Comparison of KAFO and AFO for extension thrust pattern after stroke: a randomized N-of-1 trial

Abstract

[Purpose] Extension thrust pattern (ETP) during the stance phase is a common gait deviation after stroke; however, there is no consensus regarding the optimal orthosis, such as knee-ankle-foot orthosis (KAFO) or ankle-foot orthosis (AFO). This study aimed to compare the effects of KAFO and AFO on ETP using an N-of-1 trial design. [Participant and Methods] A 54-year-old male with right thalamic hemorrhage and post-stroke hemiplegia participated in this study. A prospective multiple-crossover N-of-1 trial was conducted. Five intervention periods (3 consecutive days each) of KAFO and AFO use were randomly assigned, with a 1-day washout period between interventions. Hip, knee, and ankle joint angles during mid-stance, 10-m walking time, and step count were measured. [Results] KAFO use significantly reduced ankle plantar flexion compared with AFO. A particularly large effect size was observed for the ankle dorsiflexion angle. Improvements in walking time and step count were also noted with KAFO. [Conclusion] KAFO was more effective than AFO in improving ETP in this patient. Objective evaluation using an N-of-1 trial may support more appropriate orthotic selection in clinical practice.

INTRODUCTION

Extension thrust pattern (ETP) at the knee during the mid-stance phase of walking is a major gait deviation after stroke, which results in reduced walking ability¹⁾. It occurs due to insufficient forward tilting of the lower leg along with inadequate hip extension moment (from the line of gravity passing in front of the knee), resulting in full (0°) or hyper-knee extension (>0°)²⁾. About 20–68% of ambulatory stroke survivors experience compensatory hyperextension at the knee^{2, 3)}. A systematic review of treatment options for ETP highlighted the use of several interventions, such as functional electrical therapy, proprioceptive training, and orthotic treatment⁴⁾. In terms of orthotic treatment, a study reported that a knee-ankle-foot orthosis (KAFO) was more effective than no orthosis in improving knee hyperextension, gait speed, and gait symmetry⁵⁾; furthermore, it has been reported that using a KAFO for training immediately after onset improves gait acquisition and activities of daily living (ADL) in stroke patients with moderate to severe motor paralysis^{6, 7)}. However, when comparing KAFO with barefoot therapy, there was limited evidence supporting the short-term efficacy of KAFO. Another study suggests that combining the use of ankle-foot orthosis (AFO) with balance exercises is effective for improving mobility⁸⁾, however, some reports indicate that the benefits of AFO use are limited to the time they are worn, and the quality of the evidence is reported to be low⁹⁾ Thus, there is a lack of definitive evidence establishing the superiority of KAFO or AFO for correcting ETP in stroke survivors.

N-of-1 trials are promising research tools that provide the highest level of evidence for validating the efficacy of a particular intervention tailored to a single study subject¹⁰⁾. Presently, this design is used in developing therapeutic drugs for rare diseases and in tailoring medical care to individual needs¹¹⁾. Since orthotic fabrication is also based on customization according to an individual’s anatomy and needs, an N-of-1 design allows for determining the most effective individualized orthotic selection. In this study, we used an N-of-1 design to compare the efficacy of a KAFO versus AFO in improving ETP with post-stroke hemiplegia. This approach could offer important information influencing the decision-making process for optimizing orthotic selection for a single case.

PARTICIPANTS AND METHODS

A 54-year-old male was diagnosed with right thalamic hemorrhage based on imaging findings from a head computed tomography conducted at an emergency hospital. He had a history of hypertension and was taking antihypertensive medications. He was treated conservatively and was admitted to an acute-care hospital for rehabilitation after stroke onset, where he underwent physical therapy (1 h long sessions for five times/week). Thirty days after transfer to a convalescent rehabilitation ward, the physical therapy sessions were increased to seven times a week.

