The influence of the number of screws and additional surgical procedures on outcome in hallux valgus treatment

A retrospective cohort study with a superiority design and an internal comparison group was performed. It was part of a Master’s thesis, and the results about different study aims are described elsewhere [17‐19]. Patient charts with radiological and clinical data from the Balgrist University Hospital in Switzerland between January 2004 and December 2013 were examined. Ethics approval was given by the local ethics committee (cantonal ethics committee Zurich 2015-0480), allowing this retrospective study with a large number of patients to be performed without the need for individually signed informed consent.

The surgical technique is described in more detail elsewhere [18]. Briefly, surgery was only performed for painful HV. ReveL was carried out with a short dorsal vertical and long plantar horizontal osteotomy (beginning dorsally 1 centimeter (cm) proximal to the metatarsophalangeal (MTP) joint, then changing direction to a proximally directed osteotomy in the middle of the dorsoplantar distance before exiting the plantar cortex about 4 cm proximal to the MTP joint; keeping the biplanar option of medial wedge resection when needing decrease of the distal metatarsal angle). Fixation was achieved with one or two 2.4 mm cortex screws directed dorsoplantarly, and postoperative care consisted of unrestricted ambulation in a postoperative shoe with a rigid sole for 4–6 weeks [20].

The inclusion criteria were primary ReveL for HV in adult patients. The local database (KISIM; CISTEC AG, Zurich, Switzerland) was searched for the key term “hallux valgus” (Fig. 1). Several exclusion criteria were applied. Eventually, 827 cases were investigated in this study. Patient charts were investigated utilizing the local database (KISIM) and imaging software (IMPAX 6.4.0.6010 and IMPAX Orthopaedic Tools, Agfa-Gevaert N.V. (Agfa), Mortsel, Belgium). The data were gathered anonymously with REDCap (version 6.11.5; Vanderbilt University, Nashville, TN, USA) and cleaned with Microsoft Excel (version 2010; Microsoft Corporation, Redmond, WA, USA) as well as Stata/IC (version 13.1; StataCorp LP, College Station, TX, USA).

The main exposure variable was the number of screws (one or two) used for fixation of the osteotomy during ReveL (Figs. 2 and 3). The secondary exposure variable was the use of additional surgical techniques for HV (no or yes), which consisted of tarsometatarsal arthrodesis (no or yes) and/or Akin osteotomy (no or yes). A tarsometatarsal arthrodesis was added if clinical examination revealed hypermobility of the first ray in the tarsometatarsal joint. An Akin osteotomy was performed if HV interphalangeus (lateral deviation of the distal phalanx in the interphalangeal joint) was found. As explained in Tables 2 and 3, other documented exposure variables, potential confounders, and/or effect modifiers were considered. The time periods were chosen according to the local changes in staff that occurred in the foot and ankle team in 2007 and consecutive changes in personal preferences for the use of one or two screws. The performance of an additional surgical procedure on the foot consisted of any other simultaneous surgery on the foot, such as the Hohmann technique (proximal phalangeal head resection of the lesser toes) for hammer foot deformity or Coughlin osteotomy (longitudinal diaphyseal osteotomy of the fifth metatarsal bone) for Bunionette deformity [21, 22].

The primary (categorical) outcome variable was the radiological recurrence of HV (HV angle (HVA) > 15°) at last radiological follow-up [23]. The secondary (categorical) clinical outcome variables were limited patient satisfaction, complication, revision surgery, and elective hardware removal.

The data were non-normally distributed according to the skewness and kurtosis test for normality, and medians (interquartile range (IQR)) are provided. Categorical data were cross-tabulated and assessed with the chi-squared or Fisher’s exact test. Odds ratios (OR) and 95% confidence intervals (CI) were calculated for each category. Confounders were considered to be factors associated with the radiological recurrence of HV and the number of screws after ReveL without being on the causal pathway between both variables. A priori, age and sex were defined as confounders since they commonly influence the prevalence of diseases, including HV [24‐27]. Preoperative HVA was also chosen as an a priori confounder since it represents the severity of HV and influences the recurrence of HV after surgery [28]. Further confounding was assumed when the crude and adjusted estimates differed by ≥ 10%. Effect modification was present when stratum-specific estimates differed according to the test for heterogeneity. Adjusted Mantel-Haenszel estimates for potential confounders and stratum-specific estimates for effect modifiers were calculated for the association between radiological recurrence of HV and the number of screws after ReveL. Likewise, secondary outcomes as well as the remaining exposure variables were assessed accordingly. To study the objectives of associations between the number of screws after ReveL and additional surgical techniques for HV with radiological recurrence of HV, logistic regression was used to obtain a causal model. Likelihood ratio tests were used to compare different models with and without interaction terms and linear trends. A Wald test assessed differences in the odds in different categories of a single variable.

