Assessing brain integrity in patients with long‐term and well‐functioning metal‐based hip implants

Production of metal debris from implant wear and corrosion processes is now a well understood occurrence following hip arthroplasty. Evidence has shown that metal ions can enter the bloodstream and travel to distant organs including the brain, and in extreme cases, can induce sensorial and neurological diseases. Our objective was tosimultaneously analyze brain anatomy and physiology in patients with long‐term and well‐functioning implants. Included were subjects who had received total hip or hip resurfacing arthroplastywith an implantation time of a minimum of 7 years (n = 28) and age‐ and sex‐matched controls (n = 32). Blood samples were obtained to measure ion concentrations of cobalt and chromium, and the Montreal Cognitive Assessment was performed. 3T MRI brain scans were completed with an MPRAGE sequence for ROI segmentation and multiecho gradient echo sequences to generate QSM and R2* maps. Mean QSM and R2* values were recorded for five deep brain and four middle and cortical brain structures on both hemispheres: pallidum, putamen, caudate, amygdala, hippocampus, anterior cingulate, inferior temporal, and cerebellum. No differences in QSM or R2* or cognition scores were found between both groups (p > 0.6654). No correlation was found between susceptibility and blood ion levels for cobalt or chromium in any region of the brain. No correlation was found between blood ion levels and cognition scores. Clinical significance: Results suggest that metal ions released by long‐term and well‐functioning implants do not affect brain integrity.

][6][7][8][9][10][11][12] However, most studies are based on adverse reactions, or unsuccessful implants, which primarily investigatethe extreme release of metal ion concentrations to the bloodstream or metal hypersensitivity.There has beenno study that hasconsidered the joint evaluation of both brain anatomy and physiology (brain integrity) in well-functioning devices versus population-based controls without joint replacement.While certain THA and HRA devices have been associated with high revision rates, not all patients with these implants require revision because not all patients develop symptoms or adverse local tissue reactions. 13It is unclear whether the patients who still have well-functioning devices are at a risk of accumulating cobalt and chromium ions in their brains over long periods of implantation.Further, since trends associated with modern THA include implanting patients at younger ages and an expectation for multiple decades of service, understanding whether long-term accumulation of metal ions in the brain might be occurring is a worrying concern that needs to be addressed.Acomparison against healthy controls could indicate any correlation between metalion levels and brain integrity.
In the past decade, it has become possible to assess human brain integrity in vivo using MRI.For example, a study using MRI voxel-based morphometry revealed that patients with HRA had reduced gray matter attenuation in the occipital cortex and basal ganglia compared to patient with metal-on-polyethylene THA, as well as a smaller optic chiasm. 65][16][17] Greater accumulation of substances with similar magnetic susceptibility-such as the metals used in hip replacements-lead tolarger magnetic field distortions, potentially indicatingtheir deposition in the tissue.Conventionally, gradient echo (GRE) T2*-weighted imaging, susceptibility-weighted imaging (SWI) and R2* (transverse relaxation rate) imaging has been used to assess magnetic susceptibility in MRI.However, GRE and SWI are only suitable for qualitative assessment, and although R2* may allow quantitative valuation, it is associated with certain measurement errors. 14,164][25][26] Prior applications of QSM include measuring iron and calcification levels within specific structures of the brain. 15,16,22,27,28Due to the paramagnetic nature of cobalt and chromium, QSM may have the capability to detect the accumulation of these metal ions in the brain.
Paramagnetic materials, such as iron, have a positive susceptibility value, while structures such as myelin possess a diamagnetic susceptibility (i.e., a negative susceptibility value).The presence of toxic metal ions like cobalt and chromium might lead to a potential degradation of myelin integrity.QSM considers the net susceptibility from all elements contained in an image voxel, allowing analysis and comparison of metal accumulation between implanted and control participants.Since myelin has a magnetic susceptibility that is opposite to that of metal ions, this contrast in magnetic properties could result in a notable susceptibility difference in QSM for implanted subjects with metal accumulation in comparison to control subjects.On the other hand, myelin further elevates R2*.Therefore, complimentary information from QSM and R2* is important in understanding potential metal and myelin interactions in the brain. 29Therefore, the objectives of the present study were to (1) determine if brain integrity differs between subjects with wellfunctioning hip implants and age-and sex-matched controls, (2) evaluate correlations between QSM or R2*, blood metal ion levels, and cognitive impairment and blood ion levels.

