Pain induced by injection of hypertonic saline into the infrapatellar fat pad and effect on coordination of the quadriceps muscles


  • Paul W. Hodges,

    Corresponding author
    1. School of Health and Rehabilitation Sciences, The University of Queensland, Brisbane, Queensland, Australia
    • Centre of Clinical Research Excellence in Spinal Pain, Injury and Health, School of Health and Rehabilitation Sciences, The University of Queensland, Brisbane, Queensland 4072, Australia
    Search for more papers by this author
  • Rebecca Mellor,

    1. School of Health and Rehabilitation Sciences, The University of Queensland, Brisbane, Queensland, Australia
    Search for more papers by this author
  • Kay Crossley,

    1. School of Physiotherapy, The University of Melbourne, Melbourne, Victoria, Australia
    Search for more papers by this author
  • Kim Bennell

    1. School of Physiotherapy, The University of Melbourne, Melbourne, Victoria, Australia
    Search for more papers by this author
    • Dr. Bennell has received consultant fees, speaking fees, and/or honoraria (less than $10,000) from the Pan Pacific Rehabilitation Conference.



Musculoskeletal conditions of the knee involve changes in sensorimotor function, but it is unclear whether these changes are a cause or result of pain. Induction of experimental pain may help solve this issue. Although this is commonly achieved by injection of hypertonic saline into muscle, muscle is commonly not the source of pain. This study investigated whether pain induced by injection of saline into the infrapatellar fat pad changes motor control of the quadriceps muscles of the knee.


Ten participants performed a standardized task involving ascending and descending a series of steps. Electromyographic activity (EMG) of vastus medialis obliquus (VMO) and vastus lateralis (VL) was recorded with surface electrodes. Trials were conducted without pain, with anterior knee pain induced by injection (0.25 ml) of hypertonic saline (5%) into the infrapatellar fat pad, with anticipation of pain associated with unpredictable electrical shocks to the knee, and 20 minutes after pain cessation. EMG onset and amplitude were analyzed.


When participants ascended the steps with pain, the onset of VMO EMG was delayed relative to that of VL, in contrast to simultaneous or earlier activation of VMO EMG in the pre- and postpain trials. VL EMG amplitude was decreased significantly from the control condition.


These data show that alterations in coordination of knee muscle activity can be caused by pain, even when it is of nonmuscle origin. Treatment of pain is therefore important to facilitate performance of the quadriceps muscles, which are essential for locomotor and functional tasks as well as for knee stability.


Musculoskeletal conditions involving the knee are common and pain is a predominant feature. Changes in sensorimotor function are associated with many painful knee conditions such as osteoarthritis (1, 2) and anterior knee pain (3), and these have been linked to impaired performance in everyday tasks such as walking and stair climbing (1, 3). The temporal relationship between development of pain and the sensorimotor deficits is not clear. A better understanding of this relationship would assist in prevention and treatment.

Injection of hypertonic saline into muscle induces pain that is described using descriptors similar to those used for clinical pain (4, 5). This technique has been used since the 1930s (4) to study sensory properties of muscle pain and its effect on sensorimotor function (5–7). Most studies have investigated the function of painful muscles (6, 8) or muscles with antagonist/synergist functions (9). For instance, injection of hypertonic saline reduces movement amplitude and velocity (10), electromyographic (EMG) amplitude (10, 11), and motor unit firing rate (12) in the painful muscle. However, this technique has been criticized. Although excitation of nociceptive afferents has been reported (13–16), other axons could be affected by changed ion concentration (17), leading to sensorimotor changes independent from pain effects. Furthermore, activity of the injected muscle may simply change to avoid pain provocation from contraction. This criticism is important because muscle tissue is not the pain source in many clinical conditions (18). Thus, sensorimotor changes observed in response to experimental muscle pain can be difficult to interpret.

An alternative model that avoids these problems is injection of hypertonic saline into the infrapatellar fat pad. This structure is sensitive to mechanical stimulation (18), has a high proportion of nociceptive afferents (19), and is a source of knee pain (20) including some pain in osteoarthritis (21). Furthermore, this technique induces anteromedial knee pain with a distribution similar to that of clinical pain (22). This experimental model has the potential to determine whether knee pain leads to sensorimotor changes while controlling for problems of muscle injection.

