In vivo measured joint friction in hip implants during walking after a short rest

Introduction It has been suspected that friction in hip implants is higher when walking is initiated after a resting period than during continuous movement. It cannot be excluded that such increased initial moments endanger the cup fixation in the acetabulum, overstress the taper connections in the implant or increase wear. To assess these risks, the contact forces, friction moments and friction coefficients in the joint were measured in vivo in ten subjects. Instrumented hip joint implants with telemetric data transmission were used to access the contact loads between the cup and head during the first steps of walking after a short rest. Results The analysis demonstrated that the contact force is not increased during the first step. The friction moment in the joint, however, is much higher during the first step than during continuous walking. The moment increases throughout the gait cycle were 32% to 143% on average and up to 621% individually. The high initial moments will probably not increase wear by much in the joint. However, comparisons with literature data on the fixation resistance of the cup against moments made clear that the stability can be endangered. This risk is highest during the first postoperative months for cementless cups with insufficient under-reaming. The high moments after a break can also put taper connections between the head and neck and neck and shaft at a higher risk. Discussion During continuous walking, the friction moments individually were extremely varied by factors of 4 to 10. Much of this difference is presumably caused by the varying lubrication properties of the synovia. These large moment variations can possibly lead to friction-induced temperature increases during walking, which are higher than the 43.1°C which have previously been observed in a group of only five subjects.


Introduction
It has been suspected that friction in hip implants is higher when walking is initiated after a resting period than during continuous movement. It cannot be excluded that such increased initial moments endanger the cup fixation in the acetabulum, overstress the taper connections in the implant or increase wear. To assess these risks, the contact forces, friction moments and friction coefficients in the joint were measured in vivo in ten subjects. Instrumented hip joint implants with telemetric data transmission were used to access the contact loads between the cup and head during the first steps of walking after a short rest.

Results
The analysis demonstrated that the contact force is not increased during the first step. The friction moment in the joint, however, is much higher during the first step than during continuous walking. The moment increases throughout the gait cycle were 32% to 143% on average and up to 621% individually. The high initial moments will probably not increase wear by much in the joint. However, comparisons with literature data on the fixation resistance of the cup against moments made clear that the stability can be endangered. This risk is highest during the first postoperative months for cementless cups with insufficient under-reaming. The high moments after a break can also put taper connections between the head and neck and neck and shaft at a higher risk.

Discussion
During continuous walking, the friction moments individually were extremely varied by factors of 4 to 10. Much of this difference is presumably caused by the varying lubrication properties of the synovia. These large moment variations can possibly lead to friction-induced temperature increases during walking, which are higher than the 43.1˚C which have previously been observed in a group of only five subjects. PLOS