| METHODS
This is a level III cross-sectional study.Ethics approval was obtained from our institutional ethics board and all subjects provided written informed consent.Study participants (n = 60) were recruited from the London Health Science Centre from March 2020 to October 2022.

| Patients
Participants were recruited from (1) two previously randomized control trials at our institution, one of which comparing metal ion release in primary total hip metal bearings (NCT00962351) and the other investigating early failure of modular metal-on-polyethelyne implants (NCT 00208494), and (2) any patient who had undergone THA and received a metal-on-metal implant.Participants were mailed letters of information and followed-up by a phone call.The study population consisted of older adults of over the age of 55 who had received a metal-based implant and were otherwise healthy and able to provide informed consent.Implants included were metal-on-metal THA (n = 18), ceramic-on-metal THA (n = 1), dual-modular neck metal-on-polyethylene THA (n = 5), or metal-on-metal HRA (n = 4).
We excluded patients who had undergone revision of the device or had in implantation time of less than 7 years to look atstabilized metal ion levels in the system.Age-(±3 years) and sex-matched controls with no prior history of joint arthroplasty were recruited from patients on the surgical wait list to undergo total hip or knee arthroplasty.Controls were excluded if they had any current hardware or unsuitable implants for MRI, such as pacemakers, cardiac stents, aneurysm clips, staples, nails, and screws.
A total number of 506 patients were screened to participate in the study.For the implant group (n = 28), 49 were considered eligible, of which 32 agreed to participate in the study.Four patients were later withdrawn due to medical concerns, difficulty fitting into the MRI, or to being lost to follow-up.For the matched-control group (n = 32), 352 were excluded, 83 were considered eligible for this study, of which 32 agreed to participate and completed the MRI.No differences in cohort characteristics were found (Table 1).Data was collected at a single time point.Subjects were examined using 3T MRI (Prisma Fit; Siemens Healthineers).T1-weighted, magnetization-prepared rapid acquisition gradient-echo (MPRAGE) pulse sequence (1.0 × 1.0 × 0.9 mm 3 resolution) images were acquired.

| Data extraction from imaging
Quantitative assessment was performed through a multistep imaging pipeline.The ME-GRE sequences generate magnitude and phase images.Within a given brain region, phase images describe the change in magnetic field, from which QSM and R2* are produced.
There are three conventional steps before the QSMmap: (1) phase unwrapping, (2) background field removal, and (3) solving the inverse problem to convert phase images into susceptibility maps.A single-step QSM algorithm, which combines steps and avoids error amplification between steps, was used to directly convert multiecho GRE sequences into susceptibility images. 30Susceptibility images are expressed in parts per billion (ppb), describing the strength of the response from a voxel to the magnetic field.R2* maps are generated subsequent to the following steps: (1) phase image calculations, (2)   complex difference image calculations, (3) field map estimation, (4) phase unwrapping, and (5) calculation of R2* values, given by the formula R2* = (Δφ)/(2π × ΔTE), where Δφ is the phase difference between echoes and ΔTE is the time difference between echoes.
Five deep brain structures were identified from reports of anatomical and metabolic changes in the brains of THA patients: amygdala, caudate, hippocampus, pallidum, putamen. 4,6Four middle and cortical brain regions of interest were also identified: anterior cingulate, inferior temporal, cerebellum.Automatic subcortical segmentation with automated labeling of voxels was performed on the T1-weighted images using FreeSurfer software (v7.1)with no manual editing of segmentations.
Segmentation labels and quantitative MRI (QSM or R2*) images were co-registered by performing a rigid registration between the T1-weighted and the GRE images using ITK-SNAP software (v3.8) to allow extraction of QSM and R2* measurements and cortical volumesfor each of the nine regions of interest.Interpolation was performed to adjust for spatial resolution differences in segmentation and QSM/R2* maps.The magnetic dipole patterns, although well-known, mathematically hides the absolute susceptibility content; thus, results will bestandardized to the well-known patterns of the the cerebral spinal fluid (CSF).

