Strength muscle power and rapid force production assessment in knee osteoarthritis

Executive assessment

Muscle assessment is useful in knee osteoarthritis when the quantity, limb, contraction mode and apparatus are explicit. Maximal knee extensor force, knee flexor force, torque, bilateral leg-press load, mechanical power, rapid force production and submaximal force control describe related but distinct capacities. A clinically useful measurement system should preserve these distinctions instead of presenting them as interchangeable versions of “strength.”

The strongest immediate application is to describe impairment, guide exercise loading and monitor protocol-matched change. Stabilized handheld or fixed dynamometry can produce reproducible rankings, but their individual error is sometimes substantial. In a pre-TKA OA study, knee-extension smallest detectable differences ranged from approximately 19% to 31% between days, despite good-to-excellent ICCs; flexion error was larger. Conversely, a same-day study in 28 older women reported an extension MDC95 of 14 Nm with externally fixed handheld dynamometry. The different interval, fixation, muscle and sample explain why there is no universal handheld-dynamometry MDC for knee OA. [1, 2]

Mechanical power has meaningful concurrent associations with mobility and perceived disease burden. However, much of this evidence is cross-sectional. Accettura's knee-extensor power study found additional explained variance for walking and stairs beyond age, BMI and self-efficacy, but compared separate strength and power models rather than establishing power's incremental value after strength. Tevald's smaller study did directly examine additional information beyond strength and found it for bilateral leg-press power, not consistently for isolated knee-extension power. Task and apparatus matter. [3, 4]

Rapid force production requires even more explicit definitions. The knee OA literature includes peak first derivatives across brief pulses, early and late onset-based slopes, a 30–90% force-rise slope, peak-force/time-to-peak ratios and regression scaling factors across many submaximal pulses. These have different units, physiological interpretation and processing sensitivity. An attractive numerical value from one cannot calibrate another. [5–10]

A large longitudinal OAI RFD study is informative but limited. In OAI participants with or at risk of OA, higher 30–90% RFD was associated with lower odds of worsening self-reported WOMAC function over 36 months, but not worsening measured walking or repeated chair-stand performance. The pooled adjusted OR for the highest versus lowest tertile was 0.68 (95% CI 0.51–0.92) for WOMAC function. It does not establish an individual risk calculator, RFD superiority over maximum strength, or a treatment effect. [9]

For rehabtools, a small transparent family of measurement options is preferable to a generic strength age, universal symmetry target or recovery forecast. Record both limbs, raw units and body-size normalization; retain unsuccessful or painful tests; pair the observed change with a directly compatible error estimate; and distinguish detectable change from patient-important change. Strong correlation with a functional test is valuable construct evidence, but the strongest clinical claims require temporal prediction, validation and useful added information beyond ordinary clinical variables.

Scope and construct definitions

Population and disease stage

The primary population is adult symptomatic and/or radiographic knee OA without joint replacement. Clinical criteria and definite radiographic OA are separately labelled. Samples that include doubtful KL1 disease, isolated patellofemoral changes, previous injuries, or people merely at risk of OA should not silently become homogeneous established tibiofemoral OA cohorts. Disease stage, pain severity, alignment, obesity, age, sex, activity and bilateral symptoms can alter both force generation and task performance.

Severe OA awaiting TKA is included as a distinct preoperative group. It is not representative of all community OA, and measurements after surgery are reserved for the companion TKA review. A preoperative predictor of later surgical recovery likewise has a different endpoint from the progression of nonoperative OA. Data from mixed hip/knee OA, healthy comparators or postoperative arthroplasty may clarify methods but cannot supply unlabelled knee-specific thresholds.

Quantities that should remain separate

Maximal voluntary force is the greatest external force achieved under a stated task, commonly reported in newtons. Joint torque is force multiplied by the relevant perpendicular moment arm and is expressed in Nm. Recording force without the moment arm can make a longer lower leg appear weaker or stronger for reasons unrelated to the joint moment. Dividing force or torque by body mass changes the question to capacity relative to body size; it does not eliminate all confounding by adiposity or body dimensions.

Isometric testing holds the joint angle approximately fixed. Isokinetic testing constrains angular velocity during a moving contraction. A device called an isokinetic dynamometer can also be used in isometric or isotonic mode, so naming the machine alone is insufficient. A machine 1RM is the maximum successful load through the specified movement and range. Bilateral leg press combines multiple muscles and limbs and can conceal compensation by the less painful side.

Mechanical power is the rate of doing work: force multiplied by linear velocity, or torque multiplied by angular velocity, with watts as the unit. Peak power, average concentric power, stair-climb power and chair-rise estimates differ. Power depends on resistance, range of motion and shortening speed; the load producing peak power varies between protocols and people. During a strictly isometric contraction, external mechanical power is approximately zero even when force rises rapidly. RFD is therefore not mechanical power.

RFD or RTD is the change in force or torque per unit time. An average slope from onset to 50 ms is distinct from the slope between 100 and 200 ms, the maximum moving-window slope, or the highest numerical derivative anywhere on a trace. Dividing maximum force by time to peak is a whole-rise ratio, not an early-onset slope. A scaling factor relates the peak derivative to peak force across many submaximal pulses and has units of inverse seconds. Force relaxation is another construct. Steadiness concerns variability while maintaining a target and must remain separate from both maximal and explosive capacity.

Search approach and source accessibility

Searches completed on 2 October 2026 combined knee OA with strength, power or RFD and measurement or prognostic terms, with an explicit publication-date cutoff. The dated PubMed query yielded 675 records, verified against the official NCBI records. A complementary Scopus query used dynamometry, muscle power, force or torque development, steadiness, maximal strength or quadriceps strength with measurement and prognostic terms and yielded 198 distinct records.

