1 Introduction

The knee is the most common location for degenerative joint disease, resulting in pain, functional impairment, and reduced quality of life (QoL).⁷ Arthroplasty, specifically total knee arthroplasty (TKA), remains the gold standard for the surgical management of end-stage osteoarthritis (OA).¹⁹ However, TKA has inherent risks and limitations particularly for younger patients and those who are more active.^20,30,39 In response to TKAs limitations, focus has shifted towards surgical alternatives that aim to maximize QoL in this particular demographic of patients with knee OA. Among others, high tibial osteotomy (HTO) has emerged as a viable joint preserving alternative, aimed at correcting malalignment, redistributing joint forces, and delaying the need for TKA.³⁴

The principle behind HTO for knee OA involves realigning the mechanical axis of the lower extremity to reduce load bearing through the diseased compartment and preserve the integrity of the joint.⁴ Several surgical techniques achieve this goal; notably, opening-wedge and closing-wedge, each bearing unique advantages, limitations, and outcomes.³⁶ Medial opening-wedge technique maintains a posterior slope while avoiding the proximal tibiofibular joint and peroneal nerve.³² Lateral closing-wedge technique offers inherent and immediate joint stability facilitating a faster recovery without the need for bone grafting.^4,9 Irrespective of surgical technique, management of knee OA with HTO is supported by randomized control trials (RCTs) 2. The outcomes and clinical significance reported in such RCTs, however, remains unclear.

RCTs represent the highest level of evidence in guiding management of knee OA.¹⁶ The P-value is used throughout the orthopedic literature to indicate whether a study finding is statistically significant or not. However, it has received criticism for neglecting important study design elements.⁴³ The fragility index (FI) was introduced to convey the vulnerability of a given reported outcome to the reversal of statistical significance.⁶ The purpose of this study was to evaluate the statistical fragility of RCTs assessing the efficacy of managing OA with HTO using the FI, rFI, and FQ metrics. Specifically, we evaluated the fragility of RCTs that focused on various HTO surgical techniques for knee OA. We hypothesized that study findings would be statistically fragile, especially those outcomes initially denoted by RCTs as statistically significant.

2 Methods

2.3

2.3 Fragility analysis

Using the recorded outcomes data, P values were confirmed via a Fisher's exact test to ensure accuracy. Fragility analysis was then performed by manipulating outcome events in a 2 x 2 contingency table until a reversal of significance was determined (Fig. 1). Statistical significance was set at P < 0.05.

Close

Fig. 1 Demonstration of statistical significance reversal using a 2x2 contingency table with a resulting fragility index (FI) = 1.

Export to PPT

For study outcomes reporting statistical significance, a FI was calculated. The FI is calculated as the number of outcome event manipulations that need to be made for significance to be lost for a study finding.³¹ The reverse fragility index (rFI) was similarly defined as the number of event manipulations to convert nonsignificant outcomes into statistically significant outcomes.¹² The fragility quotient (FQ) accounts for a comparative trial's sample size by dividing the FI by the sample size.³¹

Smaller FIs represent more fragile outcomes. For outcomes that were not statistically significant, an rFI was determined in a similar manner as FI, but with the intention of determining the number of reversals to achieve statistical significance. These values were then used to calculate an FQ, dividing the FI (or rFI) by the total number of study outcomes, with higher FQs representing more robust results. Subgroup analyses were performed for each of the outcome categories. Interquartile ranges (IQRs) were calculated and reported for both the FIs and FQs. Summary statistics are presented as median FI (IQR).

3 Results

The literature review encompassed 256 studies, from which 21 RCTs focusing on HTO were identified for detailed analysis (Fig. 2). Information on all 21 studies is included in Table 1. The journals most frequently referenced in the review were Knee Surgery, Sports Traumatology, Arthroscopy and International Orthopaedics. Specifically, four articles were sourced from Knee Surgery, Sports Traumatology, Arthroscopy, while three articles originated from International Orthopaedics. Among RCTs evaluating the same intervention comparison, there were 6 studies on open wedge versus closed wedge tibial osteotomies and 2 studies on high tibial osteotomy versus knee joint distraction. The other 13 studies included RCTs such as HTO with versus without microfracture augment, pain control in HTO, HTO versus unicompartmental arthroplasty for medial compartmental osteoarthritis, and open wedge HTO with versus without navigational assistance and patient specific instrumentation. Upon thorough assessment using the revised Cochrane risk-of-bias tool, two studies were found to have a “high risk” of bias, primarily due to concerns of results being selected on the basis of results.⁴² There were 5 studies that demonstrated “some concerns” for risk of bias and the remaining studies were found to be at “low risk”.

