To ensure reproducibility and participant safety, training intensities and progressive overload mechanisms were meticulously standardized across all exercises.
The traditional resistance training program consisted of five hip and quadriceps strengthening exercises, including straight leg raise, single-leg semi-squat, seated leg press, hip abduction, and isometric knee extension. All participants performed three sets of each exercise three times per week for 8 weeks at 70–80% of one-repetition maximum (1RM). Dynamic exercises were performed for 8–10 repetitions, whereas the isometric knee extension was maintained for 10–12 s per repetition. Training progression was achieved by gradually reducing the rest interval between sets from 50 to 20 s over the intervention period, while maintaining a constant training intensity (Table 5).
To prevent exacerbating patellofemoral symptoms, 1RM was estimated indirectly every two weeks using a submaximal 10-repetition maximum (10RM) testing protocol. Participants performed a progressive submaximal strength test for each exercise, which was terminated when the participant could complete no more than 10 repetitions with proper technique, and the corresponding load was subsequently used to estimate 1RM using the Brzycki equation. The 1RM was calculated using the following equation:
$$ {\text{1RM}} = {\text{Weight}}\;{\text{Lifted}}/\left( {{1}.0{278} – 0.0{278} \times {\text{Repetitions}}\;{\text{Completed}}} \right) $$
(1)
High-precision load administration was achieved by rounding weights to the nearest 0.5 kg, utilizing fractional steel plates (0.25–1.0 kg) for the back squat, factoring the 45-kg baseline carriage weight into the 45-degree linear leg press total load, and placing rubberized micro-loading blocks (0.5–1.5 kg) on top of the weight stack for the selectorized leg extension machine.
For the knee stabilization protocol, progression and biomechanical parameters were strictly controlled to isolate specific therapeutic targets. The isometric component consisted of a seated hip adduction squeeze, performed in a high-backed chair with both hips and knees flexed at 90° and a rigid 15-cm foam block compressed between the medial femoral condyles. This movement targeted the hip adductor musculature (adductor longus, magnus, and brevis) to facilitate the neuromuscular co-activation of the vastus medialis oblique (VMO) for patellar stabilization. Furthermore, progressive overload for the step-down exercise was achieved biomechanically without introducing abrupt external weights (such as vests or dumbbells); instead, difficulty was incrementally advanced by increasing the stepping box height from 10 to 20 cm and extending the time-under-tension during the eccentric phase. All training variables were adjusted bi-weekly based on individual pain-free tolerance and updated 1RM assessments throughout the 8-week intervention.
In groups 2 & 3 the exercises included hip external rotation while lying on the side, hip external rotation while sitting on the table, step down, knee extension between 0 and 30 degrees, and isometric contraction with a ball (subjects hold a ball between their knee) as shown in Table 6. They performed all exercises with 30% of 1RM26.
In group 3 BFR was added. The 7-cm Katsu sports cuffs were employed to restrict blood flow. The arterial occlusion pressure was measured using a pneumatic cuff on the upper thigh, while standing. The participant then laid down supine, and the pedal pulse was located with an Acuson Cypress Portable Ultrasound. The cuff pressure was incremented until the pedal pulse could no longer be detected or a maximum of 250 mmHg, inflation rate 92 cm/s, and gradually released. The pressure upon pulse re-appearance was recorded as the arterial occlusion pressure3.
Blood flow restriction was initially applied at 50% of arterial occlusion pressure (AOP) and progressively increased to 80% AOP over the intervention period to facilitate gradual participant adaptation, improve tolerance to cuff compression, and maintain an effective training stimulus, while remaining within the recommended range for lower-limb BFR exercise13.
The dimensions of the cuff were suitable for encircling the upper portion of the participant’s thigh, ensuring that the edges of the cuff aligned seamlessly without any space in between3. The duration of muscle contraction and the speed at which the exercises were performed were prescribed as 2 s for the concentric phase, 0 s for the isometric phase, and 2 s for the eccentric phase. Throughout the rest between sets the cuff remained inflated providing continuous pressure. However, during the rest period between exercises, the cuff was deflated to allow for relaxation and recovery3. There was no need for any adjustments to the body or exercise technique. Independent of the BFR occlusion windows, a standardized lower-extremity flexibility program was administered during the warm-up and cool-down phases of each training session. Static stretching was systematically performed for the major muscle groups surrounding the knee joint, specifically targeting the quadriceps, hamstrings, gastrocnemius, and the tensor fasciae latae iliotibial (IT) band14. For each muscle group, participants completed 3 sets of passive static stretches held at the point of mild discomfort (without acute pain) for 30 s, followed by a 15-s relaxation period, under the direct supervision of a researcher to ensure proper alignment26. The act of stretching encompasses the muscles of the quadriceps, hamstrings, iliotibial band and plantar flexors, all of which were stretched within a comfortable and pain-free range26. In addition to performing stretching exercises, no other co-interventions were permitted. Participants were prohibited from engaging in any other form of exercise throughout the intervention period, and no home-based exercise routines were prescribed to them3 (Table 7).
