Postoperative complications across different cervical spine surgical procedures: a systematic review and meta-analysis

Search strategy

A thorough literature search was conducted in the following databases: PubMed, Embase, Web of Science, Cochrane Central Register of Controlled Trials, and Cochrane Database of Systematic Reviews from project inception until January 2024. Search terms included medical subject headings (MeSH) and keywords related to cervical spine surgery and postoperative complications: (“cervical vertebrae” OR “cervical spine” OR “neck”) AND (surgery OR operation OR procedure OR fusion OR arthrodesis) AND (complication OR adverse event OR morbidity OR infection). The review included reference lists of relevant reviews and studies, with no language or publication limitations. Two reviewers independently evaluated all titles and abstracts to determine their inclusion criteria. Full texts of potential studies were evaluated in triplicate. Disagreements were resolved through a third reviewer. The selection process was recorded using a flow diagram based on the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) framework. The objective was to acquire a comprehensive collection of literature on postoperative complications across various cervical spine surgical procedures without language limitations.

Data collection & extraction

Data were gathered using a standard methodology by the study and a specific extraction form, which was evaluated on a range of studies. Two reviewers separately retrieved data from each article, including the first author, year of publication, study design, country of origin, sample size, duration of the follow-up, and patient demographics. Information about the surgical procedures was also extracted, including condition explanation, type of cervical procedure, number of levels treated, equipment utilized, and complications, and authors were contacted for ambiguous or missing data. Each study’s methodological quality was assessed using the MINORS criteria in order to identify any potential bias.

Inclusion and exclusion criteria

The review followed a clear inclusion and exclusion criteria, focusing on postoperative complications in adult patients after cervical disc replacement, anterior cervical corpectomy and fusion, posterior cervical fusion, and anterior cervical discectomy and fusion for inclusion. Studies had to report total complications or the number of patients experiencing specific complications. Exclusions included review papers, cadaveric and animal studies, and those with fewer than thirty patients. Only prospective cohort studies, randomized controlled trials, and retrospective case series were accepted. Studies without clear complication rates were excluded. No restrictions were applied to language, status, or publication date. Two reviewers independently screened titles, abstracts, and full texts, resolving any disagreements through discussion or with a third reviewer.

Data analysis

Data were analyzed using Review Manager 5. Pooled incidence rates and 95% CIs were calculated with random-effects models for each complication outcome reported in at least three studies, with I2 values greater than 50% indicating significant heterogeneity. Subgroup and sensitivity analyses were performed, and funnel plots assessed bias. A direct comparison was made between the complication rates for several cervical surgeries (ACDF, ACCF, PCF, PCF, disc replacement). Meta-regression analysis identified variables linked to complication rate fluctuations, and evidence quality for each outcome was evaluated using GRADE criteria. Two reviewers conducted the analysis independently, resolving disagreements with a third reviewer.

Risk bias assessment

A standardized tool assessed the included studies’ quality and danger of bias. The Methodological Index for Non-Randomized research (MINORS) criteria, a validated 12-item checklist, evaluated the methodological quality of the surgical studies, with 8 items for all studies and 4 additional items for comparative studies. A total score of 24 for non-comparative research and 32 for comparative studies is possible, with each item receiving a value of 0 (not reported), 1 (reported but inadequate), or 2 (reported and adequate). Two reviewers used MINORS to grade study quality, while the Cochrane Risk of Bias Tool was used for randomized controlled trials. Any disagreement among the reviewers were settled through conversation or a third reviewer.

Literature search result

A comprehensive literature search was conducted from the study’s inception to January 2024, using the previously mentioned databases without language or date restrictions. The search strategy, based on medical subject headings and keywords related to cervical spine surgery, yielded 8763 articles. Two reviewers screened the titles and abstracts according to the qualifying criteria, which required original data on postoperative complication rates following one or more of the surgical procedures of interest. Review articles, case reports, cadaveric or animal experiments, technical notes, studies with patients less than thirty were excluded.

After removing 4425 duplicates, 4338 unique records remained, with 212 articles suitable for full-text screening. Both reviewers retrieved the complete texts and evaluated them in depth independently. Furthermore, 190 studies were disregarded for various reasons Ultimately, 13 studies were included. Both reviewers analyzed the characteristics of them separately. From fifty to more than five thousand patients were included in the samples, with follow-up duration from six months to ten years. The funnel plots were used to evaluate the risk of bias showed no obvious asymmetry. Most studies showed that they adequately reported comparability, but less than half of them described consecutive enrollment or assessment blinding, as evaluated using MINORS. Figure 9 shows a PRISMA flow chart displaying identification of studies via databases and registers.