Close

Fig. 2 Preferred Reporting Items for Systematic Reviews and Meta-Analyses PRISMA flow diagram showing identification, screening, and inclusion of eligible articles from PubMed, Embase, and Medline.

Export to PPT

Across the 21 included RCTs, a total of 158 outcomes were evaluated. Of these, 20 outcomes were statistically significant, and 138 were not statistically significant (Table 2). The median FI across all extracted outcomes was 4,^3–6 suggesting that four event reversals could change the significance of study findings. Additionally, the associated median FQ was 0.083 (0.043–0.121), implying that the outcomes could shift in significance with 8.3 % of patients experiencing event reversals. When considering only statistically significant outcomes, the median FI was 3, and the corresponding FQ was 0.053. Conversely, non-significant outcomes exhibited a higher median rFI of 4, with an associated FQ of 0.083 (Table 2).

For RCTs examining the same intervention comparison, the 6 studies assessing open vs. closed wedge HTO exhibited the highest fragility, with a median of FI of 5 and FQ of 0.065. Additionally, we found that when looking specifically at limb length discrepancies among the open vs. closed wedge HTO RCTs, the FI was 1.5.

The two RCTs evaluating HTO vs. knee joint distraction (KJD) showed a median FI of 3, with an associated FQ of 0.047 (Table 3). However, when looking specifically at the outcome category hardware complications, we calculated an FI of 2.

The most fragile subgroups across all included RCTs were limb length discrepancy and hardware complications. They had an FI of 1.5 and 2, respectively. Complications and adverse events were the most common outcome type with 100 identified across the included RCTs, demonstrating a median FI of 4 and FQ of 0.066. Of all the subgroup analyses based on outcome category (Table 4), reoperations/intervention had the highest median FI of 9.5, and associated FQ of 0.149.

4 Discussion

In this study we used the FI, rFI, FQ to determine the robustness of outcomes reported in RCTs assessing management of osteoarthritis with HTO. In the current assessment of (RCTs) on high tibial osteotomies (HTO), it was observed that the overall median FI was 4 and the FQ was 0.083 (IQR 0.043–0.121). An FI of 4 indicates that reversing just 4 patient outcome events would be sufficient to alter the significance of the results. This illustrates the sensitivity of HTO RCT findings to small shifts in patient outcomes, a feature that is expected in small sample surgical trials. It is important to note that no universally acceptable thresholds exist for what constitutes a clinically meaningful FI or FQ, and these values should therefore be interpreted descriptively rather than as definitive markers of study quality. Given the sample size, an FQ of 0.083 infers that approximately 8 out of 100 patients would need to experience a different outcome to change the significance across the 158 total outcomes. Statistically significant findings had a median FI of 3 and an associated FQ of 0.053, indicating higher fragility in the HTO literature for findings with a statistically significant p-value.

These findings align with results from previous fragility analyses in the orthopedic literature.^2–11 A median FI of 3 for significant outcomes is comparable to studies that look at the fragility of femur fractures, radius fractures, achilles ruptures, fibula fracture, cartilage restoration, orbital fractures, and rotator cuff repair.^{21–24,27–29}

RCTs that report statistically significant outcomes help physicians to make informed decisions using objective data.⁴⁰ While P values provide some interpretation of data, they fail to take into account effect size, loss to follow-up, or sample size, and can be influenced by study designs.^3,44 These shortcomings can lead to fragile statistical findings and unintentional type I (alpha) errors where null hypotheses are rejected despite being true for a study outcome.³⁸ This kind of error can lead to unsubstantiated claims being made that lead to one intervention being recommended over another, when in actuality there is not convincing evidence to support it. Therefore, relying solely on P values may be inadequate. They should be supplemented by considering effect size, study design, and methodological integrity.

Fragility analyses are a relatively novel area of research, and produce values such as FI and FQ which help to adequately represent uncertainty in RCTs to clinicians.⁴¹ However, it is important to note that there is currently no established threshold for an optimal FI or FQ in the literature. The use of other clinical evaluation tools such as the minimal clinically important difference (MCID), substantial clinical benefit (SCB), and patient-acceptable symptomatic state (PASS), and maximum outcome improvement (MOI) could help create recommended or optimal FI and FQ thresholds for statistically and clinically significant outcomes.^26,35,37 Creating a recommended FI and FQ threshold could further standardize RCTs and help physicians gain a more comprehensive understanding of data. Until this recommended threshold is created the use of FI and FQ in determining the fragility of RCTs can provide clinicians with a more thorough understanding of significant and nonsignificant trial outcomes.