Complete articles were examined for key handheld, fixed-device, power, RFD, OAI and MOST studies. Original tables were inspected for Accettura’s power study, Chopp-Hurley and Hu to resolve numeric details. Kean’s strength and activation error study, Skou’s older handheld reliability paper, Holm’s power-rig reliability report, Luc-Harkey’s onset-RTD paper and several older longitudinal studies remain abstract-only. Their unobserved methods and tables are not inferred from titles or secondary discussion. The bibliography identifies source access and distinguishes these limitations from complete article and table access.

Maximal voluntary strength

Standardization determines what the result means

Knee and hip angle, axis alignment, contact point, pelvic/trunk restraint, ankle position, gravity correction and permitted bracing change the mechanical task. A device secured by a belt, a dynamometer manually resisted by the examiner, a tension device fixed to a door and a rigid laboratory dynamometer are different systems. “Handheld” can describe the sensor rather than the actual resistance arrangement.

Instructions should match the purpose. Gradually building force and holding a plateau is appropriate for certain maximal-strength protocols; it cannot be used to interpret explosive-onset RFD. Conversely, an instruction to contract as fast and hard as possible combines two objectives and may require separate trials if the application needs both reliable peak force and early RFD. Familiarization, rest, feedback and the choice of the maximum versus mean of several contractions all influence results.

Pain during exertion may reduce voluntary drive, while fear, discomfort at the pad and loss of fixation can limit the force transmitted to the device. Low force does not prove muscle atrophy. Voluntary activation assessed with an interpolated twitch provides different information, but its result depends on stimulation, muscle potentiation, stimulus timing and contraction adequacy. A normal-looking force ratio between knees can conceal bilateral impairment.

Direct clinical force measurement evidence

Fransen and colleagues compared isometric force in 113 OA patients with published age- and sex-matched reference data. Patients aged 60–79 generated approximately 40–53% of reference knee-extensor force and 35–46% of flexor force. One-week ICCs ranged from 0.79 to 0.95. This supports the clinical relevance of measuring both muscle groups, but historical normative comparisons and incompletely verified protocol details do not establish an individualized expected value or prognostic threshold. [11]

The 2011 modified-HHD study is particularly useful for distinguishing relative from absolute reliability. Thirty-two patients awaiting TKA contributed interrater data, and 13 contributed between-day intrarater data. Patients needed 90° knee flexion and were excluded when pain interfered with testing. The hip and knee were positioned at 90°, straps supported the sensor, extension contact was 5 cm proximal to the medial malleolus, and the mean of three measured contractions followed one familiarization. Force was normalized to body mass in N/kg. Participants built force over two seconds and held three seconds, explicitly avoiding explosive contraction. [1]

For affected-leg extension, interrater ICC was 0.96 (95% CI 0.91–0.98), SEM 0.21 N/kg and SDD95 0.58 N/kg, or 21.7% of the mean. Between-day affected-extension SDD was 0.84 N/kg for examiner A and 0.48 N/kg for examiner B, corresponding to 31.4% and 19.0%. Flexion between-day errors were substantially larger, reaching 57.5% for one condition. The retest subgroup was small, the interval ranged from 2 to 27 days and examiner order was fixed. The study also excluded a participant as an outlier. These factors limit a universal “HHD reliable/unreliable” verdict; the central lesson is that high ICC did not ensure sensitivity to small individual changes.

Chopp-Hurley and colleagues studied 28 women with clinical knee OA, mean age 66 years. A strap-and-suction fixation system held the HHD perpendicular to a metal door, with knee flexion at 90° and a matched distal cuff position. The mean of three peak contractions was tested twice on the same day, with approximately ten minutes between sets and randomized device/exertion order. HHD force was converted to torque using the measured moment arm. This is strong evidence for that same-session fixed arrangement, not for an examiner-held device between clinic visits. [2]

Original Table 2 reports HHD extension ICC 0.95 (99% CI 0.90–0.98), SEM 6.8% and MDC95 14.0 Nm. For flexion the corresponding values were 0.83 (99% CI 0.68–0.92), 9.5% and 11.2 Nm. The fixed laboratory device had extension MDC95 15.9 Nm and flexion 7.6 Nm. Confidence intervals are labelled 99%, an easily missed distinction. All testing occurred on one day and assessors were not blinded to measured values.

Cross-device agreement was less reassuring than repeatability. The HHD generally underestimated extension torque, with mean absolute difference 17.1 Nm, and the agreement ICC was 0.76 with a wide interval crossing zero for extension. Larger torques produced larger discrepancies. A fitted conversion regression with r² approximately 0.82 is a sample-derived relationship, not an externally validated universal conversion between devices. In a longitudinal record, retain the same apparatus rather than “correcting” readings with that equation.

Fixed dynamometry and voluntary activation error

Kean's 20-patient one-week study reported ICCs of 0.93–0.98 for isokinetic strength, isometric strength and interpolated-twitch voluntary activation. Reported SEMs were 14.57 Nm, 10.76 Nm and 2.84 percentage points, with MDCs 33.90 Nm, 25.02 Nm and 6.60 percentage points. The between-session mean differences were small relative to those individual change limits. [12]

The confidence convention is consequential. The published abstract does not label its MDC confidence level; dividing each reported MDC by its SEM gives approximately 2.33, compatible with 1.645×√2 rather than the 1.96×√2 multiplier for MDC95. Brisson's later paper explicitly describes Kean's isometric value as MDC90. Until the complete original method is checked, the values should be labelled “reported MDC” rather than silently advertised as 95% limits. They are detectable-change estimates, not anchor-based clinically important improvements.