Initial assessment after transfer to the convalescent rehabilitation ward showed severe motor paralysis of the left upper and lower limb on neurological examination. We used the Stroke Impairment Assessment Set (SIAS) to classify motor function (knee-mouth test: 1; finger function test: 1A; hip flexion test: 1; knee extension test: 2; and foot pat test: 1) and sensory function (tactile sensation: 3; position sense: 2). The total SIAS score was 46 points. In terms of balance, he scored 13 points on the Berg Balance Scale. He scored 0 on Functional Ambulation Categories; in terms of independence in ADL, he scored 44 points (motor: 25; cognitive: 19) on the Functional Independence Measure.

As a result of medical treatment and physical therapy during convalescent rehabilitation the patient was able to walk ~5 m without assistance; however, he developed ETP during the mid-stance phase (Fig. 1A), which necessitated corrective measures to obtain long-term gait stability before discharge and ensure social reintegration. Therefore, we decided to use an ankle foot orthosis to correct the gait deviation. We used a KAFO and an AFO, both of which resulted in gait improvement; however, the patient was unable to choose one of them because of insufficient information about which orthosis was more suitable for him.

Fig. 1.

Analysis of joint angles and gait training using ankle-foot orthosis and knee ankle-foot orthosis.

A: Example of kinematic analysis of the mid-stance phase using OpenPose. B, C: Patient performing walking exercises using (B) an ankle-foot orthosis and (C) a knee-ankle-foot orthosis.

The mean value of the mid-stance phase of three gait cycles for each of the five intervention periods was used for calculation.

The AFO was a bilateral metal upright orthosis with a double-stop ankle joint weighing 880 g (W-Klenzak joint; Keiai Co., Ltd., Tokyo, Japan) and having a fixed 5° ankle dorsiflexion angle to prevent ETP (Fig. 1B). We used a quick adjustable KAFO (Gait Innovation; Pacific Supply Co., Ltd., Osaka, Japan) (Fig. 1C). In the KAFO, the knee joint was a ring lock type, and the ankle joint had a hydraulic braking stop with plantar flexion braking and no dorsiflexion (weight=2,720 g). Both orthoses were rental devices and did not fit his foot size perfectly. The fitting process was supported by a physical therapist (PT).

We used a prospective multiple-crossover N-of-1 trial to compare the effectiveness of the two orthoses. The duration of the study protocol was based on N-of-1 trials guidelines¹²⁾ and CENT Statement¹³⁾. The KAFO intervention was designated as period A, and the AFO intervention was designated as period B. The intervention period was 3 consecutive days, with a 1-day washout period between interventions. The interventions were randomly conducted five times each, over a total intervention session of 39 days. During period A, walking exercises were carried out with assistance from behind; during period B, walking exercises were carried out by the patient himself using a cane. The patient was treated five times a week by the PT in charge; a second PT supervised the sessions on the other two days. The intervention order is shown in Fig. 2.

Fig. 2.

Timeline of the measurement procedures (progressing from left to right).

In period A, a knee-ankle-foot orthosis was applied and an ankle-foot orthosis was employed in period B. Phase wo indicates the washout phase.

Throughout the study period, the patient underwent regular physical therapy, including balance exercises and ADL training, in addition to gait exercises. During the washout period, the patient only underwent physical therapy. On the last day of each intervention phase, the patient underwent physical evaluation after the intervention was done; barefoot gait filming and photography were done using a tablet device (iPad mini4, iOS14.2; Apple Inc., Cupertino, CA, USA) mounted on a tripod.

The study outcomes were sagittal plane joint angles at mid-stance (hip flexion, knee flexion, and ankle dorsiflexion), 10 m walking time (seconds), and number of steps taken to walk 10 m (step). Joint kinematics were analyzed using a markerless motion capture system based on OpenPose (version 1.7.0) (Carnegie Mellon University, Pittsburgh, PA, USA), a deep learning-based human pose estimation algorithm. OpenPose estimates two-dimensional joint coordinates from video images without the need for reflective markers^14,15,16). The analysis was conducted using video recordings captured with a single camera under standardized conditions. The mean (± standard deviation) values from the mid-stance phases of three gait cycles were used for calculation. A comparison of the treatment effect during the intervention period was performed using the Wilcoxon signed-rank sum test; the significance level was set at 5%. Statistical analyses were performed using Stata (version 17.0; Stata Corp., College Station, TX, USA).