An alternative view is that pain has a less specific effect. For instance, threat of pain, without direct stimulation of nociceptive afferents, changes trunk muscle coordination (23). In this case, structures responsible for symptoms appear unimportant. To determine whether stimulation of nociceptive afferents in the fat pad is responsible for effects on motor control, nonspecific effects must be controlled for.

Quadriceps muscles are important for functional tasks and knee stability. Clinically, anterior (patellofemoral) knee pain and knee osteoarthritis are associated with changes in quadriceps control. Knee control in pain-free individuals is characterized by simultaneous activation of medial (vastus medialis obliquus [VMO]) and lateral (vastus lateralis [VL]) quadriceps muscles or earlier VMO activation (3, 24, 25). However, VMO activity is delayed relative to that of the VL in anterior knee pain in some (26, 27), but not all (24, 28, 29), studies. Differences are likely due to differences in experimental methods and tasks. For instance, studies with no timing difference have investigated the time of peak EMG activity rather than onset (28), expressed EMG onset as a percentage of the gait cycle rather than relative latency between muscles, which is less sensitive (24), or used subjects with a history of knee pain but no current symptoms (29). Alternatively, knee osteoarthritis has been associated with delayed VL activity relative to foot contact (1), increased duration of VL activity during gait (2), but no change in the relative timing of the medial and lateral quadriceps (30, 31). Changes in amplitude are difficult to interpret due to difficulty with normalization, and both decreased VMO (32, 33) and VL EMG (34) have been reported. Changes in neural drive such as decreased synchronization of motor unit activity in VMO and VL have been reported (35). Whether changes are replicated in experimental pain is unclear.

The present study investigated whether experimental pain induced in a nonmuscle structure changes coordination of the medial and lateral quadriceps muscles in a similar manner to clinical knee pain. A secondary aim was to determine whether changes are replicated by anticipation of knee pain to control for nonspecific effects of pain.



Ten healthy individuals (7 women) with a mean ± SD age, height, and mass of 33 ± 9 years, 171 ± 10 cm, and 64 ± 11 kg, respectively, were recruited. Individuals were excluded if they reported any history of lower limb pathology or injury to either knee for which treatment was sought or that interfered with function for more than 1 week. The Institutional Human Research Ethics Committee approved the study. All participants provided written informed consent.

Stair stepping task.

Participants ascended and descended a set of 2 stairs, 20 cm in height (Figure 1), at a rate of 96 steps/minute as paced by an external metronome. EMG activity was recorded during the stance phase of stair ascent (concentric contraction) and descent (eccentric contraction) with the right leg for 10 consecutive trials (<1 minute).

Figure 1.

Methods. A, Electromyographic (EMG) electrode sites, site for injection of hypertonic saline, and site for fixation of electrodes for unexpected painful electrical stimulation. B, Experimental setup for stair ascent. VMO = vastus medialis obliquus; VL = vastus lateralis.


EMG activity was recorded using pairs of Ag/AgCl gel–filled surface cup electrodes (20 mm interelectrode distance; GrassTelefactor, West Warwick, RI) (Figure 1). The electrode for VMO was placed ∼4 cm superior to and ∼3 cm medial to the superomedial patella border, and orientated 55° to the vertical (36); the electrode for VL was placed ∼10 cm superior and ∼6–8 cm lateral to the superior border of the patella, and orientated 15° to the vertical (36). Cross-talk between muscles has not been reported using this electrode configuration (37). The ground was placed over the anterior tibia. EMG data were preamplified and further amplified with a gain of 2,000x, band-pass filtered between 20 Hz and 1 kHz, and sampled at 2 kHz using a Power 1401 and Spike2 software (CED, Cambridge, UK). Data were exported for analysis with Matlab 6.1 (Mathworks, Natick, MA).

Pain and anxiety measures.