Skou's 20-person, one-week HHD study offers a complementary warning. ICCs were 0.78–0.91 using the highest examination and 0.86–0.94 using the mean of the three highest. Yet reported 95% limits-of-agreement percentages were approximately 38–47% for the highest and 40–53% for the mean. Averaging improved relative reliability without uniformly reducing proportional agreement limits. Full muscle- and side-specific tables remain necessary before applying those ranges to a patient. [13]

Newer devices do not eliminate measurement error

Sahu and colleagues' 2024 device locks a weight stack and measures transmitted force with a load cell. It removes examiner resistance and offers a potentially practical route to stabilized isometric testing. However, only 15 of 44 participants had mild-to-moderate knee OA, and their mean age was 37.6 years. Their low pain and age make them a different target population from typical older symptomatic or preoperative OA. [14]

The pooled same-day intrarater ICC was approximately 0.97, but the OA subgroup's between-day interrater results were less favourable: ICC 0.78 (95% CI 0.416–0.919), SEM 81 N and MDD95 224 N for the new device. HHD had ICC 0.788 (0.402–0.924), SEM 35 N and MDD95 97 N. The paper reports 51.9% and 26.3% respectively, but percentage denominators vary or are unclear; 51.9% corresponds approximately to the second-day new-device mean rather than the two-day average.

Table 2 contains additional inconsistencies in the OA within-day SEM/MDD values and percentages. Its reported pooled SD and ICC do not reproduce the printed SEM using the stated conventional formula, and percentages do not reproduce the listed means. Therefore the tempting smaller 80-N same-day OA threshold is not adopted here as an uncontested clinical rule. Concurrent force-curve correlation of 0.96 with the laboratory device does not validate onset detection or high-frequency RFD, especially when the study's primary target is peak strength. Failure to find a significant mean difference between devices is also not proof of equivalence.

Table 1 Selected strength and power error estimates

Values retain the study’s original task, units, interval and confidence level. These are not interchangeable knee OA thresholds.

Muscle power and task specific capacity

Direct mechanical power

Accettura and colleagues tested 55 independently ambulatory adults with clinical OA, predominantly women, with 54 contributing the power-model analysis. Peak isometric strength was measured at 60° knee flexion. Isotonic knee extension–flexion power was assessed at 25%, 50% and 75% of MVIC torque, with ten consecutive repetitions; the peak from repetitions three to seven was extracted. Increasing numbers could not complete the higher loads: one at 25%, 11 at 50% and 24 at 75%. The analyzed power therefore used 25% MVIC. Peak knee-extensor power and torque were normalized to body mass (W/kg and Nm/kg, respectively); the regression relationships apply to these normalized quantities. This feasibility pattern matters as much as the association coefficients. [3]

Original Table 2 shows that adding power to age, BMI and self-efficacy increased explained variance from 0.33 to 0.39 for six-minute walking, 0.52 to 0.60 for stair ascent and 0.44 to 0.47 for descent. Power terms were significant, while strength terms in separate covariate-adjusted models were not. These are modest, task-specific concurrent additions. Power was not added after strength in the same model, so the result should not be described as demonstrating independent incremental prediction beyond maximal strength. Neither the variance explained nor the regression slope describes the within-person improvement expected after power training.

Reid and colleagues studied baseline data from 190 symptomatic radiographic OA participants in a trial. Bilateral pneumatic leg-press 1RM was followed by five maximum-velocity attempts at 40% and 70% 1RM, retaining the highest power and corresponding velocity. Age, sex, race, height, weight, medication count and depressive symptoms were included in separate adjusted models. Power at 70% 1RM was associated with WOMAC pain and SF-36 physical component score; the latter coefficient was 0.013 points/W, SE 0.004, p=0.003. Strength was not independently associated with those outcomes in the same way. These are cross-sectional associations rather than evidence that power causes pain or forecasts later quality of life. [16]

The trial sample also illustrates missingness: 14 eligible randomized participants did not complete strength/power testing, including three who felt unsafe, two who refused and one whose abdominal obesity prevented the protocol. Analyses of the 190 completers do not establish feasibility for all 204 randomized people. Four included cases had KL0 tibiofemoral radiographs but definite patellofemoral osteophytes, showing why radiographic definitions should be reported rather than simplified.

Tevald's 40-person exploratory study directly compared bilateral leg press and unilateral knee extension. After covariates and strength, leg-press power explained additional variance in functional tests, whereas knee-extensor power did not consistently do so. The abstract reports leg-press power models explaining 44–57% of variance compared with 24–34% for involved-side knee power and 28–48% for the other side. Because full resistance, adjustment and endpoint-specific tables were not obtained, these ranges support the task-specific conclusion rather than a numerical clinical conversion. [4]

Reliability of power

Holm's 40-person knee OA reliability study tested a Nottingham Power Rig and handheld knee-extensor strength across a three-day interval. All reported ICCs were at least 0.97, but 95% agreement limits were ±32.3 W (±22%) for leg-extensor power and ±22.7 Nm (±24%) for strength. These values show why “excellent reliability” is not synonymous with detecting a small training gain in one patient. Power-rig output represents the complete leg-extension task and should not inherit an isolated knee-dynamometer MIC or be assumed equal to bilateral leg-press power. [15]

Brisson and colleagues analyzed 46 OA participants measured at baseline, six months and 24 months. Isometric quadriceps strength ICCs were 0.91–0.93 and isotonic power ICCs 0.84–0.88; the protocol used a 25%-MVIC resistance for repeated knee movement. This is useful evidence about long-term stability in an observational sample. It is not a pure short-term error experiment: real disease, activity, pain and comorbidity changes can contribute to within-person variance. The authors acknowledge this. An MDC from these long intervals should not automatically be used as the minimum change required to show a short rehabilitation treatment is effective. Group treatment effects and individual detectable change are separate questions. [17]

Functional power estimates

Chair-rise and stair-climb power estimates are useful functional summaries when their equations and assumptions are explicit. They combine body mass, displacement and time, with assumptions about the moving mass, vertical distance and movement phase. They do not isolate knee extensors or directly measure joint power. A faster ascent may reflect confidence, balance, handrail use or a changed movement strategy; a bilateral rise can conceal unloading of the painful knee.