Introduction
Total hip joint replacement is performed more than 200,000 times in Germany alone [1]. During recent decades, the patients became younger and more active. Hence, their demands for functionality and lifetime of the implants have increased. Loosening of the artificial cup and inlay is one of the most common reasons for the failure of total hip replacements [2][3][4]. Polyethylene wear and aseptic loosening of the cup account for 26% and 48%, respectively, of reoperations [2,5]. Another study demonstrated that 30% to 40% of all revisions require a change of cup or inlay [6]. Wear is caused by friction, and aseptic loosening can be due to moments that stress the fixation in the acetabulum. These moments are not only determined by the patient's activities and, thus, the frequency and magnitude of the contact forces but also by the amount of friction in the joint. In vitro studies using different test conditions [7][8][9][10][11][12][13][14][15][16][17][18] demonstrated that the materials of the implant head and inlay primarily influenced the friction. In several studies, the stability of cup-bone bonding was investigated in cadavers or using plastic bone substitutes. A loosening moment of 8.8 Nm was reported for cementless cups [19], but values as low as 2.2 Nm were reported [20], both with an under-reaming of 1 mm. In comparison, our own in vivo load measurements with instrumented hip implants [21] have determined average friction moments during walking between 2.25 ±0.29 Nm (three months postoperatively) and 1.76 ±0.83 Nm (12 months postoperatively) [22,23]. This indicates that already during typical activities of daily living, such as walking, critical friction moments can occur in total hip joint replacements.
Analogous to higher moments in technical joints after movement started, it was suspected that friction in joint implants may also be higher after a short break, during which time the lubrication film breaks down [24,25]. Based on in vivo measurements [26] of typical activity times and resting periods in hip patients and using a pin-on-ball test, joint friction was investigated in vitro after movement began [24]. Friction after 5 s resting was 30% higher in ceramic-UHMWPE pairings than during the following continuous movement. The increase depended on the tribological pairing of the implant and correlated to the rest time. The results of this study confirmed that the peak moments in hip implants are higher after a rest than during continuous movements and may jeopardize the cup fixation. However, the in vitro data cannot directly be applied to in vivo situations because the kind of movement, the force amplitudes and the lubricant are different. The lubricating property of the synovia has an especially large influence on friction in the joint, as a difference of up to 451% in the friction moments in a group of ten subjects suggests [23]. Such individual differences of lubrication may also effect the possibly increased moments after a rest.
The main goal of this study was to obtain in vivo data on the increases of friction moments and friction coefficients after joint movement began. This enables the estimation of the potential risk of cup loosening due to high moments. For this purpose, contact forces and friction moments in instrumented hip implants of ten subjects were measured during walking, and the friction coefficients were calculated from these data.

Instrumented hip implant
To measure friction between the head and cup in vivo, instrumented hip implants were used [21]. The titanium implant (CTW, Merete Medical, Berlin; Germany) was combined with a 32 mm Al 2 O 3 ceramic head and an XPE inlay. The neck of this clinically proven standard implant was modified to house an inductive power supply, six strain gauges, signal amplifiers and telemetric data transmission [21]. The external measurement system [27,28] supplied the inductive coil around the patient's hip joint. The received signals were recorded simultaneously with the patient's images on video tape. Joint contact forces and friction moments. The femur-based coordinate system [29] was located in the head center of a right-sided implant [30]. Data from left implants were mirrored to the right side. The resultant joint contact force F res was calculated from the three force components in the lateral (F x ), anterior (F y ), and superior (F z ) directions. The resultant friction moment M res was determined from the three components M x , M y , M z , rotating positively around the corresponding axes.

Calculation of coefficient of friction
Based on all force and moment components, the magnitude of the three-dimensional coefficient of friction μ was calculated [23], assuming Coulomb friction and a head radius R: The lever arm H is given by the following equation, see also [23] H The coefficient μ was only determined for F res ! 25%BW and M res ! 0.02%BWm to ascertain an accuracy of μ better than 5%.

Patients and measurements
Ten patients with instrumented implants participated in the study (Table 1). They gave their informed written consent to participate. The study was approved by the ethical committee of the Charité-Universitätsmedizin Berlin, Germany (EA2/057/09) and was registered in the German Clinical Trials Register (DRKS00000563). The measurements were performed an average of 17 months (12-31 months) postoperatively during 10m of level walking at a selfselected walking speed. Five to eighteen trials per subject were recorded. The subjects stood still on both legs for 12 s, on average, before they started walking with the ipsilateral leg.