Among others, bias was limited by calculation of a sample size (type II error), exclusion of cases with a reason for a specific number of screws (selection bias), hypothesis testing without knowledge of surgeons and patients (blinding, performance, and reporting bias), adjusting for confounders including time period as a surrogate for changes in staff and preferences for the number of screws (performance bias), and disclosure of missing values (selection bias). A clinically relevant difference in radiological recurrence of 10% would be important to detect. At a significance level of 5% and power of 80%, this required ≥ 199 cases per group. Statistical significance was assumed at a 5% level. Analyses were carried out with Stata/IC (StataCorp LP).

While two screws were more commonly used in the earlier time period (one screw, n = 53 [11.5%] vs two screws, n = 409 [88.5%], p < 0.001), one screw was more commonly used in the second period (one screw, n = 266 [76.4%] vs two screws, n = 82 [23.6%], p < 0.001).

More cases were treated with two screws after ReveL (n = 491 [60.6%]) compared to one screw (n = 319 [39.4%]) (Table 1). The stand-alone use of ReveL for HV (n = 514 [63.5%]) was more common than additional surgical techniques for HV (n = 296 [36.5%]). An additional Akin osteotomy (n = 289 [35.7%]) was more common than tarsometatarsal arthrodesis (21 [2.6%]). The majority of cases did not receive additional surgery on the remaining foot (n = 496 [61.2%]).

The logistic regression model demonstrated moderate evidence that radiological recurrence of HV was 45% less likely with the use of one screw for fixation after ReveL; irrespective of age, sex, preoperative HVA, additional surgical technique for HV, time period, and BMI (OR_adjusted = 0.55 [95% CI 0.30–0.98], p = 0.043) (Table 2). It also showed moderate evidence that radiological recurrence of HV was 162% more likely with the use of additional surgical technique for HV than without the use of additional surgical technique for HV (OR_adjusted = 2.62 [1.24–5.52], p = 0.011).

There was no evidence of an association of the secondary outcome variables (limited patient satisfaction, complication, revision surgery, and elective hardware removal) with the number of screws used for fixation after ReveL in the multivariate analysis (Table 3).

No evidence of associations of limited patient satisfaction with the number of screws used for fixation after ReveL (OR_adjusted = 0.64 [95% CI 0.37–1.08], p = 0.096), additional surgical technique for HV (OR_adjusted = 0.74 [95% CI 0.47–1.18], p = 0.202), and time period (OR_adjusted = 1.69 [95% CI 0.97–2.94], p = 0.062) were noted.

There was no evidence for associations of complications with the number of screws after ReveL (OR_adjusted = 0.50 [95% CI 0.20–1.24], p = 0.133), additional surgical technique for HV (OR_adjusted = 0.54 [95% CI 0.71–3.31], p = 0.272), and time period (OR_adjusted = 1.51 [95% CI 0.59–3.88], p = 0.391).

No evidence for associations of revision surgery with the number of screws after ReveL (OR_adjusted = 0.39 [95% CI 0.69–2.16], p = 0.279), additional surgical technique for HV (OR_adjusted = 1.23 [95% CI 0.42–3.62], p = 0.710), and time period (OR_adjusted = 1.18 [95% CI 0.26–5.47], p = 0.830) were observed.

There was no evidence of associations of elective hardware removal with the number of screws (OR_adjusted = 0.86 [95% CI 0.56–1.33], p = 0.503), additional surgical techniques for HV (OR_adjusted = 1.14 [95% CI 0.78–1.66], p = 0.492), and time period (OR_adjusted = 1.20 [95% CI 0.76–1.89], p = 0.492). The cumulative probability of complication and revision surgery did not differ between cases with one screw and those with two screws after ReveL (p = 0.902 and p = 0.453, respectively) (Fig. 4a, b). The cumulative probability of no elective hardware removal was lower in cases with one screw after ReveL than in those with two screws after ReveL (p = 0.001) (Fig. 4c). The median time to hardware removal was shorter in cases with one screw after ReveL (12.0 [IQR 8.0–19.0] months) than those with two screws after ReveL (14.5 [9.0–32.0] months).