LAUSR

BACKGROUND The accurate measurement of pelvic tilt is critical in hip and spine surgery. A sagittal pelvic radiograph is most often used to measure pelvic tilt, but this radiograph is not always routinely obtained and does not always allow the measurement of pelvic tilt because of problems with image quality or patient characteristics (such as high BMI or the presence of a spinal deformity). Although a number of recent studies have explored the correlation between pelvic tilt and the sacro-femoral-pubic angle using AP radiographs (SFP method), which aimed to estimate pelvic tilt without a sagittal radiograph, disagreement remains about whether the SFP method is sufficiently valid and reproducible for clinical use. QUESTIONS/PURPOSES The purpose of this meta-analysis was to evaluate the correlation between SFP and pelvic tilt in the following groups: (1) overall cohort, (2) male and female cohort, and (3) skeletally mature and immature cohorts (young and adult groups, defined as patients older or younger than 20 years). Additionally, we assessed (4) the errors of SFP-estimated pelvic tilt angles and determined (5) measurement reproducibility using the intraclass correlation coefficient. METHODS This meta-analysis was reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines and registered in PROSPERO (record ID: CRD42022315673). PubMed, Embase, Cochrane, and Web of Science were screened in July 2022. The following keywords were used: sacral femoral pubic, sacro femoral pubic, or SFP. The exclusion criteria were nonresearch articles such as commentaries or letters and studies that only investigated relative pelvic tilt rather than absolute pelvic tilt. Although the included studies had different patient recruitment strategies, study quality-wise, they all used an adequate amount of radiographs for landmark annotation and applied a correlation analysis for the relationship between the SFP angle and pelvic tilt. Thus, no risk of bias was found. Participant differences were mitigated via subgroup and sensitivity analyses to remove outliers. Publication bias was assessed using the p value of a two-tailed Egger regression test for the asymmetry of funnel plots, as well as the Duval and Tweedie trim and fill method for potential missing publications to impute true correlations. The extracted correlation coefficients r were pooled using the Fisher Z transformation with a significance level of 0.05. Nine studies were included in the meta-analysis, totaling 1247 patients. Four studies were used in the sex-controlled subgroup analysis (312 male and 460 female patients), and all nine studies were included in the age-controlled subgroup analysis (627 adults and 620 young patients). Moreover, a sex-controlled subgroup analysis was conducted in two studies with only young cohorts (190 young male patients and 220 young female patients). RESULTS The overall pooled correlation coefficient between SFP and pelvic tilt was 0.61, with high interstudy heterogeneity (I2 = 76%); a correlation coefficient of 0.61 is too low for most clinical applications. The subgroup analysis showed that the female group had a higher correlation coefficient than the male group did (0.72 versus 0.65; p = 0.03), and the adult group had a higher correlation coefficient than the young group (0.70 versus 0.56; p < 0.01). Three studies reported erroneous information about the measured pelvic tilt and calculated pelvic tilt from the SFP angle. The mean absolute error was 4.6° ± 4.5°; in one study, 78% of patients (39 of 50) were within 5° of error, and in another study, the median absolute error was 5.8º, with the highest error at 28.8° (50 female Asian patients). The intrarater intraclass correlation coefficients ranged between 0.87 and 0.97 for the SFP angle and between 0.89 and 0.92 for the pelvic tilt angle, and the interrater intraclass correlation coefficients ranged between 0.84 and 1.00 for the SFP angle and 0.76 and 0.98 for the pelvic tilt angle. However, large confidence intervals were identified, suggesting considerable uncertainty in measurement at the individual radiograph level. CONCLUSION This meta-analysis of the best-available evidence on this topic found the SFP method to be unreliable to extrapolate sagittal pelvic tilt in any patient group, and it was especially unreliable in the young male group (defined as patients younger than age 20 years). Correlation coefficients generally were too low for clinical use, but we remind readers that even a high correlation coefficient does not alone justify clinical application of a metric such as this, unless further subgroup analyses find low error and low heterogeneity, which was not the case here. Further ethnicity-segregated subgroup analyses with age, sex, and diagnosis controls could be useful in the future to determine whether there are some subgroups in which the SFP method is useful. LEVEL OF EVIDENCE Level III, diagnostic study.