the beneficial effect of the experimental intervention3 (25.00)4. Conclusion claims safety based on non-statistically significant results with a wide confidence interval1 (8.33)5. Conclusion claims the beneficial effect of the experimental treatment despite high risk of bias in the primary studies4 (33.33)6. Selective reporting of or overemphasis on harm outcomes or analysis favoring the safety of the experimental intervention07. Conclusion extrapolates the review’s findings to a different intervention0 8. Conclusion extrapolates the review’s findings from a surrogate marker of a specific outcome to the global improvement of the disease0 9. Conclusion claims the beneficial effects of the experimental treatment despite reporting bias5 (41.67) Nine most severe types of spin per Yavchitz et al. AMSTAR-2 rating The AMSTAR-2 survey was used to examine the quality of included articles. Seven of 12 (58%) were rated critically low, and 5 of 12 (42%) were rated low. None of the included articles were given ratings of moderate or high (Table 2). Of note, all articles which received a critically low rating had at least 1 type of spin bias. Only 2 of 12 (17%) of included articles provided a list of excluded studies and justified their exclusions. Nine of 12 (75%) studies did not carry out adequate investigation of publication bias. Not providing a list of exclusions and not performing an analysis on publication bias were the 2 most common reasons precluding articles from receiving a rating above low. Two of 12 (17%) studies reported the sources of funding for their articles. All included studies clearly defined population, intervention, comparison, outcome (PICO) elements, defined review methods before the initiation of the literature search, and reported potential sources of conflicts of interest. Table 2Characteristics of studies which did and did not contain spinCharacteristicSpin (n=11)No spin (n=1)AMSTAR-2 rating High00 Moderate00 Low41 Critically low70Impact factor2.64±0.682.5Yearly citations4.84±2.953Data are presented as mean ± standard deviation or number. Characteristics of studies which did and did not contain spin Data are presented as mean ± standard deviation or number. Multivariable regression models Using our Poisson regression mode to analyze for associations with number of types of spin bias, we found no significant predictor for quantity of spin: AMSTAR-2 rating [IRR =2.48 (0.431–14.9), P=0.30], total number of citations [IRR =0.973 (0.914–1.04), P=0.39], citations per year [IRR =1.26 (0.723–2.18), P=0.41], impact factor [IRR =0.484 (0.091–2.58), P=0.39], reported funding [IRR =2.80 (0.215–36.5), P=0.43], presence of PRISMA [IRR =6.33 (0.083–481), P=0.40], and performance of meta-analysis [IRR =1.05 (0.093–11.8), P=0.97]. Similarly, AMSTAR-2 rating [OR =0.180 (1.61e−03 to 3.02), P=0.23], total number of citations [OR =0.959 (0.773–1.11), P=0.45], citations per year [OR =1.55 (0.607–8.83), P=0.36], impact factor [OR =0.343 (1.40e–3 to 13.7), P=0.48], reported funding [OR =0.129 (3.79e−4 to 7.07), P=0.30], presence of PRISMA [OR =0.141 (7.19e−7 to 1.22e+3), P=0.64], and performance of meta-analysis [OR =0.706 (2.84e−3 to 145), P=0.87] were not able to significantly predict the presence of spin bias, using a Firth’s logistic regression. Discussion This study assessed the prevalence of spin bias by type, quality, and potential predictors of spin in systematic reviews and meta-analyses comparing PCF and ACDF. There was a high prevalence of spin in 11 of 12 (92%) of articles, with a maximum of 2 types of spin per article. All studies received AMSTAR-2 ratings of either critically low (58%) or low (42%), with no studies deemed as high quality. Notably, all studies which received a critically low AMSTAR-2 rating had spin bias. However, no predictors were identified through multivariable logistic regression. To our knowledge, this is the first study to report the high prevalence of spin bias in systematic reviews and meta-analyses regarding the highly debated topic of PCF vs. ACDF, highlighting the importance of scrutinizing recommendations and improving the quality of research. This study mirrored prior work investigating spin within orthopaedic literature. Specifically in orthopaedic spine research, Taka et al. analyzed the prevalence of spin bias in systematic reviews and meta-analyses regarding the treatment of lumbar stenosis with spondylolisthesis through decompression alone or decompression with fusion. In 13 articles, the authors found spin bias in 6 of 13 (46.2%) articles (21). Avetisian et al. later found spin bias in 64% of systematic reviews comparing laminoplasty and laminectomy and fusion for degenerative cervical myelopathy (22). These trends have also been demonstrated in other orthopaedic subspecialties. In the first analysis of spin bias in orthopaedic trauma randomized controlled trial abstracts, Shepard et al. identified spin in 35 of 53 (66%) included articles (24). Jones et al. found that 25 of 73 (34.2%) of systematic review abstracts regarding the treatment of proximal humerus fractures contained spin bias (18). In 2024, Fathi et al. demonstrated at least 1 type of spin was present in 27 of 29 (93.1%) primary and secondary studies reporting the clinical outcomes of subacromial balloon spacers in the treatment of massive irreparable rotator cuff tears (20). Our findings of 92% spin bias among articles reflect the upper range of prior studies. Of the 11 articles containing spin bias, 7 articles supported PCF. Type 9 spin bias was the most common form of spin bias found in 5 of 12 (42%) articles (12,30,31,36,37). This occurs when the author’s conclusion claims the beneficial effect of the experimental treatment despite reporting bias within the article. For example, one article concluded that PCF is safe and effective compared to ACDF based on increased visual analog scale (VAS)-arm, VAS-neck, and neck disability index (NDI) scores; however, the article reports significant heterogeneity among the primary studies. In addition, the article reported a mean Methodological Index for Nonrandomized Studies (MINORS) score of 12.9 with 4 out of 9 primary studies receiving a score of 10, indicating poor study quality (37). Similarly, another