In our analysis by outcome categories across all included RCTs, limb length discrepancy and hardware complications were the two most fragile subgroups analyzed. Limb length discrepancy and hardware complication are of interest in HTO studies due to their potential impact on quality of life.^8,14 Limb length discrepancies following HTO can lead to back pain and gait abnormalities.¹⁴ Similarly, hardware complications, which encompass a variety of different issues such as irritation, symptomatic hardware, infection, non-union, delayed wound healing, can lead to compromised stability and prolonged recovery time.⁸ Studies report limb lengthening after open wedge HTO and shortening following closed wedge HTO.^1,15,17

Current literature generally favors open wedge HTOs over closed wedge HTOs. Factors that favor open wedge HTOs include superior control over the correction angle, sparing the proximal tibiofibular joint, and less severe complications.^8,13,25,33 In contrast, some studies have shown that closed-wedge HTOs produce more accurate corrections and less morbidity.² However, long term follow-up studies found no difference in clinical outcomes or radiographic alignment.^5,18 Using subgroup analysis, we compared open vs closed wedge HTOs. We reported that open vs closed wedge TO related outcomes had a median FI of 5. A study by Kim et al. found that close wedge HTOs caused fewer limb length discrepancies than open wedge HTOs. A separate study by TK Nerhus et al. found that open wedge HTOs led to fewer reoperations/interventions than closed wedge HTOs. Both of these results were statistically significant, but the FI for both studies was 1. Therefore the results should be interpreted with caution due to the subgroup analyses being underpowered. It also demonstrates how FI can be used to help interpret these findings. These results add to the current literature by suggesting that a median of five patients can reverse outcomes in studies related to open vs closed wedge HTOs.

Knee joint distraction (KJD) is a newer joint-preserving surgical technique. Studies have reported significant cartilage tissue repair and long-term clinical benefits with KJD.^10,11,45 Jansen et al.¹¹ concluded that KJD is a valid alternative to HTO based on two-year clinical, radiographic, and biochemical outcomes. Our study reported a median FI of 3 with respect to HTO vs knee joint distraction outcomes. These findings demonstrate that studies comparing outcomes of HTO and knee joint distractions are somewhat fragile in that 3 patients may be sufficient to produce outcome reversal.

When comparing hardware complications in HTO vs. KJD we calculated an FI of 2, Additionally, we compared limb lengthening discrepancy outcomes in open wedge HTO vs. closed wedge HTO and calculated an FI of 1.5. Post-op weight/ambulatory ability was another outcome we analyzed. For this outcome, all patients received HTO but then either delayed weight bearing activity or started it immediately after their procedure. The FI for Post-op weight/ambulatory ability was fragile with an FI of 3.5.

5 Conclusion

Our fragility analysis of RCTs evaluating HTO efficacy for knee OA revealed that reported outcomes are sensitive to small changes in patient events. These findings are consistent with current studies in orthopedic literature and indicate that changing a small percentage of outcomes may alter the significance of the study. This study does not provide direct clinical recommendations for HTO but instead evaluates the robustness of current RCTs. Incorporating FI, rFI, and FQ alongside traditional metrics in RCTs can provide clinicians with more comprehensive data to inform their clinical decision making.

Guardian patient consent

This study did not require patient or guardian consent since all of the data was obtained from publicly available websites.

Disclosures

Robert L. Parisien, MD.

Arthrex, Inc: Other financial or material support

The following individuals have no conflicts of interest or sources of support that require acknowledgement:, Reginald M. Brewster, Jamie Frost, Michaela E. Corvi^,Avanish Yendluri, John J. Corvi, Junho Song, Nikan K. Namiri, and David Kantrowitz.

Patient involvement statement

Not applicable.

Data sharing statement

Data are available in a public, open access repository. The data was obtained on publicly available websites.

Credit statement

Avanish Yendluri, Reginald M. Brewster, Jamie Frost: Conceptualization, Methodology, Software Reginald M. Brewster, Jamie Frost, Avanish Yendluri, Michaela Corvi:Data curation, Writing- Original draft preparation. Reginald M. Brewster, Jamie Frost, Michaela Corvi: Visualization, Investigation. Robert L. Parisien: Supervision John Corvi, Junho Song, Nikan K. Namiri, David E. Kantrowitz: Writing- Reviewing and Editing.

Ethical statement

This was not a human study and did not include human subjects.

Statement of financial disclosures

No Funding was received from external sources. I affirm that I have no financial affiliation (including research funding) or involvement with any commercial organization that has a direct financial interest in any matter included in this manuscript, except as disclosed and cited in the manuscript. Any other conflict of interest (i.e., personal associations or involvement as a director, officer, or expert witness) is also disclosed and cited in the manuscript. No.

Funding information

No Funding was received from external sources.