Recent knee OA studies relating sit-to-stand or stair power to clinical performance are cross-sectional, even when “predicts” appears in the title or they use baseline data from a randomized trial. They support concurrent construct relationships and potential practical measurement choices, not future-outcome forecasts. Full treatment-response and chair-height interpretation belongs in the sit-to-stand report, with the power quantity clearly labelled as estimated rather than directly measured when appropriate. The recovered Jørgensen article/PDF text appears to omit gravitational acceleration while defining mass in kilograms and output in watts. The literal extracted expression is dimensionally inconsistent, but the rendered equation has not been independently recertified and text extraction may omit symbols. The actual calculation is unknown; withhold equation implementation pending clarification without declaring all study results erroneous. [18, 19]

Rapid force production and force control

Why acquisition and processing cannot be optional metadata

An early RFD interval may contain only a handful of samples in a low-rate device. Onset choice, baseline noise, preload, strap compliance, filtering and the differentiation algorithm can materially change the slope. A fixed absolute onset threshold selects a different proportion of capacity in a weak versus strong participant. Normalizing RFD to MVC answers how rapidly available capacity is expressed, but can make a weak person appear relatively fast when the denominator is small. Preserve the absolute result as well as any normalized version.

A maximum derivative is noise sensitive and depends on the length of the searched interval. A best 100-ms window can occur well after contraction onset. A slope between 30% and 90% force excludes the initial rise and depends on reaching those fractions of the selected peak. A mean rise to peak depends heavily on contraction duration. These are all legitimate defined variables, but they cannot be averaged in a common RFD meta-value or treated as having the same error threshold.

Peak derivative during brief pulses

Winters and Rudolph compared 26 medial-OA participants with 23 controls. Subjects sat with hips near 90° and knees near 70° in a strapped chair. A rigid padded cuff minimized material compression. Three maximal contractions established MVIC, followed by multiple rapid small, medium and large force pulses without visual targets. The study retained the highest first derivative across pulses and the force level of the pulse at which it occurred. This differs fundamentally from measuring the first 200 ms of an MVIC. [5]

Highest peak RFD did not differ between groups: approximately 983 versus 1,000 N/s, p=0.763. Force at highest peak RFD was lower in OA, 64.0% versus 71.1% MVIC, p=0.008. MVIC itself was not significantly different in this sample. The force-at-peak measure related to current KOOS-ADL and aspects of knee power during walking; peak RFD did not explain additional stair-climb performance after strength. The null findings are central to the interpretation, not exceptions to an assumed universal RFD deficit.

The paper also contains reporting inconsistencies: KL category counts sum to 27 despite 26 OA participants, and some hierarchical-regression R, R², change and p values do not reconcile cleanly. Accordingly, the robust interpretation is an exploratory, task-specific association of force-pulse characteristics with current function. The exact regression table should not be converted into a calculator. Speculation about falls prevention or slowing structural progression was not tested prospectively.

Onset based early and late torque development

Luc-Harkey and colleagues assessed 76 people with symptomatic radiographic tibiofemoral OA. They separately quantified early 0–50-ms RTD, late 100–200-ms RTD and overall peak RTD, including involved-side and bilateral-average values. Greater bilateral-average late RTD explained an additional 20% of fast-walking variance and 11% of stair-climb variance after quadriceps strength. Involved-side associations were weaker, and none of the RTD measures related significantly to WOMAC function. [6]

Despite a “prognosis” evidence label in the abstract, this was a controlled laboratory cross-sectional design. Its results concern current task performance. The value of a bilateral average also does not imply that the contralateral limb is healthy; it may simply represent overall lower-limb capacity more accurately for a bilateral functional task. Full sampling, onset, filtering and precision details remain required before the particular RTD procedure can be reproduced or its change interpreted clinically.

Suzuki and colleagues compared 58 mild/doubtful-to-mild cases (KL1–2) with only eight KL3–4 cases, mostly older women. A belt-fixed HHD at 90° knee flexion recorded force at 1,000 Hz; measured moment arms converted force to torque. Two five-second contractions were attempted as fast and hard as possible. Onset was defined as torque reaching 4 Nm above baseline, and the first-200-ms slope was normalized to body mass and MVC. The more painful knee was analyzed. [7]

Severe cases had lower normalized RFD, while maximum torque did not differ significantly. The reported standardized effect for RFD was −1.07 (95% CI −1.83 to −0.30), versus −0.22 (−0.96 to 0.52) for maximal strength; adjustment for age, sex and pain retained the RFD difference. However, eight severe cases, post hoc adjustment, no prior power calculation and a cross-sectional design preclude concluding that low RFD predicts progression. The reported normalized units are unconventional and require careful equation verification before implementation. The early 0–100-ms comparison was not significant in the supplementary analysis described in the discussion. A positive 200-ms result should not be generalized to every RFD interval.