Pain was assessed using an 11-point numerical rating scale (NRS) anchored with “no pain” and “worst possible pain” and marked in 1-cm increments. In trials with injection of hypertonic saline, pain intensity was verbally rated by the participant every 30 seconds. Pain was also assessed with the McGill Pain Questionnaire (38) after the pain trial. Anxiety during the trials with anticipation of pain was assessed using an 11-point NRS anchored with “not at all distressing” and “completely distressing” (23). Participants were asked, “How distressing did you find that trial?”


Participants performed the stair stepping task under 4 experimental conditions: control trials, trials with anticipation of pain, trials following saline injection when pain was rated ≥4 out of 10, and trials 20 minutes after saline injection when pain free. Time of foot contact on the stair was identified with a custom-built foot switch. This device involved a conductive surface on the stair connected via a 9-volt battery to the data acquisition system and a wire connected to the leg of the subject. Contact of the foot with the step surface closed the circuit and indicated the instant of foot contact regardless of which part of the foot touched first. Order of trials with pain and anticipation of pain were randomized.

In trials with experimental pain, a bolus injection (0.25 ml) of hypertonic saline (5%) was injected into the right infrapatellar fat pad, medial to the patella tendon and proximal to the joint line (Figure 1). The injection was directed at 45° in a superolateral direction. Prior to injection, 0.05 ml of lignocaine (lidocaine 1%) was injected subcutaneously to minimize cutaneous sensations. If a participant reported pain <4 after 2 minutes, an additional 0.1 ml bolus injection was administered (n = 2).

In trials with anticipation of pain, 2 surface EMG electrodes (Ag/AgCl discs, 1 cm diameter, 1.5 cm interelectrode distance) were placed over the patella (Figure 1). Perceptual threshold to cutaneous stimulation (single pulse, 200 microseconds, maximum voltage 250V, 10–20 mA) was determined as the intensity at which the participant could “definitely feel something.” The stimulus was then presented with increasing intensity and participants were instructed to indicate when they would rate the stimulus as 5 out of 10. Because motor stimulation would evoke postural demands separate to those caused by the task, electrodes were placed on the patella. Participants were instructed that, at random and without warning, they would receive a shock that was 80–120% of the target intensity. Stimuli were delivered at random intervals (∼1 stimulation every 3 steps). No stimuli were provided at foot contact so that the period of data used for analysis was not affected.

Statistical analysis.

EMG onsets were identified visually from raw EMG data. Automated methods were not used as they are commonly affected by background EMG (39). EMG data were analyzed in a blinded manner; data were displayed individually without reference to the trial, muscle, or biomechanical events. EMG onset was identified as the point at which the amplitude increased above baseline. EMG onset was expressed as the onset of VMO relative to that of VL, and the onset of each muscle relative to foot contact. Reliability of EMG onset detection has been reported (39, 40) and the standard error of the measurement for the stair stepping task is 5.9–6.2 msec (40). EMG amplitude was calculated for 50 msec after EMG onset. EMG data were normalized to the peak activity across conditions.

The onset of VMO EMG relative to VL and the timing of VMO and VL EMG relative to foot contact were averaged over 10 repetitions and compared between tasks using a repeated-measures analysis of variance (ANOVA) with 2 repeated-measures factors (condition: prepain, anticipation of pain, pain, followup; stair direction: ascent, descent). Amplitudes of VMO and VL EMG were compared between tasks and directions with separate ANOVAs, as the normalization to peak across tasks does not permit comparison between muscles. Post hoc testing was performed with Duncan's multiple range test. The relationship between changes in timing and the amplitude of pain was investigated using Pearson's correlation coefficient and the regression lines were plotted. Data are presented as the mean ± SD throughout. P values less than 0.05 were considered significant.


Experimentally induced pain.

Using the McGill Pain Questionnaire, the most common pain descriptors chosen by participants were aching (86%) and boring (57%). All participants reported pain in the anterior inferomedial region with retropatellar pain also reported by most (70%). Severity of pain peaked at ∼2 minutes when the average pain level reported was 5.5 out of 10 (range 4–9). Pain gradually declined and most participants were pain free by 15 minutes (Figure 2).

Figure 2.