| Statistical analysis
Correlation of susceptibility to metal ion levels wasour main outcomes of interest, for whicha sample size of at least 28 participants was required to detect moderate correlation (r = 0.51) or better with an α value of 0.05 and power of 0.8.Descriptive statistics were generated for patient characteristics, susceptibility, R2*, whole blood ion, and cognitive data.Susceptibility and R2* measurements between anatomical regions were compared using a Friedman multiple comparisons test.Two-tailed Pearson's correlation were computed for all correlation analyses between QSM and R2* values, blood ion levels, and MoCA scores.The effect of brain regions and the presence of an implant on susceptibility was assessed using a two-way ANOVA.All statistical tests were completed using Prism version 8.2.1 (GraphPad Software).c Whole blood cobalt and chromium was not collected for controls.

In
There was no difference (p = 0.9938) in susceptibility between the subjects with implants and the control subjects (Figure 1).There were differences in susceptibility between brain regions that were consistent for both implanted subjects (p < 0.0001) and controls (p < 0.003).Deep brain structures had significantly higher susceptibility values compared to the middle and cortical brain regions (p < 0.0001).Susceptibility levels were greatest in the pallidum.No differences were found between left and right brain hemispheres.
There was also no difference (p = 0.9851) in R2* between the subjects with implants and the control subjects (Figure 2).As with susceptibility, there were differences in R2* between brain regions that were consistent for both implanted subjects (p < 0.0001) and controls (p < 0.0001).Deep brain structures showed larger R2* values compared to the middle and cortical brain regions (p < 0.0001).R2* was greatest in the cerebellum.No differences were found between left and right brain hemispheres.
No differences in cortical volume were found between both cohorts (p = 0.9947).Mean cortical volume (in voxels) was 4439.No correlation was found between susceptibility and blood ion levels of cobalt or chromium or chromium for any brain region.
Correlation between ion levels and susceptibility in the pallidum (where susceptibility was greatest) is presented in Figure 3.No correlation was found between cobalt or chromium ion levels and R2* for any brain region.Correlation between ion levels and R2* in the cerebellum (where R2* was greatest) is presented in the Figure 4.
Four implanted subjects had cobalt concentrations higher than MoCA scores and susceptibility or R2* for any brain region in either the implanted subjects or the control subjects (Figure 6).