Comparisons of effects were examined by calculating the effect sizes (Cohen’s d) of the two groups. Cohen’s d is the difference between the means of the two groups divided by the standard deviation and is an indicator of how large the difference between the means is. The value obtained from this calculation is the standardized difference between the means for each group, and the criteria for the size of the effect size are d=0.2 to 0.4 for a small effect size, d=0.5 to 0.7 for a medium effect size and d=0.8 or greater for a large effect size¹⁷⁾.

The authors confirm that written informed consent has been obtained from the patient, who has also provided approval for the publication of this case report.The study protocol was approved by the Committee of Medical Ethics, Saitama Rehabilitation Center (Approval No.: R5-105); the study follows the principles of the Declaration of Helsinki and complies with the CARE guidelines for reporting.

RESULTS

The results were divided between KAFO and AFO (Table 1). The mean hip flexion angle was 10.9 ± 2.2° in period A and 17.1 ± 0.9° in period B. The mean knee flexion angle was −1.3 ± 1.8° in period A compared to −2.4 ± 0.7° in period B. The mean ankle dorsiflexion angle was −16.8 ± 3.8° in period A and −24.8 ± 1.5° in period B. The average 10-m walking time was 47.7 ± 9.1 sec in period A and 57.2 ± 14.5 sec in period B. Lastly, the average 10-m step count was 32.2 ± 3.5 steps in period A and 35.2 ± 3.6 steps in period B (Table 1). Comparison of the two intervention periods showed statistically significant differences in hip flexion angle, ankle dorsiflexion angle, 10-m walking time, and 10-m step count, with period A showing significantly smaller values (p<0.05). However, the knee flexion angle showed no statistically significant difference between the two intervention period (Table 1, Fig. 3). Comparison of effect sizes (Cohen’s d), hip flexion angle: d=3.69, knee flexion angle: d=0.81, ankle dorsiflexion angle: d=2.77, 10-m walking time: d=0.79, 10-m step count: d=0.85 (Table 1). A supplementary post hoc power analysis using G*Power (version 3.1) (Heinrich Heine University, Düsseldorf, Germany) showed limited statistical power for the knee joint angle (power=0.29), despite the observed large effect size.

Table 1. Outcome measurements and results of the statistical analysis.

	Session	KAFO	AFO	p-value	Cohen’s d
Hip flexion angle (°)	1	14.4	18.2	0.03*	3.69
	2	11.8	17.3
	3	10.1	16.5
	4	9.4	17.5
	5	8.9	15.8
	mean	10.9	17.1
	SD	2.2	0.9
Knee flexion angle (°)	1	−3.1	−3.3	0.08	0.81
	2	−2.5	−2.3
	3	−2.0	−2.2
	4	−0.3	−2.0
	5	0.8	−1.7
	mean	−1.3	−2.4
	SD	1.8	0.7
ankle dorsiflexion angle (°)	1	−21.1	−22.4	0.03*	2.77
	2	−21.0	−26.1
	3	−18.6	−25.7
	4	−15.2	−25.3
	5	−12.4	−24.7
	mean	−16.8	−24.8
	SD	3.8	1.5
10 m walking time (sec)	1	63.0	82.4	0.04*	0.79
	2	48.9	53.4
	3	43.3	55.1
	4	42.9	48.6
	5	40.2	46.4
	mean	47.7	57.2
	SD	9.1	14.5
10 m walking steps (step)	1	38.0	41.0	0.04*	0.85
	2	32.0	36.0
	3	32.0	34.0
	4	30.0	32.0
	5	29.0	33.0
	mean	32.2	35.2
	SD	3.5	3.6

The mean value of the mid-stance phase in the three gait cycles for each of the five trial periods was used for calculation. *p<0.05. KAFO: knee ankle foot orthosis; AFO: ankle foot orthosis; SD: standard deviation.