Data extraction was conducted separately and in triplicate, utilizing a standardized form, capturing trial characteristics, patients demographics, procedures specifics, complications assessment methods, outcomes, and ORs with CIs. Reviewers resolved disagreements through consensus. Cochrane Review Manager and Comprehensive Meta-Analysis were utilized to carry out the meta-analyses and risk of bias evaluation, respectively.

Neurological injury risk after anterior cervical spine surgery

Table 2 presents data from five studies on neurological impairment after anterior cervical spine surgery, comparing experimental and control groups. The largest group³ involved 14,583 patients in the experimental group and 11,774 in the control group, with 14,583 neurological incidents reported in the experimental group. Yielding an OR of 6959.50 (95% CI: 435.16 to 111304.42), indicating a significant increase in risk, though the exact effect size is unclear.

In contrast, Donk et al.¹² reported fewer incidents (350 total), with 35 patients in the experimental and 71 in the control group. Nonetheless, the experimental intervention showed a lower incidence of neurological injury, as indicated by the OR of 0.44 (95% CI: 0.28 to 0.68). Studies by Ghogawala et al.¹⁴ and Grosso et al.¹⁵ included 180–260 patients, found too low neurological incident rates (ORs near 1.0, confidence intervals crossing 1), identifying no significant difference.

The study by Li and Dai,²² with 441 patients overall, reported 62 neurological incidents in the experimental group (62 patients) and 415 in the control group (415 patients), showed the most disparity. This resulted in an OR close to 0, demonstrating a substantial protective impact. However, the large confidence interval (0.01 to 0.02) indicates that there is ambiguity because of the low occurrence rates. A high heterogeneity statistic (I²=100%, p<0.00001) confirmed significant variability between studies, with no overlap between individual CIs, as indicated by the I² value of 100% as shown in Figure 1. Overall, the effect test yielded a high significance (Z=34.08, p<0.00001), indicating that the pooled data detected an association between the risk of neurological injury from surgical interventions; however, the direction of this association varied amongst the studies.

Figure 1. Shows forest plot of predicting neurological injury risk after anterior cervical spine surgery.

Postoperative complications of cervical ossification surgery

Table 3 summarizes five trials comparing postoperative complication rates for cervical ossification surgeries. 26 patients were in Grosso et al.¹⁵ trial, with 2 complications in the experimental group and 8 in the control group. With an OR of 0.19 (95% CI: 0.04-0.99) indicating a significant reduction in complications with the experimental technique. Li and Dai²² having the greatest sample size of 443 patients, reported 5 problems in the group and 13 in the experimental group. Leading to an OR of 0.51 (95% CI: 0.26-1.00), which suggests a diminished but indeterminate impact.

Mostafa et al.²⁴ with only 30 patients, reported 30 complications in the control group and 6 in the experimental group, having an OR near 0 (95% CI: 0-0.08), suggesting a significant difference, and that the new technique had a strongly protective effect. In contrast, Noh et al.²⁶ with 100 patients in total, recorded 70 complications in the experimental group versus 30 in the control group, for an OR of 5.44 (95% CI: 2.97-9.97), favoring the control intervention. Sclafani and Kim³⁰ had 62 patients in total, with 5 complications in the control group and 4 in the experimental group, yielding an OR of 0.79 (95% CI: 0.20-3.08) suggesting no significant difference.

With an I² of 92% and a heterogeneity statistic of 51.91 (p<0.00001), substantial variations between trials was confirmed. Figure 2 illustrates the non-significant results of the overall effect test (Z=0.26, p=0.80). As a result, the pooled data from all the studies failed to find an association. The ORs varied, making it unclear whether experimental cervical ossification increased, reduced, or maintained the risk of postoperative problems. From the pooled data, no strong conclusions could be reached.

Figure 2. shows forest plot of predicting the Postoperative complications of cervical ossification surgery.

Postoperative pain after cervical spine surgery

Table 4 summarizes five studies quantifying postoperative pain in control and experimental surgery groups. The study conducted by Donk et al.¹² had 71 patients. On a scale of 0 to 10, the experimental group had mean scores of 6.9 and the control group of 7.1, yielding a non-significant weighted mean difference of -0.20 (95% CI: -1.20 to 0.80). Both Gao et al.¹³ and Ghogawala et al.¹⁴ measured pain in about 80 individuals using a 0-100 visual analog scale, and they discovered minimal differences (95% CI: 0). Grosso et al.,¹⁵ with 26 individuals, found slightly less pain in the experimental group (1.8 vs. 2.2), but the confidence interval (-0.97 to 0.17) included zero, indicating no significance.