study recommended MI-PCF while stating that each included study in the meta-analysis had a high risk of bias when assessing selection, performance, detection, attrition, and reporting biases (31). Conversely, another study claimed no clear superiority between MI-PCF and ACDF while also reporting a high risk of bias in primary studies (36). The next most common types of spin were Type 5 in 4 (33%) articles, Type 3 in 3 (25%) articles, and Type 1 in 2 (17%) articles. Type 5 spin occurs when the article’s conclusion claims a beneficial effect despite a high risk of bias in the primary studies. For example, one review stated that MI-PCF had the lowest rates of adverse effects within the abstract; however, within the main text, the author stated that one primary study comparing ACDF and MI-PCF was at high risk of bias due to its technique of randomization utilizing alternation in order of presentation (33). Type 3 spin is characterized by selective reporting or overemphasis on the efficacy outcomes of the experimental treatment. In one systematic review, PCF was favored due to a higher success rate in the abstract with no mention of clinical outcomes, while the main text reported improvements in VAS-neck, VAS-arm, and NDI scores with ACDF (30). Lastly, Type 1 spin occurs when the conclusion formulates recommendations for clinical practice that are not supported by the article’s findings. For instance, one systematic review without meta-analysis recommended PCF, even though one primary study showed increased good and excellent results with ACDF and a greater need for revision surgery with PCF (34). AMSTAR-2 rating, a commonly used tool to assess systematic reviews and meta-analyses, was used to analyze the quality of included articles. AMSTAR-2 is a 16-question checklist designed to critically appraise systematic reviews for methodological quality. In our study, 5 of 12 (42%) of articles received a rating of low, indicating they contained 1 critical flaw, while the remaining (58%) articles received a critically low rating, meaning they contained at least 2 critical flaws. The most common critical flaw was not providing a list of excluded studies with justification for exclusion, which occurred in 10 of 12 (83%) articles. The next most common critical flaw occurred when the systematic reviews did not analyze the risk of publication bias and discuss its impact on the results of the review, which occurred in 9 of 12 (75%) articles. These 2 critical flaws precluded most systematic reviews from receiving a rating above low. Other potential, more severe, critical flaws occur when the authors do not account for the risk of bias in primary studies using proper tools, such as the Newcastle-Ottawa Scale, or interpreting the risk of bias within the context of the results, which occurred in 3 of 12 (25%) and 4 of 12 (33%) articles, respectively. Simple tools such as the AMSTAR-2 checklist online ensure higher quality systematic reviews and meta-analyses, allowing for more confidence in a study’s conclusions (47). Our study failed to show an association between AMSTAR-2 rating and the presence of spin bias; however, it is important to note that all articles rated as critically low also contained a form of spin bias. This emphasizes the importance of critically analyzing systematic reviews for methodological quality. Surgeons often rely on an article’s abstract and systematic reviews as efficient ways to digest information on a specific topic; however, this makes them susceptible to spin bias when making clinical decisions. Even with the advent of abstract reporting guidelines such as consolidated standards of reporting trials (CONSORT) and PRISMA, Nascimento et al. demonstrated that articles across all healthcare areas presented poor reporting quality with over 5% of an abstract’s information not fully reported and more than 5% of inconsistency between the abstracts and full texts (48). Boutron et al. reported that physicians are more likely to favor a treatment if spin is present in the article (49). For example, the majority of articles in this study favored PCF even though some reported similar complication rates and an increased rate of reoperation (10,31,36). There are a few strategies proposed to avoid the influence of spin bias. Scrutinizing an article’s abstract and corroborating it with the results in the main text is an effective way to avoid spin in the conclusion (50). Raising awareness of spin bias and applying Yavchitz’s guide to critically appraise systematic reviews are crucial (17). Journal reviewers should be trained to identify spin bias, and journals should require authors adhere to strict reporting guidelines. Lastly, because some articles are only accessible as abstracts, the inability to review the full text may hinder critical appraisal. Ensuring that article data and full manuscripts are openly available and free of charge would allow spine surgeons to corroborate abstract conclusions with complete findings. This study has limitations worth noting. There is potential human error when assessing an article for spin bias types, due to its subjective nature. The small sample size of included articles may have caused the under- or over-reporting of spin on this specific topic. Four included studies also compared cervical disc replacement (CDR) to ACDF and PCF, which may limit the generalizability of our findings. With a larger sample size, our logistic regression may have shown potential predictors of presence of spin bias. However, this study was strengthened using methodology from prior spin studies analyzing the most objective types of spin and independent evaluation by two reviewers with a kappa statistic showing substantial agreement (21-23). Conclusions This study identified spin bias in 92% of systematic reviews and meta-analyses comparing ACDF and PCF. All included studies received either a low or critically low AMSTAR-2 rating, emphasizing the need for higher quality reviews. Efforts and strategies to mitigate spin bias in orthopaedic literature are imperative to collectively enhance the published literature. Supplementary The article’s supplementary files as Available at https://jss.amegroups.com/article/view/10.21037/jss-2026-1-0021/prf