Data evaluation
All forces were determined as a percent of the patient's bodyweight (%BW); the friction moments in %BW Ã m. For a subject with a body weight of 100 kg, as an example, the values must be multiplied by a factor of 9.81 to obtain numbers in N or Nm. The continuous time patterns and the numerical peak values of forces and moments were analyzed separately for each of the first four steps after rest. Each complete step started and ended at the instants when F res became a minimum (Figs 1 and 2). The 'Start' phase, preceding the first step, started when the ipsilateral leg started to move after stance and ended when F res became a minimum before heel strike. If not mentioned as being 'individual', all reported data refer to results from the average subject.
Time patterns: First, the durations of the Start phases and the four steps were averaged ( Fig  1) from all individual trials (Table 1). Then, the time patterns of all six force and moment components and the results from all trials were averaged, separately for the Start phase and the four steps. For averaging, a dynamic time warping procedure was applied [31], which delivered an output that retains the typical maxima and minima of the included signals. Finally, the obtained individual averages were averaged again from all subjects to obtain the load-time behavior of an average subject (Fig 2). The friction coefficient μ throughout each trial was calculated from these individual or average force and moment components.
Numerical values: For the Start phase (Fig 1), the absolute maxima F start , M start , and μ start of the resultant force, the resultant moment and the friction coefficient were analyzed. During each of the following steps, the curves of the force F res always exhibited two peak values F CTO and F CHS (Figs 1 and 2), at approximately the instant of contralateral toe off (CTO) and contralateral heel strike (CHS). One of both peak values was always the absolute maximum of F res . For the moment M res , the two values M CTO and M CHS were determined at the same instants as the peak forces. The maximum resultant moment M max throughout the entire cycle duration mostly acted very shortly after M CHS and is about 10% to 15% higher ( Table 2). Two numerical values μ CTO and μ CHS of the friction coefficient were also identified at the instants of the two peak forces. The absolute maximum μ max of the friction coefficient was denoted as μ max . Separately for the Start phase and the four steps, these distinct force, moment and coefficient values were first determined numerically from the time patterns of the single trials and then averaged in the same sequence as the load-time patterns. First, the individual averages were calculated and then, based on the obtained numbers, the values for the average subject plus the corresponding minima, maxima and standard deviations. Information about the variations of all parameters in the individuals and their average is supplied by the reported standard deviations. The procedure used for averaging the time patterns minimized the summed errors between all included patterns throughout the entire measurement time. Therefore, the peak values in the curves of the average subject (Fig 2) can slightly deviate from the corresponding, numerically averaged peak values (Table 2).
To determine whether the values of the eight F res , M res and μ measurements during the first step were different from the corresponding values during the last step, each measurement from step 1 of the individual and average subjects was compared to the value from step 4. The obtained differences in percent were statistically analyzed (Wilcoxon, p 0.05).
Rest times: To investigate whether the duration of the rest time influenced the changes of M res and μ, the individually averaged rest times per trial (Table 1) were correlated to the changes of the corresponding six M res and μ values between steps 1 and 4.

Joint contact forces F res
Time patterns: During the early gait phase until CTO, F res was much higher during step 1 than during steps 2 to 4 (Fig 2). This surplus was 59% when the steps started and 35% at ipsilateral heel strike (HS). After reaching CTO and until the end of the cycles, F res was nearly the same for all four steps.  [1]

Start
Step 1 Step Numerical values: The individual variations of the two force peaks F CTO |F CHS were large ( Table 2). For each of the four steps the standard deviations were approximately 23|13% of the average values. From step to step, both average force peaks were nearly unchanged ( Table 2, Figs 2 and 3). F CTO during step 1 exceeded the value from step 4 by only 1.3% (Table 3); for F CHS this difference was -0.8%, that is, F CHS was slightly smaller during step 4 than step 1.  Table 2. https://doi.org/10.1371/journal.pone.0174788.g003