Li and Dai²² conducted the largest study, which involved 443 patients. Mean scores were nearly identical: 77.9 (experimental) and 78.2 (control), yielding a negligible WMD of -0.30 (95% CI: -1.90 to 1.30). The p-value was 0.96, and heterogeneity was minimal (I²=0%, statistic=0.63). As seen in Figure 3, the overall effect test (Z=1.29, p=0.20) showed no significant difference in pain levels between the two groups. Although some studies suggested slightly less pain in the experimental surgeries, combined data showed no statistically or clinically significant difference. To make more definitive conclusions, closer CIs around WMD would be required.

Figure 3. Shows forest plot of comparison of postoperative pain after cervical spine surgery.

Neurological outcomes after cervical kyphosis correction

Table 5 summarizes five studies on neurological function post-surgery. Ajiboye et al.³ conducted the largest study with 14,583 patients, found a WMD of -0.08 (95% CI: -0.10 to -0.66) favoring the experimental group. Grosso et al.¹⁵ with 26 patients, found that the means of the control group were higher (2.6) than the experimental group (1.0). This suggests that the experimental operation had a greater advantage of -1.60 (CI: -2.09 to -1.11).

Li and Dai,²² with 443 patients, found that the experimental group was preferred by -3.20 (CI: -4.44 to -1.96) with control and experimental averages of 8.3 and 5.1, respectively. Mostafa et al.²⁴ reported a significant impact; with thirty patients, they used the same scale and discovered a mean of 69.2 for controls and 0.0 for the experimental group, demonstrating a significant benefit of -69.20 (CI: -74.00 to -64.40). Lastly, Sclafani and Kim³⁰ had 62 patients. The results showed averages of 2.6 and 2.2 for the treatment and control groups, respectively, for a moderate -0.40 WMD (CI: -0.84 to 0.04).

The heterogeneity test (Chi²=861.01, p<0.00001, I²=100%) confirmed that there was no overlap of CIs, as illustrated in Figure 4. At 8.08 and p<0.00001, the overall effect Z score was extremely significant. Even though the effects of the individual studies varied, a pooled analysis revealed compelling evidence that experimental cervical kyphosis corrective procedures produced better neurological results than control surgeries, though benefits likely vary by patient and procedure as shown in Figure 5.

Figure 4. Shows forest plot of comparison of neurological outcomes after cervical kyphosis correction.

Figure 5. Shows piechart about cervical pain and surgical procedures.

Risk of biased assessment

The methodological quality of the 13 papers on complications after cervical spine surgery was assessed using the risk of bias assessment. The majority lacked enough information on the techniques for allocation concealment and sequence generation. Utilizing computer-generated random numbers, only three research provided sufficient random sequence creation. The remaining ten papers either mentioned “randomization” without providing a mechanism or did not explain the sequence creation process. Allocation concealment was properly documented in just two studies; eleven studies reported no concealment at all. Two cohort studies’ non-randomized designs made them susceptible to selection bias by nature. Overall, 10 studies had a high risk of selection bias, and 3 were uncertain, as shown in Figure 6.

Figure 6. Shows risk of bias graph.

Blinding in cervical spine surgeries is impractical, making all 13 studies highly prone to performance bias. Eleven studies provided no blinding information, and just two reported using outcome assessors. The danger that knowledge of the intervention received may affect result measurement was introduced by the fact that outcome evaluation in the remaining trials mostly relied on medical records without specific blinding. Due to the possibility of ascertainment bias, there was a substantial probability of detection bias as shown in Figure 7.

Figure 7. Shows risk of bias summary.

Every study reported loss to follow-up, with the majority exceeding 80%, indicating low risk of attrition bias threshold. Two trials were missing 5-10% of the data, and the intention-to-treat method was used to analyze the remaining randomized individuals. Imputation techniques were used in 4 research to address missing outcome data, and in 2 studies, the findings were claimed to be unaffected. Attrition bias was low in 11 trials but unclear in two. Since published research makes up the majority of the research, it is possible that significant and positive results will be reported more often than null or negative results. It is difficult to identify publishing biases and selective reporting without access to research methods. The reporting bias risk for 12 trials was unclear due to the unavailability of procedures. One conference abstract high risk due to its insufficient outcomes reporting as shown in Figure 8. Concerns about modest study effects were also raised by the fact that most research couldn’t identify clinically significant effects.

Figure 8. Shows funnel plot showing asymmetry.

Figure 9. Shows PRISMA flow diagram.⁴³

Clinical heterogeneity was identified by contrasting surgical techniques, outcome definitions, and inclusion/exclusion criteria. Baseline imbalances increased bias risk in several studies, while two industry-funded studies raised conflict-of-interest concerns. For several of the included investigations, biases in selection, performance, and detection were very likely. Attrition bias was low risk, but there is less reporting, which lowers confidence. There were also significant reporting hazards and other biases. The review results’ level of confidence is reduced by these methodological flaws. Larger, better RCTs are necessary to provide more reliable results.