Scaling of contraction and relaxation

Šarabon and colleagues studied 24 KL2–3 OA patients and 24 controls. In a fixed dynamometer at 60° knee flexion, separate maximal and explosive contractions preceded approximately 30 pulses at each submaximal target. Controls performed 20%, 40%, 60% and 80% of maximum torque; OA participants performed only the lower three levels to reduce pain. Signals were sampled at 1,000 Hz and filtered at 5 Hz. Regression slopes related peak RTD and peak torque, or peak relaxation rate and peak torque, across pulses; the associated r² assessed linearity/consistency. [8]

Comparison of the reduced versus standard target ranges was performed in healthy controls, with ICCs above 0.77 and CVs below 10%. This supports agreement of two calculation protocols in those controls, not between-day reliability in painful OA. In the OA-control comparison, maximal torque was lower, but peak RTD and RTD scaling factor were not significantly different. Relaxation scaling and pulse consistency showed clearer differences. The affected-versus-unaffected comparison did not demonstrate a general scaling-factor deficit limited to the symptomatic knee.

The practical attraction is assessment across submaximal efforts that may be better tolerated. The limitations are a substantial pulse count, dependence on a preceding MVC, fixed apparatus, processing choices and incomplete clinical outcome validation. This method is not a short substitute for a maximal RFD contraction, and its inverse-second scaling factor cannot inherit an N/s threshold from OAI. Neither a falls endpoint nor a treatment-important-change anchor was established.

Whole rise ratio in advanced OA

Izadi and colleagues assessed 50 advanced symptomatic/radiographic OA patients awaiting unilateral TKA. Their “RFD” was the best peak force divided by the time taken to reach that peak, recorded at 2 kHz. It was not an onset-to-50-ms or onset-to-200-ms slope. The single-session battery also included MVC/body weight, bilateral leg-press 1RM/body weight, KOOS-PS, a 47-cm 30-second chair stand, 40-m fast walking with turns and ten-step ascent. [10]

After adjustment for age, sex and pain, MVC and 1RM were associated with current chair stand, stairs and self-report. The RFD ratio was associated with chair stand and more weakly with stair ascent, but not significantly with KOOS-PS or fast walking. MVC/body-weight added 38% explained variance for chair stand, whereas the RFD ratio added 22% in separate models. The absence of a significant fast-walking association is relevant and opposes a universal “stronger means faster” equation in advanced OA.

These models do not forecast postoperative recovery or establish how many additional stands a patient will gain after an increase in strength. The word “predictor” describes a regression input, not temporal prediction. Body-mass normalization and the bilateral nature of leg press also make direct comparison with unilateral absolute RFD more complex than comparing coefficients alone.

Force steadiness

Submaximal target tracking examines control of force around a requested level. It can be impaired even when maximum force is relatively preserved. Target intensity, visual feedback, duration, fatigue and whether variability is represented by SD, coefficient of variation or tracking error are integral to the definition. These measures are not interchangeable with rate of force relaxation or RFD scaling linearity.

In Sørensen's 41-person cross-sectional knee OA study, submaximal quadriceps steadiness did not significantly explain peak external knee-adduction moment during walking; adjusted R² was 0.05 and p=0.41, with the conclusion unchanged after covariates. The null result is useful: better laboratory steadiness cannot be assumed to imply lower frontal-plane knee loading. Neither load progression nor future falls was measured. [20]

Prognosis and clinically important change

Thirty six month RFD outcomes in OAI

Hu and colleagues analyzed OAI participants aged 45–79 with or at risk of OA, excluding healthy reference participants and selected replacement, injury and inflammatory-arthritis categories. After those exclusions, 3,623 were eligible, but baseline RFD/outcome completeness left 2,238 for walking, 2,017 for chair stand and 2,630 for WOMAC function. These denominators should accompany the result; they are not three independent cohorts. [9]

The Good Strength Chair measured isometric force with the knee fixed at the reported 60° position. The highest peak-force trial from three per side was selected after gravity correction. RFD was the difference between 90% and 30% of that maximum divided by the time between those force levels. The more painful limb was used. This measure omits the initial 30% rise and is not a direct early-onset RFD metric. Device strength reliability quoted in the paper should not be assumed to establish between-day reliability of this derived slope.

Worsening WOMAC function was an increase of at least nine on the 0–68 scale; walking worsening was a gait-speed decrease of at least 0.14 m/s. The authors lacked a validated chair-stand MIC and substituted an MDC90 approximation. Knee surgery/replacement during follow-up also entered the worsening definition. These are therefore composite, definition-dependent outcomes, not purely continuous biological decline.

The highest versus lowest RFD tertile was associated with lower odds of WOMAC worsening: OR 0.68 (95% CI 0.51–0.92). The corresponding walking OR was 0.89 (0.68–1.18) and chair-stand OR 0.94 (0.59–1.50), both null. There was no meaningful correlation between baseline RFD and baseline function measures, with all r<0.14. The result supports an endpoint-specific longitudinal association; it does not show that RFD is uniformly the best functional marker.

Original Table 3 gives a pain-adjusted WOMAC OR of 0.60 (0.40–0.89) for women, while narrative/abstract wording uses 0.57 (0.38–0.86), the table's model without knee-pain adjustment. The table is retained here with an explicit discrepancy. Methods list age, sex, BMI, depression, chronic diseases and pain as confounders, whereas the table footnote lists a shorter set; the exact annotation is not fully consistent. A significant association in women and nonsignificant association in men does not, by itself, demonstrate a statistically tested sex interaction.

The study does not adjust the RFD association for maximum strength as an incremental-prediction question, nor report external validation, discrimination, calibration or clinical utility. Tertile boundaries of approximately 224.8 and 491.6 N/s are sample splits, not diagnostic or patient-important cutoffs. Selection due to missing measurements, coexisting pain and at-risk participants restrict generalization to an OA-only clinical service. Observational association also cannot establish that increasing RFD will prevent the outcome.