| DISCUSSION
6][7]32 However, no previous study has evaluated such possible changes at long-term follow-up in a cohort of subjects with well-functioning implants with a known elevated risk of revision who remain clinically asymptomatic and therefore unrevisedas compared to background controls.5][16][17] SF-12, WOMAC, and Harris Hip scores were used to assess joint function.A 10-year survivorship analysis of cementless primary total hips using cobalt chrome on highly cross-linked the scores from that study and its conventionally accepted standard of care implants to the implanted cohort in the present study, mean scores in the implant cohort were within 6.4 points for SF-12 MCS, 3.9 points for SF-12 PCS, 6.2 points for WOMAC, and 1.1 point for Harris Hip, with our implanted cohort having higher scores indicating, on average, better patient outcomes and related functionality.We also included cognitive assessment and measures of blood ion levels alongside the imaging to better appreciate any correlations between metal deposition and effects in both brain anatomy and physiology.
Age-and sex-matched controls who had not yet undergone joint arthroplasty were examined as comparators.We detected no findings that would indicate differences in cognition orbrain integrity occurring in subjects with long-term and well-functioning implants.
We can therefore infer that the subjects with long term wellfunctioning hip implants are not more susceptible to cognitive decline as a side effect of their implants.
No significant differences in QSM susceptibility or R2* were observed in any brain region between the implant and control groups.This finding can be attributed to several factors.First, iron concentration is orders of magnitude more abundant than cobalt and chromiumdue to its role in oxygen transport, energy metabolism, and neurotransmitter and myelin synthesis, so subtle concentration differences of cobalt and chromium likely are not affecting the overall susceptibility.8][39][40][41][42][43][44] A 2012 study compared QSM methods to published estimates of iron in the brain from in vivo post mortem data and reported the same ordering of iron concentration by brain structure as we found in this study, with the highest concentration in the pallidum. 42,45However, an alternative reasoning may be cobalt and chromium toxicity to myelin integrity, 46,47 which counters iron concentrations in the overall susceptibility. 23,48Consequently, the negative impact of these metals on myelin could indirectly indicate their presence rather than their actual concentration in QSM.
Langkammer and colleagues validates the assumption, concluding that iron is the most dominant source of magnetic susceptibility in deep gray matter, whilst myelin anisotropic susceptibility of tissue microstructure may confound the susceptibility effects from metal in white matter. 49Second, the lack of statistically significant differences suggests that the measured levels of metal ions in patients may still fall within biologically acceptable ranges.Alternatively, it is possible that the blood-brain barrier restricts the entry of these metal ions into the brain, or once inside the brain, they bind with other components, like albumin, to mitigate their toxicity, thus preserving the diamagnetic susceptibility of healthy myelin that counters the paramagnetic susceptibility from iron, cobalt, and chromium. 12ross both implanted and control subjects, the deep brain structures were observed to have significantly greater susceptibility compared to their middle and cortical brain counterparts, suggesting that the increased susceptibility of the deep brain structures is due to normal brain physiology, rather than implant wear and corrosion.This is consistent with studies reporting that increased susceptibility of deep gray matter structures occurs with normal aging. 21,50,513][54] In QSM, the pallidum consistently had the greatest susceptibility, while in R2* the cerebellum had the greatest values followed by the pallidum.3][54][55][56][57] The difference between QSM and R2* may be attributed to a larger presence of myelin in the cerebellum.Moreover, R2* is susceptible to nonmetallic influences on the signal, particularly from myelin and tissue microstructure in white matter of the brain, which may explain the much wider standard deviations seen as compared to QSM. [58][59][60] One possible explanation for increased pallidum susceptibility involves the choroidal artery, the vessel supplying the two most vulnerable cerebral structures: the globus pallidus and the hippocampus.It is hypothesized that the vulnerability of an artery to pathological damage may be expressed by the ratio of its length over its diameter, and the choroidal artery has the longest free subarachnoid course for its diameter. 2As such, defects in vascular permeabilities may allow metal ions to transport across the blood-brain barrier.
No correlation was found between susceptibility and blood cobalt or chromium levels for any brain region in the implant group.
Only four subjects were found to have cobalt levels above the toxicity threshold, defined as 7 μg/L. 61,62Of these four, one had levels above the toxicity threshold, with serum cobalt of 40 μg/L, but still showed comparable susceptibility results to the rest of the cohort, suggesting a nonlinear relationship between blood ion levels and brain metal deposition.We believe that this nonlinear relationship arises, most likely, from the blood-brain barrier that regulates components into the brain.Similarly, it might be possible thatrisk of developing diseases appears at 100 μg/L, or more, as suggested in 12 and that cobalt concentrations slightly above the risk threshold could beordinary for implanted patients.2][63][64][65][66] Studies looking into metal ion transport across barriers found that the presence of metal-on-metal implants is associated with significantly higher plasma ion levels, but not cerebrospinal fluid levels and that cerebrospinal fluid cobalt concentrations are influenced by plasma concentrations of cobalt in a nonlinear fashion. 3though there are important differences between blood-brain barrier and blood-cerebrospinal fluid barrier, we can derive information about metal ion transport across barriers and into the nervous system.[69][70] No significant differences in terms of cognition were observed between the implant and control groups.Furthermore, there were no differences in the number of subjects in each cognitive impairment group.Even for the four subjects above the 7 μg/L blood cobalt toxicity threshold, two demonstrated mild cognitive impairment, and two had no cognitive impairment.Prior studies suggest that rarely will patients with even failed primary hip replacements develop neurological damage.2][73][74][75][76] The presence of an implant alone does not seem to make patients vulnerable to visuospatial or memory impairments; however, studies have raised the concern for psychological and neurological symptoms in patients with long-term metal bearings. 32Previous studies have reported the various sources of potential causality between increased ion levels and neurocognitive symptoms, such as ion-induced genotoxicity, altered brain metabolism, protein accumulation, impaired-iodine uptake, and metal-induced polyneuropathy. 4,5,10,23,48,51However, those cases are not from well-functioning implants.Further research in confirming the mechanisms behind elevated ion levels and neurocognitive impairment is therefore indicated.
This study has limitations that should be acknowledged.First, we assumed that control patients without implants had negligible levels of cobalt and chromium in their bloodstream; however, we did not collect blood samples from the control group due to the invasive nature of the procedure.Potential differences in environmental or other exposure to these ions due to occupation, smoking, or from medicine and vitamins may be present, thus introducing confounders.However, due to the local environment homogeneity anddemographic matching of our cohorts, we can reasonably assume comparable metal ion exposure due to the environment and industrial activities, pollution levels, and metal metabolism.A second limitation is that the study included a diverse range of devices with various implant bearings, making the results more generalizable but less specific to a particular implant type.
This study included a variety of devices across multiple combinations of implant bearings, so the results are not specific to a certain implant type, though this may be more generalizable than studying a single device.Third, the image segmentation of brain regions using QSM or R2* maps directly was not readily obtainable, so segmentation was performed on T1-weighted MR images instead.This introduces the possibility of errors in the rigid registration between susceptibility and T1-weighted images, as well as residual distortions that may not have been fully corrected.However, it is important to note that the imaging pipeline was automated, and rigorous quality and accuracy checks were conducted.QSM, R2* and T1-weighted images were acquired during the same session, minimizing potential systematic errors that could affect all examinations.8][79][80][81] Lastly, although the CSF has been reported to be a suitable reference for QSM since it is not affected by age and some diseases, the CSF may be affected by cobalt and chromium concentrations. 82However, the robustness of the CSF to cobalt and chromium concentrations indicated by Harrison-Brown et al. 3 suggests a CSF rate of change significantly lower than those of other brain structures.To test this hypothesis, we standardized our results to brain structures of minimum susceptibility variance across subjects and with no reference at all, 29 and found no significant differences in our results.While these limitations should be considered, efforts were made to minimize potential biases and errors throughout the study.Future work could include the development of an in vitro QSM phantom with anticipated cobalt and chromium blood or serum ion levels to further assess the accuracry of susceptibility-based measurements as previously presented in iron and calcium concentrations, 23,48 as well as provide a suitable and specific calibration phantom during the imaging process.Including ratios against iron and myelin may also need to be considered as they provide the most abundant source of paramagnetic and diamagnetic content in the brain.
In conclusion, no differences were found in brain tissue susceptibility or cognition in subjects with long-term and wellfunctioning THA and HRA implants in comparison to age-and sexmatched controls.There was no correlation observed between susceptibility measurements or cognition scores with blood ion levels of cobalt or chromium.This finding suggests that the relationship between metal ions in the bloodstream and the central nervous systemfor well-functioning and long-term implants are within biologically acceptable ranges.Therefore, patients with well-functioning longterm hip implants are not at higher risk of compromising their brain integrity due to metal deposition in the brain.