Pain response. A, Pain following injection of hypertonic saline into the infrapatellar fat pad. Data show the mean and SD of the pain reported on the numerical rating scale for the duration of the trial. B, Relationship between pain and change in timing of vastus medialis obliquus (VMO) relative to vastus lateralis (VL). The change in timing of VMO relative to VL during pain is plotted against the peak pain. Negative values indicate VMO is delayed relative to VL during pain. The regression line and 95% confidence intervals are shown.

Anxiety associated with anticipation of pain.

The reported level of anxiety with anticipation of pain ranged from 2–7 (mean 4.4). The average level of pain induced by the electrical stimulation during these trials ranged from 4–7 (mean 5.4).

Temporal changes in EMG activity.

When participants stepped up onto a step with the right leg during the control trial, the mean ± SD onset of VMO EMG activity occurred 13.2 ± 31.2 msec before VL (Figure 3A). The onset of VMO occurred after that of VL in only 3 of 10 participants. Following injection of hypertonic saline into the infrapatellar fat pad, the mean ± SD onset of VMO EMG occurred later relative to VL (12.8 ± 22.8 msec after VL onset) compared with the control trials (main effect: P < 0.05 [condition]; P < 0.01 [post hoc ]). The onset of VMO relative to VL was delayed relative to the control values for 10 of 10 participants as shown in Figure 3A. Due to the delay in VMO onset, this muscle was activated after VL in 7 of 10 participants. Furthermore, there was a significant correlation between pain and the size of the change in relative timing (r2 = 0.35, P < 0.05) (Figure 2B). If the outlier at approximately −90 msec was removed from the data, the correlation increased (r2 = 0.65). Although VMO and VL onsets preceded foot contact by a greater amount on the stair descent than stair ascent (main effect: P < 0.001 [direction]), there was no change in VMO or VL onset relative to foot contact between conditions (main effect: P = 0.067 [condition]). During the followup period, after the resolution of knee pain, the relative latency between VMO and VL onset returned to values that were not different from the control trials (mean ± SD VMO EMG onset 1.7 ± 27.7 msec before onset of VL: P = 0.22 [post hoc]). When painful electrical stimuli were applied to the knee to induce anticipation of pain, the mean difference between the onsets of these 2 muscles was not altered (onset of VMO EMG occurred a mean ± SD of 2.0 ± 32.2 msec before VL; post hoc: P = 0.21).

Figure 3.

Temporal electromyographic (EMG) data. Data (mean and 95% confidence interval) are shown for the onset of EMG activity of vastus medialis obliquus (VMO) and vastus lateralis (VL) relative to the onset of foot contact on the step for walking up and down stairs (A and C). B and D, Individual (circles) and group (rectangle) data for onset of VMO relative to that of VL EMG during each condition when walking up and down stairs. Negative onset indicates onset of VL EMG before that of VMO. * P < 0.05.

When participants stepped down from the step and landed on the right leg, the onset of VMO EMG occurred a mean ± SD of 4.0 ± 32.0 msec after the onset of VL activity in the control trials (Figure 3B). The relative latency between the onset of VMO and VL EMG was not systematically altered by pain or anticipation of pain (post hoc: both P > 0.17). Data from one participant was not available due to movement artifact.

Spatial changes in EMG activity.

VMO and VL EMG amplitudes were greater during stair ascent than descent (main effect [both muscles]: P < 0.001 [direction]) (Figure 4). When participants ascended the stairs, the amplitude of VL EMG was reduced during pain trials (condition × direction interaction: P < 0.033; post hoc: P < 0.01). There was no change with pain anticipation (post hoc: P = 0.16) and EMG amplitude returned to control values after pain resolution (post hoc: P < 0.14). VL EMG amplitude did not differ between conditions when stepping down (post hoc: all P > 0.33). VMO EMG amplitude did not change with either stair ascent or descent (main effect: P = 0.43 [condition]; condition × direction interaction: P = 0.18).

Figure 4.

Spatial electromyographic (EMG) data. Group data are shown for amplitude of vastus lateralis (open circles) and vastus medialis obliquus (solid circles) EMG, as a proportion of peak activity, during each condition when walking A, upstairs and B, downstairs. Means and 95% confidence intervals are shown. Note that the amplitude of VL was reduced during pain when walking upstairs only. * P < 0.05.