Friction moments M res
Time patterns: The friction moment M res had a different time behavior than F res and revealed only one maximum at or shortly after CHS (Fig 1). During step 1, M res rose sharply after heel strike until CTO, while F res increased, and then changed only a little until CHS (Fig 2). In contrast to this, M res increased nearly linearly between HS and CHS during the following steps. After CHS, M res always fell, approximately until the joint movement changed from extension to flexion. This instant was controlled by the synchronous videos. Then, it rose to an intermediate peak value, most pronounced for step 1, and continuously fell until the step ended. During the entire cycle time, M res was higher during step 1 than later. Numerical values: In all subjects, except H10R, all three moment values, M CTO , M CHS and M max were higher during step 1 than later ( Table 2). The individual variations of all three moments, were large. For step 1, for example, the standard deviations were 39%, 38% and 33% of the average values. The small red circles shown in Fig 3 illustrate the huge range of individual values of M CHS during step 1. During the next three steps the standard deviations were even higher. In the average subject, however, uniform step to step changes were observed ( Table 2). M CTO fell from step to step (Fig 3) and was most pronounced from step 1 to 2. With 143% (p = 0.005), the surplus from step 1 relative to step 4 was very large (Table 3). M CHS was much higher than M CTO and decreased continuously but was less pronounced than M CTO from step 1 to 2. The value during step 1 was only 35% (p = 0.013) higher than during step 4. The maximum moment M max only slightly exceeded M CHS . The step to step changes of M max were similar to those of M CHS , with a total surplus during step 1 of 32% relative to step 4. The moment courses in Fig 3 indicate that all three friction moments will probably only slightly decrease further after step 4.

Coefficient of friction μ
Time patterns: The charts of μ from the average subject (Figs 1 and 2) show that it was permanently higher during step 1 than during the following steps. During step 1, μ in the average subject already rose at HS and stayed at a high level after CTO. The rise during the following three steps only started after CTO. When hip flexion began, μ was nearly the same for all four steps. It then uniformly and sharply increased to the absolute maxima at around ipsilateral toe off (TO), which were more than twice as high than the values during the whole stance phases. After TO, μ continuously decreased during the remaining swing phase.
Numerical values: All three individual friction values, μ CTO , μ CHS and μ max , varied a lot, as observed from the ranges and standard deviations in Table 2. In subject H10R, all three friction values were much higher than in all other patients. The individual values of μ CTO |μ CHS exceeded the average ones by up to 176%|239% In the average subject, μ CTO was 122% (p = 0.005) higher during the first than during the last step (Table 3, Fig 3). For μ CHS , this surplus was 36% (p = 0.014). The extreme individual surplus of μ CHS from step 1 to 4 was 105%, observed in subject H7R. The maximum coefficient μ max was on average by only 8% higher during step 1 than step 4. The declines in all three μ values with the step number (Fig 3) was less pronounced than the drops observed for the moments. During step 2, μ max was even larger than during step 1, an effect observed for six of the ten individual subjects.

Start phase
During the Start phase, the patterns of F res , M res and μ were different from those during step 1 (Fig 1). The numerical value of F start ( Table 2) was on average 48% lower, compared to F CTO during step 1, with individual variations between -60% to -32%. The values M start |μ start , however, were on average 3%|7% higher, compared to F CTO during step 1. However, these changes again varied a lot, with ranges of -49% to +135% for M start and -51% to +50% for μ start .

Rest times
No or very poor correlations existed between the individually averaged rest times (Table 1) and the six values of M res and μ (Table 3); R 2 was always below 0.23.

Discussion
High friction moments in hip implants increase wear in the joint, especially during the early postoperative weeks, when the fixation stability of cementless implants is lower than later; high friction moments can possibly also endanger the fixation of the cup. It was shown [30] that the peak friction moment during some activities can already reach values that were reported in the literature to jeopardize the cup fixation. [24,25] It has been reported that the in vitro friction moments are higher during the first loading cycle after a rest than during continuous movement. This indicated that frequently increased moments after rests might increase the risk for cup loosening or lead to more wear. Because the test conditions in these studies were not realistic, we examined whether the friction moments and the coefficient of friction are higher in vivo when walking starts after a rest.
With regard to the reported large individual variations of all load parameters, the average values cannot be generalized. The friction coefficient μ CTO , for example, was by 542% larger during step 1 than step 4 in one subject, but by only 16% in another one. The low significance of the load changes is indicated by the low p-values in Table 2. However, decreases of all moments and friction coefficients except μ max were observed in at least eight of ten subjects. Therefore, the moment and friction increases after a rest prior to walking can be expected for the majority of subjects, but their extend cannot be predicted exactly for a specific individuum. Other limitations to this study are the small number of ten investigated subjects and that only one tribological paring was investigated (Al 2 O 3 /XPE).