Maximum strength and later knee replacement

The MOST cohort analysis included 1,257 participants with frequent knee pain; 1,252 completed follow-up contacts, representing 1,682 knees, with 394 replacements in 331 people over 84 months. The endpoint included both total and partial knee replacement. Ipsilateral strength data were available for 1,108 people and 1,480 knees; the female strength model included 220 replaced knees among 969, and the male model 119 among 511. Baseline knee strength was peak isokinetic torque at 60°/s, retaining the maximum of four trials. Cox models used robust estimates for clustering of two knees per person. [21]

Greater strength was associated with less replacement in women in crude and several adjusted analyses, but the association disappeared after KL-grade adjustment: HR per Nm 1.00 (95% CI 0.99–1.01), p=0.97. No corresponding protective association was established in men. Replacement is influenced by symptoms, structural severity, access, preferences and surgical decision-making; it is not simply a biological OA-progression endpoint. The sample also included knees with KL0–1. These results argue against a stand-alone replacement-risk rule based on strength and illustrate why adjustment can materially change an apparently promising prognostic relationship.

Within person change versus between person difference

Ruhdorfer's cross-sectional OAI analysis of 4,553 participants related approximately 4% lower absolute strength, or 6% lower body-weight-normalized strength, to a six-point difference in WOMAC function. Those are between-person regression relationships. They do not establish that a 4% gain is a valid within-person MIC, especially when some direct repeated-testing errors are considerably larger. The original authors specifically called for longitudinal confirmation. [22]

The subsequent 2,675-person OAI analysis separated strength change during years two to four from strength change preceding that period. Concurrent knee-extensor losses were larger in those with worsening function than those without relevant functional change, but the preceding strength decline did not differ significantly between those later worsening and stable groups (−4.5% versus −4.3%, p=0.87). Functional improvement was accompanied by a concurrent strength increase but preceded by greater strength loss. The authors found no corresponding flexor-change differences. These findings demonstrate that concurrent change relationships do not necessarily provide a useful early warning of future deterioration. Full original adjustment details were not retrieved, so this study is used at its verified abstract-level scope. [23]

An earlier three-year established-OA cohort by Sharma and colleagues also found endpoint-specific results: strength related to better chair-stand outcome, but the association attenuated with pain or self-efficacy adjustment, while psychosocial and activity variables contributed to self-reported function. Its outcome classified change or persistence within baseline quintiles, rather than a universally meaningful patient-change anchor. This reinforces the need to define the endpoint before interpreting a “prognostic factor.” [24]

Clinical selection and implementation

A practical minimum measurement set

For routine serial assessment, select a tolerated knee-extensor protocol with adequate external stabilization and retain absolute force or torque together with normalization. Assess the other limb rather than assuming it is a healthy control. Add knee-flexor or other muscle groups when they answer a specific limitation; their error cannot be borrowed from quadriceps. Include a functional performance test and a patient-reported measure because strength alone does not capture confidence, symptoms, balance or endurance.

A leg-press 1RM or mechanical power assessment can be useful when dynamic capacity or exercise loading is the question and equipment and safety permit. Specify unilateral/bilateral action, range, resistance, speed instructions and the trial selected. For patients with poor single-limb tolerance, report why isolated testing is limited instead of substituting a bilateral score without explanation.

Use RFD when the equipment, stabilization and analysis can support the intended interval, and when the added information justifies the burden. Store the raw force trace. A device's peak-force validation does not establish RFD accuracy; force-curve correlation does not establish equivalent derivatives. Currently, protocol-specific between-day OA error and anchor-based importance are less well established for rapid-force metrics than for many maximal-force tests. This is a research and validation priority rather than a reason to relabel other statistics as MIC.

Table 2 Select the assessment for the clinical question

Practical synthesis of the evidence appraised in this report. Task, population and protocol determine interpretation.

Required data fields

Record diagnosis and stage, test side and bilateral symptoms, body mass, height, muscle/action, joint positions, sensor/contact point, moment arm, fixation, device range and calibration, contraction instruction, familiarization, rest, number of attempts, trial-selection rule, pain and assistance. For dynamic testing, include range of motion, external load and movement velocity. For RFD, additionally retain sampling rate, filtering, onset criterion, baseline window, preload/countermovement checks, derivative window, normalization and software version.

Report raw values for both limbs with the ratio, if one is calculated. A ratio can appear favourable because both limbs are weak or because the comparison side has declined. Do not import a sports return-to-play limb-symmetry threshold into older knee OA without direct validation. Handgrip and sarcopenia classifications may add information about general vulnerability but should not stand in for knee-extensor capacity or an OA-specific falls forecast.

Displaying uncertainty without false precision

An appropriate longitudinal display distinguishes observed change, whether it exceeds a protocol-matched measurement-error estimate, and whether a separately validated patient-important-change anchor exists. Show confidence level and original units. A same-day MDC should not be presented as a guarantee of between-day detectability, and a cohort-derived percentage should not be applied to a different denominator or force setup.

Where the source is abstract-only, formulas are inconsistent or the intended protocol differs, show the raw change and describe uncertainty rather than issuing a categorical recovery verdict. Keep equipment changes visible in the record. Do not automatically convert one device to another with an unvalidated regression or combine maximum force, RFD and estimated chair-rise power into a single score whose measurement properties are unknown.

Conclusions

Knee OA muscle assessment is most useful when it is specific, reproducible and tied to a clinical question. Maximal-force testing has substantial direct measurement evidence but heterogeneous individual error. Mechanical power adds task-relevant information, with mostly cross-sectional evidence and important device/load dependence. Rapid-force and force-control measures illuminate different aspects of neuromuscular performance, but definitions vary so much that careful acquisition and interpretation are essential.