These data show that knee pain induced in nonmuscle soft tissue by injection of hypertonic saline induces changes in motor control of quadriceps muscles. This adaptation to experimental pain cannot be explained by a strategy to reduce activity in painful muscle because the muscles were not painful in this study and pain does not increase with contraction using this model (22). As no saline was injected into knee muscles, the study also confirms that direct effects on axons cannot explain changes. Furthermore, as the response could not be replicated by pain anticipation, changes in quadriceps activity are not a generalized response to knee pain.

The changes in quadriceps activation are consistent with those associated with anterior knee pain, but not knee osteoarthritis. Previous studies have identified delayed activity of VMO relative to VL in anterior knee pain during stair stepping (3) and challenges to knee posture (26, 27). This has been debated by others (24, 28, 41). Absence of changes may be due to methodologic issues including investigation of nonfunctional tasks such as open chain knee extension (28, 41), issues related to detection of EMG onset (25), absence of pain at the time of testing (29), and parameters used to compare onset (e.g., peak activity versus onset of activity [28], relative latency between muscles versus onset relative to gait cycle [24]). Recent data using alternative methods to quantify coordination have identified changes in anterior knee pain such as reduced synchronization between firing of motor units in VMO and VL (35). In contrast, changes in quadriceps EMG relative to foot contact have been reported in knee osteoarthritis (1). This was not replicated in the present study.

Is the change in VMO EMG onset relative to VL clinically relevant? During experimental pain, VMO onset was, on average, ∼26 msec later than trials before pain. This was associated with a mean VMO onset ∼13 msec after VL, which is more than double the measurement error. Although the mechanical effect of this delay cannot be deduced from this study, earlier modeling data suggest delayed VMO activity of 5 msec significantly changes patellofemoral joint loading (42). Furthermore, data from longitudinal studies suggest that changes in quadriceps timing of several milliseconds are associated with increased risk for pain development (43). Interestingly, the timing change during stair ascent was related to the reported pain. This is perhaps surprising considering the subjective nature of pain. Other experimental pain studies have demonstrated similar relationships between pain and motor performance (44, 45).

It has been debated whether changes in quadriceps control in anterior knee pain are caused by pain and injury or whether changes contribute to their development (43, 46). Theoretically, poor VMO (the only quadriceps muscle with a medially directed force) activation may increase the potential for lateral patella glide during quadriceps contraction due to the lateral force vector of other quadriceps, and this may change joint loading and mechanically irritate the medial soft tissues (46). As mentioned above, small changes in timing may produce relevant changes in mechanical joint loading. Although it is possible that timing changes may cause pain, the present data suggest the converse is true: changes in quadriceps timing may be caused by pain. This is supported by evidence that improved timing of VMO activity relative to VL in stair stepping is associated with reduced pain in a clinical trial of therapeutic interventions for anterior knee pain (47, 48). The correlation between pain and changes in quadriceps timing reported here supports this relationship.

Pain also induced changes in VL EMG amplitude, but not VMO. Although this is inconsistent with reduced activity of VMO relative to VL observed in clinical pain (32, 33), reduced activity of VL relative to VMO has been reported (34). Unlike previous studies, the present data are not affected by problems of normalization to maximum voluntary contraction (e.g., failure to perform true maximum contraction during pain). Regardless of changes in VL EMG amplitude, the delayed VMO onset relative to VL is consistent with compromised patellar mechanics.

Pain only affected motor control during stair ascent, not stair descent. This differs from clinical pain, which is associated with delayed VMO during both (3). Differences may relate to the relative mechanical demands of the tasks; previous work has loosely associated timing of the quadriceps with knee flexion during stair ascent but not descent (49). Differences in fat pad loading are also possible between the tasks. The stair stepping task was selected for several reasons. First, it is associated with higher loads than walking and is associated with reports of clinical pain (50). Second, this task is associated with changes in vasti coordination in clinical populations (3), and thus provides an ideal model for comparison with earlier data.