Friction moment and coefficient
During the first step after standing, the friction moment M res at the instant of the first|second force maximum was 143%|35% higher than during step 4. At the same time, the friction coefficient μ from step 1 exceeded that from step 4 by 122%|36%. This means that the decreases of M res and μ throughout the first four steps are approximately proportional. Figs 1 and 2 show that M res and μ, during the initial step, rise sharply after heel strike, when the contact force F res increases, and stay at increased levels until the CHS, when F res falls again. This behavior is in sharp contrast to the changes of M res and μ during continuous walking (assumed to be represented by step 4), when both measures increase continuously throughout the whole stance phase. The force F res during step 1 is only initially higher than later.
Possible explanations for these observations are as follows: During the initial rest, all or most of the synovia is squeezed out of the joint, leading to a nearly non-lubricated contact between head and cup surfaces. Throughout the first step, until toe off, no synovia can be transported back into the contact zone because this area is pressurized by the high contact force F res . When F res falls after toe off, the joint movement during the swing phase transports synovia back into the joint. The second and all following steps, therefore, start with a sufficient lubricating film. The high contact force then again reduces the lubricating film throughout the stance phase, which leads to the continuous increase of M res and μ. The high values of F res in the beginning of step 1 are probably required for accelerating the body to walking speed.
If these explanations are valid, the friction moments during the one legged stance, when small joint movements cannot be avoided, should also be high. This assumption is confirmed by some exemplary measurements in the public data base OrthoLoad.com (parameters: Implant = 'Hip Joint III' and Activity = 'One Legged Stance'). Such high moments will be the focus of a future study.
With 0.045|0.057, the average values of μ CTO |μ CHS during step 1 were very close to the maximum of 0.055 found in simulator tests [25]. Only μ max , acting at TO when F res had already fallen to approximately one-third of the two maxima, was three times higher than this literature value. Because all three μ values have fallen to nearly constant levels until step 4; the numbers from this step can be compared to those previously reported for continuous walking [22,23].
All subjects received implants with the same tribological pairing (Al 2 O 3 /XPE). Nevertheless, the friction moments and the coefficients of friction individually varied a lot. Data from step 4, assumed to be representative for continuous walking, demonstrated variations of M CTO |M CHS | M max by factors of 10|5|4. For μ CTO |μ CHS |μ max , these factors were nearly identical. Subject H10R especially stands out as the moment and coefficient values were always extremely higher than those in the other subjects. The three values of M res and μ for H10R exceeded the averages from all subjects by up to 167%|239%. Such large variations are probably caused by different individual lubrication conditions, which can be influenced by (i) the lubricating quality of the synovia [14], (ii) the roughness of the gliding surfaces [14,32], (iii) the joint clearance [14,32,33] and (iv) the orientation of the acetabular cup [14,34], which influences the load transmitting area. Data on the impact of other factors, such as the sliding speed or joint contact area, will be investigated in a further study.

Wear
Frequent reasons for revisions of hip joint replacements are still wear and pathological reactions to wear particles [2,5,35]. Simulator studies demonstrated a correlation between the friction between sliding partners and the wear rates [33]. In theory, increased friction moments and coefficients during the first step of walking could, therefore, increase the wear. However, during walking and other repetitive activities, a single starting cycle with high moments is typically followed by many continuous loading cycles with lower moments. Because the wear volume not only depends on the height of the friction moment but also on the number of loading cycles, much increased wear due to increased moments after rest should not be expected.