The prognostic literature contains genuine longitudinal associations and equally important null or attenuated results. Higher RFD related to later self-reported deterioration in a mixed established/at-risk cohort without predicting measured walking or chair-stand worsening; strength's association with future replacement disappeared after structural-severity adjustment; concurrent strength change did not consistently precede future functional change. These findings support transparent measurement and cautious clinical reasoning, while setting a clear evidence threshold for any future automated prognosis.

Primary study characteristics

Primary studies supporting the narrative are grouped by measurement or prognostic question. Any consensus recommendation is explicitly identified. Population, protocol, endpoint and source access constrain interpretation. The linked bibliography identifies source-access limitations. Related publications from one cohort are not independent replications.

Table 3 Maximal strength and measurement error

Table 4 Muscle power and rapid force production

Table 5 Prospective outcomes and related evidence

References

References are numbered in first citation order. Study specific source descriptions identify the material examined and do not constitute a study quality rating. Links identify the original publication or the explicitly named primary source version.

1. Koblbauer IF, Lambrecht Y, van der Hulst ML, Neeter C, Engelbert RH, Poolman RW, et al. Reliability of maximal isometric knee strength testing with modified hand-held dynamometry in patients awaiting total knee arthroplasty: useful in research and individual patient settings? A reliability study. BMC musculoskeletal disorders. 2011;12:249. DOI 10.1186/1471-2474-12-249 Source examined: Complete original body including tables.

Source note: SRC-ff8074992423 Koblbauer IF 2011

2. Chopp-Hurley JN, Wiebenga EG, Gatti AA, Maly MR. Investigating the Test-Retest Reliability and Validity of Hand-Held Dynamometry for Measuring Knee Strength in Older Women with Knee Osteoarthritis. Physiotherapy Canada. Physiotherapie Canada. 2019;71(3):231-238. DOI 10.3138/ptc-2018-0051 Source examined: Complete official PMC HTML including tables.

Source note: SRC-e09bcd3da8a6 Chopp-Hurley JN 2019

3. Accettura AJ, Brenneman EC, Stratford PW, Maly MR. Knee Extensor Power Relates to Mobility Performance in People With Knee Osteoarthritis: Cross-Sectional Analysis. Physical therapy. 2015;95(7):989-95. DOI 10.2522/ptj.20140360 Source examined: Complete original article and original Table 2.

Source note: SRC-c9ea82bb35a7 Accettura AJ 2015

4. Tevald MA, Murray AM, Luc B, Lai K, Sohn D, Pietrosimone B. The contribution of leg press and knee extension strength and power to physical function in people with knee osteoarthritis: A cross-sectional study. The Knee. 2016;23(6):942-949. DOI 10.1016/j.knee.2016.08.010 Source examined: Original abstract.

Source note: SRC-c4fe87f0e5a3 Tevald MA 2016

5. Winters JD, Rudolph KS. Quadriceps rate of force development affects gait and function in people with knee osteoarthritis. European journal of applied physiology. 2014;114(2):273-84. DOI 10.1007/s00421-013-2759-8 Source examined: Complete original body including tables.

Source note: SRC-b1b6f9fd523b Winters JD 2014

6. Luc-Harkey BA, Blackburn JT, Ryan ED, Harkey MS, Davis HC, Gaynor BR, et al. Quadriceps Rate of Torque Development and Disability in Persons With Tibiofemoral Osteoarthritis. The Journal of orthopaedic and sports physical therapy. 2018;48(9):694-703. DOI 10.2519/jospt.2018.7898 Source examined: Original abstract; full article unavailable.

Source note: SRC-c2d9c793fe99 Luc-Harkey BA 2018

7. Suzuki Y, Iijima H, Nakamura M, Aoyama T. Rate of force development in the quadriceps of individuals with severe knee osteoarthritis: A preliminary cross-sectional study. PLOS ONE. 2022;17(1):e0262508. DOI 10.1371/journal.pone.0262508 Source examined: Complete original body and main tables; separate supplemental files not inspected.

Source note: SRC-8f46af9b99a5 Suzuki Y 2022

8. Šarabon N, Čeh T, Kozinc Ž, Smajla D. Adapted protocol of rate of force development and relaxation scaling factor for neuromuscular assessment in patients with knee osteoarthritis. The Knee. 2020;27(6):1697-1707. DOI 10.1016/j.knee.2020.09.023 Source examined: Complete original body; some tables unavailable.

Source note: SRC-a77325db5cc6 Sarabon N 2020

9. Hu B, Skou ST, Wise BL, Williams GN, Nevitt MC, Segal NA. Lower Quadriceps Rate of Force Development Is Associated With Worsening Physical Function in Adults With or at Risk for Knee Osteoarthritis: 36-Month Follow-Up Data From the Osteoarthritis Initiative. Archives of physical medicine and rehabilitation. 2018;99(7):1352-1359. DOI 10.1016/j.apmr.2017.12.027 Source examined: Complete official PMC HTML including formulas and tables.

Source note: SRC-afcd76743666 Hu B 2018

10. Izadi M, Tøien T, Øhrn FD, Husby OS, Schnell Husby VK, Winther SB, et al. Muscle strength is a key predictor of physical function in advanced knee osteoarthritis. Musculoskeletal science & practice. 2025;78:103339. DOI 10.1016/j.msksp.2025.103339 Source examined: Complete original body.

Source note: SRC-87d1255499cc Izadi M 2025

11. Fransen M, Crosbie J, Edmonds J. Isometric muscle force measurement for clinicians treating patients with osteoarthritis of the knee. Arthritis and rheumatism. 2003;49(1):29-35. DOI 10.1002/art.10923 Source examined: Original abstract.