Pain anticipation did not change quadriceps timing. This is inconsistent with data from trunk muscles that indicate similar changes in motor control in association with anticipated or induced back pain (23, 51). At the trunk, this finding was interpreted to suggest that cognitive processes contribute to changes in sensorimotor function and that changes were independent of nociceptor firing (23). The present data argue that nociceptor firing is necessary to induce changes in quadriceps activity, and this includes fat pad nociceptors.

The present data rule out direct effects of hypertonic saline on large diameter axons and suggest changes in control are mediated by effects of nociceptor input at spinal or supraspinal levels. It is generally regarded that motor adaptation to pain is an attempt to guard or avoid pain (52). For instance, voluntary movements during pain are associated with decreased activity of agonist muscles and their synergists, whereas activity of antagonists increases to limit movement and force production (7, 10). However, it is uncertain how delayed activity of VMO and reduced activity of VL may be beneficial in this respect during stair stepping. Reduced VL amplitude tends to suggest altered movement strategy during pain, e.g., lower knee flexion moment. This is consistent with decreased knee flexion when people with clinical symptoms walk on stairs (49). The timing deficit is more difficult to explain.

One possibility is that delayed VMO EMG may decrease the medial vector and reduce medial fat pad loading. Alternatively, differential effects on activity of the quadriceps muscles may be maladaptive. Data from other body regions have identified changes in muscle activity that would be expected to negatively affect joint control. For instance, activity of the deep abdominal muscle, transversus abdominis, is delayed in back pain (53), the deep paraspinal muscle multifidus atrophies quickly after injury to an intervertebral disc (54), and activity of ankle everter muscles is delayed in ankle sprain (55). Although difficult to confirm, each of these situations is consistent with the proposal that movement control would be impaired.

Pain affects movement control at multiple levels of the nervous system. Changes in properties along the pathway from the motoneurone (56) to the motor cortex (57–59) have been identified, and changes at any level(s) may be responsible for changes in relative activity of quadriceps. Although not possible to conclude from the present data, previous studies suggest VMO is more susceptible to the influence of pain and injury than other quadriceps muscles. For instance, VMO H-reflex amplitude is reduced with smaller volumes of knee joint effusion than other quadriceps muscles (60). Thus, it is possible that the delayed onset of VMO relative to VL could be due to differential effects of pain on motoneurone excitability or, as the H-reflex may be affected presynaptically (61), differential effects on presynaptic inputs on the Ia afferent may be responsible.

The infrapatellar fat pad is a potential source of pain referral to the anteromedial knee (18) and a source of pain in knee osteoarthritis (21). In a case study, Dye et al (18) reported that mechanical stimulation of the fat pad during arthroscopy, without anesthesia, induced intense pain. A high proportion of nerve fibers in the fat pad express substance P (19). Consistent with this finding, injection of hypertonic saline into the fat pad induces pain that is referred over the anteromedial knee (22). This was likely due to chemical irritation because injection of a similar volume of isotonic saline did not produce pain (22). Magnetic resonance imaging showed that the saline remained within the fat pad, which suggests the effect was due to afferents in the fat pad. The present study suggests that pain from this structure also changes control of knee muscles, confirming the validity of this model to study the interrelationship between pain and motor control. Although control trials with isotonic saline were not used in the present study, other studies of injection of isotonic saline into muscle demonstrated no effect on motor control (5, 10).

This study demonstrates that knee pain induced in nonmuscle soft tissue changes sensorimotor function consistent with changes identified in clinical anterior knee pain. This confirms that changes are not simply due to avoidance of contraction of a painful muscle or direct effects of saline on axons. These data support the importance of consideration of pain control during activities of daily living and in the development of exercise protocols. Further investigation is required to determine if pain control restores normal coordination of the vasti in a clinical population or if other or additional strategies are needed.


Dr. Hodges had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study design. Hodges, Mellor, Crossley, Bennell.

Acquisition of data. Hodges, Mellor, Bennell.

Analysis and interpretation of data. Hodges, Bennell.

Manuscript preparation. Hodges, Mellor, Crossley, Bennell.

Statistical analysis. Hodges.

Acquisition of funding. Hodges, Bennell.