AUTHOR CONTRIBUTIONS
Shahnaz Taleb and Vishal Kalia conducted data analysis and interpretation.Gabriel Varela-Mattatall contributed to the interpretation of the results.Abbigail Allen directed data acquisition.Roy Haast, Demographic details were recorded for each subject including age, sex, height, weight, and body mass index.Device details including model and duration of implantation were also recorded for the subjects with a THA or HRA.All patients underwent the Montreal Cognitive Assessment (MoCA) to evaluate cognitive impairment.Blood samples were collected for the implanted group at the day of imaging to measure whole blood cobalt and chromium ion levels.Samples were collected as per standard of care for metal ion testing, with certified metal-free EDTA-Royal Blue tubes specifically used for trace metal testing.All participants recruited in the implant group completed SF-12, Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC), and Harris Hip assessments to gauge functionality of the implanted joint.
Pulse sequence specifications were as follows: 2300.0 ms repetition time (TR); echo time (TE) of 2.98 ms; 9°flip angle; and 256 × 256 × 192 mm field of view.These images were used to for segmentation of regions of interest, identification of cortex volumes, and for assessment by a radiologist.For QSM and R2*, a multiecho gradient echo (ME-GRE) sequence (0.5 × 0.5 × 2.0 mm 3 resolution) was used to generate maps.Sequence parameters were as follows: TR of 52.0 ms; TE 1 to TE 5 of 8.08, 14.56, 21.05, 27.53, 34.01, and 40.49ms; 20°flip angle, and a field of view of 224 × 200 × 190 mm.