Outcomes
The primary outcome was PFP reduction, measured by the visual analog scale (VAS, 0–10 cm). Secondary outcomes included proprioception (knee joint repositioning error test, degrees), and static balance (Stork test, seconds) and dynamic balance (Y balance test, centimeters). All measures were assessed pre-and post-intervention by blinded examiners.
Measurements
Pain intensity evaluation
The visual analog scale (VAS) was depicted as a 10-cm line, serving as a visual representation for comprehending the notion of pain. Patients were requested to rate their pain levels on this scale, ranging from zero to ten centimeters. Zero denoted the absence of pain, while the range of one to three indicated mild pain, four to six represented moderate pain, and seven to ten signified severe pain. The reliability of this scale for patients with patellofemoral pain (PFP), as reported in previous literature, ranges between 0.77 and 0.7927. Participants were instructed not to use analgesic or anti-inflammatory medications during the intervention period. Pain intensity during exercise sessions was monitored using the visual analog scale (VAS). If participants reported discomfort with a VAS score above 2 cm, exercise intensity was reduced by 20% as a safety precaution.
Static postural balance testing
Static postural stability was measured using the Stork Balance Test. Participants stood barefoot on their dominant leg, placed the foot of the non-weight-bearing leg against the medial aspect of the opposite knee, and placed both hands firmly on their hips. Upon raising the heel of the standing foot to enter a tiptoe position, the examiner initiated a digital stopwatch. The maximum duration (in seconds) the participant could maintain this exact posture was timed. The trial was terminated if the hands left the hips, the standing foot moved or swiveled, or the elevated foot lost contact with the support knee. The reported reliability of the test ranged from 0.93 to 0.9728.
Dynamic postural balance testing
Dynamic postural control was evaluated using the Y-Balance Test (YBT) protocol. Participants stood barefoot on a centralized platform and were required to maintain single-leg stability on the testing limb while using the contralateral limb to reach as far as possible in three standardized directions: anterior, posteromedial, and posterolateral. The maximum reach distance for each direction was recorded across three successful trials. To account for anthropometric differences, all raw reach distances were subsequently normalized to the participant’s true leg length (measured from the anterior superior iliac spine to the medial malleolus) and expressed as a percentage. If any errors occurred, such as movement of the supporting leg or balance disturbance, the participant was asked to repeat the test. The reported reliability of the test ranged from 0.99 to 1.029.
Knee joint proprioception assessment
Knee joint position sense was evaluated via the active joint repositioning error (JRE) test using a high-precision digital goniometer. Participants sat blindfolded on the edge of a standard examination table with their lower extremities hanging freely. The examiner passively moved the target limb to a predetermined target angle of 60° of knee flexion, maintained the position for 5 s to allow for somatosensory consolidation, and then returned the limb to the baseline resting position. Participants were then instructed to actively replicate the target angle. The absolute deviation (in degrees) between the target angle and the actively achieved angle was recorded over three trials, with smaller error values indicating superior proprioceptive acuity. This test was repeated three times and the angle repositioning error value was recorded. The reliability reported was between 0.994 to 0.99430.
Data analysis
Data normality and homogeneity were confirmed using Shapiro–Wilk and Levene’s tests (p > 0.05). A linear mixed model assessed group-by-time interactions for each outcome, with Bonferroni-corrected post-hoc tests for between-group comparisons. Exact p values, mean differences, and 95% confidence intervals were calculated. Cohen’s d was calculated using the pooled standard deviation formula for independent groups:
$$ {\text{d}} = \left( {{\text{M}}_{{1}} – {\text{M}}_{{2}} } \right)/{\text{SD}}\_{\text{pooled}}, $$
where
$$ {\text{SD}}\_{\text{pooled}} = \surd \left[ {\left( {{\text{SD}}_{{1}}^{{2}} + {\text{SD}}_{{2}}^{{2}} } \right)/{2}} \right]. $$
The 95% confidence intervals for Cohen’s d were calculated using the Hedgesand Olkin (1985) formula:
$$ {\text{SE}}\_{\text{d}} = \surd \left[ {\left( {{\text{n}}_{{1}} + {\text{n}}_{{2}} } \right)/\left( {{\text{n}}_{{1}} \times {\text{ n}}_{{2}} } \right) + {\text{d}}^{{2}} /\left( {{2} \times \left( {{\text{n}}_{{1}} + {\text{n}}_{{2}} } \right)} \right)} \right], $$
and
$$ {95}\% \,{\text{CI}} = {\text{d}} \pm \left( {{1}.{96} \times {\text{SE}}\_{\text{d}}} \right). $$
Significance was set at α = 0.05, and data analyzed using IBM SPSS (Version 28). Missing data (< 5%) were handled via maximum likelihood estimation within the linear mixed model.