Source note: SRC-187895ef22c6 Fransen M 2003

12. Kean CO, Birmingham TB, Garland SJ, Bryant DM, Giffin JR. Minimal detectable change in quadriceps strength and voluntary muscle activation in patients with knee osteoarthritis. Archives of physical medicine and rehabilitation. 2010;91(9):1447-51. DOI 10.1016/j.apmr.2010.06.002 Source examined: Original abstract; full article unavailable.

Source note: SRC-0eddd44c0f77 Kean CO 2010

13. Skou ST, Simonsen O, Rasmussen S. Examination of muscle strength and pressure pain thresholds in knee osteoarthritis: test-retest reliability and agreement. Journal of geriatric physical therapy (2001). 2015;38(3):141-7. DOI 10.1519/jpt.0000000000000028 Source examined: Original abstract.

Source note: SRC-4a0911762d2c Skou ST 2015

14. Sahu PK, Goodstadt N, Ramakrishnan A, Silfies SP. Test-retest reliability and concurrent validity of knee extensor strength measured by a novel device incorporated into a weight stack machine vs. handheld and isokinetic dynamometry. PloS one. 2024;19(5):e0301872. DOI 10.1371/journal.pone.0301872 Source examined: Complete original body including tables.

Source note: SRC-677c91e03bf0 Sahu PK 2024

15. Holm PM, Nyberg M, Wernbom M, Schrøder HM, Skou ST. Intrarater Reliability and Agreement of Recommended Performance-Based Tests and Common Muscle Function Tests in Knee Osteoarthritis. Journal of geriatric physical therapy (2001). 2021;44(3):144-152. DOI 10.1519/jpt.0000000000000266 Source examined: Original abstract; final issue 2021.

Source note: SRC-bf15fdcd5a62 Holm PM 2021

16. Reid KF, Price LL, Harvey WF, Driban JB, Hau C, Fielding RA, et al. Muscle Power Is an Independent Determinant of Pain and Quality of Life in Knee Osteoarthritis. Arthritis & rheumatology (Hoboken, N.J.). 2015;67(12):3166-73. DOI 10.1002/art.39336 Source examined: Complete original body including tables.

Source note: SRC-428fbffe8d01 Reid KF 2015

17. Brisson NM, Stratford PW, Maly MR. Relative and absolute test-retest reliabilities of biomechanical risk factors for knee osteoarthritis progression: benchmarks for meaningful change. Osteoarthritis and cartilage. 2018;26(2):220-226. DOI 10.1016/j.joca.2017.11.003 Source examined: Complete body; final issue 2018; some tables unavailable.

Source note: SRC-7b73c2e20195 Brisson NM 2018

18. Sørensen B, Aagaard P, Couppé C, Suetta C, Johannsen FE, Magnusson SP. Stair climb muscle power is associated with gait speed, sit-to-stand performance, patient-reported outcomes and objective measures of mechanical muscle function in individuals with knee osteoarthritis - secondary analysis from an RCT. Musculoskeletal science & practice. 2025;77:103332. DOI 10.1016/j.msksp.2025.103332 Source examined: Original abstract and verified publication; used only for design/construct context.

Source note: SRC-6fe7ba20434d Srensen B 2025

19. Langgård Jørgensen S, Mechlenburg I, Bagger Bohn M, Aagaard P. Sit-to-stand power predicts functional performance and patient-reported outcomes in patients with advanced knee osteoarthritis. A cross-sectional study. Musculoskeletal science & practice. 2024;69:102899. DOI 10.1016/j.msksp.2023.102899 Source examined: Original article; displayed equation issue noted.

Source note: SRC-a38a5d373489 Langgard Jrgensen S 2024

20. Sørensen TJ, Langberg H, Aaboe J, Bandholm T, Bliddal H, Henriksen M. The association between submaximal quadriceps force steadiness and the knee adduction moment during walking in patients with knee osteoarthritis. The Journal of orthopaedic and sports physical therapy. 2011;41(8):592-9. DOI 10.2519/jospt.2011.3481 Source examined: Original abstract.

Source note: SRC-3ecde1a9af15 Srensen TJ 2011

21. Skou ST, Wise BL, Lewis CE, Felson D, Nevitt M, Segal NA, et al. Muscle strength, physical performance and physical activity as predictors of future knee replacement: a prospective cohort study. Osteoarthritis and cartilage. 2016;24(8):1350-6. DOI 10.1016/j.joca.2016.04.001 Source examined: Complete original body including tables.

Source note: SRC-48f41e13436e Skou ST 2016

22. Ruhdorfer A, Wirth W, Eckstein F. Relationship between isometric thigh muscle strength and minimum clinically important differences in knee function in osteoarthritis: data from the osteoarthritis initiative. Arthritis care & research. 2015;67(4):509-18. DOI 10.1002/acr.22488 Source examined: Complete original body including tables.

Source note: SRC-5f0b1be8742a Ruhdorfer A 2015

23. Ruhdorfer A, Wirth W, Eckstein F. Longitudinal Change in Thigh Muscle Strength Prior to and Concurrent With Minimum Clinically Important Worsening or Improvement in Knee Function: Data From the Osteoarthritis Initiative. Arthritis & rheumatology (Hoboken, N.J.). 2016;68(4):826-36. DOI 10.1002/art.39484 Source examined: Original abstract.

Source note: SRC-f399cde05e94 Ruhdorfer A 2016

24. Sharma L, Cahue S, Song J, Hayes K, Pai YC, Dunlop D. Physical functioning over three years in knee osteoarthritis: role of psychosocial, local mechanical, and neuromuscular factors. Arthritis and rheumatism. 2003;48(12):3359-70. DOI 10.1002/art.11420 Source examined: Original abstract.

Source note: SRC-f53d1b67cf13 Sharma L 2003