Fig. 3.

Comparison of study outcomes between intervention periods A and B.

(a) Hip joint angle; (b) Knee joint angle; (c) Ankle joint angle; (d) 10 m walking time; (e) 10 m walk step count

DISCUSSION

We found that using the KAFO resulted in a significant reduction in ankle plantar-flexion angles and an improvement in walking time and step count (Table 1, Fig. 3). It has been reported that KAFO use has limited effects on knee hyperextension and gait performance¹⁸⁾, which concurs with our results. Unlike previous studies, our study corroborated these effects by comparing and validating the intervention over a relatively long period (40 days).

Although no statistically significant difference was observed in the knee joint angle, the use of KAFO resulted in a biomechanical shift from knee hyperextension toward a more neutral alignment during mid-stance. In addition, large effect sizes were observed for both hip flexion and ankle dorsiflexion angles. These findings suggest that KAFO may contribute not only to knee stabilization but also to improved lower limb alignment through coordinated biomechanical effects at the hip and ankle joints. When evaluating the actual effects, the effect size (Cohen’s d) was larger in the hip flexion angle (d=3.69) and ankle dorsiflexion angle (d=2.77). Although the knee joint angle showed a relatively large effect size, the difference did not reach statistical significance. The post hoc power analysis indicated limited statistical power for the knee joint angle (power=0.29), suggesting that the nonsignificant result may be partly attributable to the small number of trials. Therefore, the effect of KAFO on knee joint control should be interpreted cautiously. This study primarily examined ETP of the knee joint during the mid-stance phase. The results suggest that the use of a KAFO contributed to improved lower limb alignment during the mid-stance phase. This is thought to be due to improved control of the ankle joint’s rocker function, which allowed for tibial anterior translation and suppressed excessive knee extension⁶⁾. Hence, a KAFO may be effective when ankle joint control alone cannot control the knee joint and causes ETP, as observed in our patient. Furthermore, the improvements in hip extension and ankle dorsiflexion movements may have contributed to increased step length and reduced 10-m walking time. However, achieving these results required more effort than standard clinical procedures. Planning the trial design in detail, ensuring the patient’s understanding and cooperation, and conducting more intricate measurements than those required in routine clinical settings complicated the implementation process.

The study used a prospective multiple-crossover N-of-1 design, implementing interventions A and B five times each for effective comparison. The insights offered by the study results provided us with more reliable outcomes and allowed the patient to select the more suitable orthosis owing to the reduced subjectivity compared to typical simple comparisons. Moreover, it is theorized that the accumulation of case reports using N-of-1 designs could contribute to evidence building^{10, 11)} and improve the quality of evidence supporting the selection of optimal orthotic therapy.

A limitation of this study is related to the accuracy of ankle joint angle measurements obtained using OpenPose-based two-dimensional analysis. Although OpenPose has been shown to provide valid estimates of lower limb kinematics^14,15,16), the accuracy of ankle joint angles may be affected by out-of-plane movements, such as foot rotation in the transverse plane. In particular, slight external or internal rotation of the foot may lead to measurement errors when estimating ankle dorsiflexion angles from a sagittal view. Therefore, the results should be interpreted with caution, especially for ankle kinematics. In addition, this study evaluated only the immediate effects of the orthoses, and further studies using different research designs are needed to clarify the long-term effects of orthotic gait training. Another limitation of this study is the potential influence of spontaneous recovery. As the patient was in the recovery phase after stroke, natural neurological and functional improvements may have contributed to the observed changes. Therefore, the effects observed in this study should be interpreted as a combination of orthotic intervention and spontaneous recovery. Although the use of an N-of-1 design enhanced the reliability of the results, these results are specific to this patient, necessitating caution regarding their widespread applicability.

Conflict of interest

The authors have no conflicts of interest to declare.