Cup loosening
The primary and long-term stability of the cup fixation depends, except for the height of the friction moments, on the quality of the cup-bone interface, the fixation technique, the type of porous coating and the bone quality. Simulator studies demonstrated that insufficient underreaming of the acetabulum decreases the primary fixation stability (Curtis M. J. et al., 1992;Tabata et al., 2015). With an under-reaming of only 1 mm, the loosening moments lay between 2.2 and 8.8 Nm. The average maximum moment of M res = 2.15Nm (0.236%BWm, assumed BW = 1000 N), reported here for step 1 just meets the lowest reported value. However, the highest individual moment from subject H10L was 4.28 Nm (0.43%BWm, BW = 995 N) and this is much higher than the lowest stability level reported in the literature.
An additional risk factor, not considered in the current study, is the fact that joint friction changes during the first months after replacement (Damm et al. 2015). Three months after surgery, M res was on average by 47% higher than after 12 months. The extreme moment value in H10L would then have further risen to 6.33 Nm. Possible factors for higher postoperative friction may be a still lacking smoothening of the gliding surfaces and initially insufficient synovia properties. Therefore, even higher values than reported here must be expected shortly after surgery.
The high maximum in vivo friction moments and the lowest reported in vitro loosening moments together indicate that the higher moments at the beginning of walking can put the cup fixation at risk in subjects with high body weight, inferior lubricating properties of the synovia and cementless cups, which are inserted with too small under-reaming. This risk is highest during the first postoperative months.

Joint temperature
High temperatures in artificial joints could be a potential risk factor for the longevity of the implant system. For the combination of conventional polyethylene and ceramic cups with ceramic and metal heads, friction-induced temperature increases up to 43.1˚C after one hour of walking were reported [36]. These increases varied dramatically and individually, depending on the lubrication properties of the synovia [37] and the tribological materials.
By the same reasons as for the wear a distinct influence of increased moments after breaks on the implant temperature during walking can be excluded. However, the current study again demonstrates that friction moments and coefficients are extremely varied from person to person. The three moment measures individually differed by factors of 4 to 10. A factor of four approximately corresponds to the individual differences of temperature increases measured in vivo during walking [36]. The higher factor of ten let us assume that the implant temperature may rise much more in some subjects than observed previously.

Head-stem connection
Friction influences the mechanical stress in the head-neck taper region and, if existent, the neck-stem connection. Up to 143% higher friction moments after a rest, compared to continuous walking, will lead to an increase of these stresses. Therefore, higher friction after a rest can be a potential risk factor for the mechanical stability of suboptimal taper connections. If the mechanical connection becomes loose, micromotion between the components increases and this begins a corrosion cascade [38,39] that can eventually cause implant loosening. Furthermore, increased wear and corrosion products provoke various biological and chemical effects in the surrounding tissues [39,40], which also lead to implant failure.

Conclusion
It was demonstrated that hip joint friction during the first step of walking after a rest is much higher than that during continuous movement. The initial friction moments were raised on average by 35% to 143% and individually by up to 621% compared to continuous walking.

Wear and cup loosening
The higher moments will probably increase wear in the joint only very slightly, but they can endanger the fixation of the cup in the acetabulum. This risk is highest for cementless cups with insufficient under-reaming and during the first postoperative months.

Head-stem connection
The high moments when walking begins can also put taper connections between the head and neck and between the neck and shaft at a higher risk.

Joint temperature
In five subjects, friction-induced temperature increases up to 43.1˚C were observed in hip implants during continuous walking [36]. There was much individual variation, and this was explained by different lubricating properties of the synovia. The measured friction moments during continuous walking, reported here, individually varied by factors up to 10, which is more than the differences of the reported temperature increases. Therefore, it seems worthwhile to perform another investigation with a larger group of patients to determine whether even higher implant temperatures during walking may endanger the long-term outcome of the replacement.

Additional data
Selected examples of the in vivo measurements, on which this study was based, are published in the public data base www.orthoload.com.