7. 0
μg/L.Two were metal-on-metal THA (40 and 13.5 μg/L), one was a metal-on-metal HRA (10.9 μg/L), and one was a metal-onpolyethylene THA with a dual-modular neck (10.4 μg/L).No subjects had chromium levels above 7 μg/L.No implanted or control subjects had severe (MoCA score < 10) or moderate (MoCA score 10-17) scores.Nine subjects in the implant group were found to have mild cognitive impairment (MoCA score 18-25) compared to 14 in the control group.The difference was not significant (p = 0.2684).A moderate negative correlation was found between blood cobalt concentration and MoCA scores (R 2 = 0.2026, p = 0.0232), but no correlation was found between blood chromium concentration and MoCA (Figure5).The correlation with cobalt was no longer presentwhen the high cobalt outlier patient (40.42 ug/L) was excluded (R 2 = 0.0479, p = 0.1840).No correlation was found between

1
Comparison of susceptibility (ppb) between nine selected anatomical regions of interest in the (A) left and (B) right brains of subjects with implants at long-term follow-up (n = 28) and age-and sex-matched control subjects without implants (n = 32).polyethylene bearingsperformed at our institution found mean SF-12 Mental Health Composite Scores of 52.8 ± 9.7, mean Physical Composite Scores of 40.6 ± 11.8, mean WOMAC scores of 79.7 ± 21.0, and mean Harris Hip scores of 92.4 ± 10.0. 36Comparing

F I G U R E 6
Correlation between Montreal Cognitive Assessment (MoCA) scores and quantitative MRI in the brain region with highest QSM or R2* values, where (A) and (B) depict correlation between susceptibility (ppb) in the pallidum and MoCA scores in the implanted (p = 0.8676) and control (p = 0.6647) subjects, respectively, and (C) and (D) present correlation between R2* (s −1 ) in the cerebellum and MoCA scores in the implanted (p = 0.6842) and control (p = 0.4768) subjects, respectively.
a Mann-Whitney test, two-tailed.b Fisher's exact test, two-sided.
14 ± 718.34 and 4532.99 ± 477.96 in the putamen, 1803.98 ± 298.12, and 1908.42 ± 330.70 in the pallidum, 4161.99 ± 519.68 and 4047.99 ± 577.24 in the hippocampus, 3244.48 ± 568.25 and 3284.37 ± 362.84 in the caudate, 1505.39 ± 234.91 and 1524.12 ± 152.01, and 10,970.74± 1507.07 and 10,402.57± 1336.85 in the midfrontal, 3266.16 ± 589.92 and 2517.38 ± 708.97 in the inferior temporal, 31,892.68± 3683.72 and 32,531.52 ± 3748.89 in the cerebellum, and 3612.74 ± 608.99 and 2626.40 ± 2332.24 in the anterior cingulate regions of interest, in the implant and control groups, respectively.Qualitative assessment of imaged brains showed no pertinent findings.Most implanted subjects displayed normal anatomical features for their age; however, three subjects displayed chronic ischemic changes, with one showing concern for pituitary lesions.Control subjects showed similar age-related ischemic changes.