Hip resurfacing arthroplasty at a minimum of 14 years: revision is concentrated among female patients with dysplasia and small components

Abstract

Introduction

Hip resurfacing arthroplasty (HRA) is a bone-preserving option for young, active patients. The influence of the underlying diagnosis on long-term outcome remains unclear, particularly the distinction between primary osteoarthritis (OA), developmental dysplasia of the hip (DDH) and non-OA conditions.

Materials and methods

We retrospectively reviewed 104 consecutive metal-on-metal HRAs in 101 patients (three bilateral) performed by one surgeon between 2006 and 2011, stratified into Primary OA (n = 50), DDH (n = 29) and Non-OA (n = 25: avascular necrosis [AVN] n = 15, inflammatory arthritis n = 5, post-traumatic arthritis n = 3, Legg-Calvé-Perthes disease n = 2). AVN was analysed additionally as an independent group, and exploratory analyses restricted to female patients and to cups ≤ 50 mm addressed the collinearity of dysplasia, sex and size. Survivorship used Kaplan-Meier estimation with revision as the endpoint.

Results

Implant status was known for all 104 hips; the 91 unrevised assessed hips had a mean follow-up of 16.2 ± 1.5 years. Overall 15-year survivorship was 95.2%. Primary OA, Non-OA and AVN each showed 100%; DDH 82.8% (versus Primary OA p = 0.002, versus Non-OA p = 0.031). All five revisions occurred in female DDH patients with a cup ≤ 50 mm; DDH survivorship remained lower within female patients (p = 0.025) and within small cups (p = 0.006), with no event in any comparator arm. Final HHS was lower in DDH (89.5 ± 8.5) than in Primary OA (93.9 ± 6.8) or Non-OA (95.0 ± 4.9; p = 0.012), a difference below the minimal clinically important improvement threshold of 15.9 to 18 points. The Co/Cr ratio did not differ between groups (p = 0.37).

Conclusions

HRA achieved 100% 15-year survivorship in primary OA and in non-OA diagnoses including AVN. Dysplasia may be associated with an increased risk of revision, but dysplasia, female sex and small component size were too collinear, with no revision outside the dysplastic group, for their independent contributions to be separated.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00402-026-06481-3.

Introduction

Hip resurfacing arthroplasty (HRA) has been established as a bone-preserving alternative to total hip arthroplasty (THA) for young, active patients with advanced hip disease [1]. The procedure preserves the femoral head and neck, maintaining near-normal biomechanics and facilitating future revision to a standard THA if required [2]. Long-term studies report 20-year survivorship of 96.5% in men and 87.0% in women after Birmingham Hip Resurfacing [3], falling to 89.5% and 66.9% respectively at 25 years [4].

Patient selection remains the cornerstone of successful HRA. Multiple studies have identified risk factors for failure, including female sex and smaller component size [5], and underlying developmental dysplasia of the hip (DDH) [6, 7]; Subbiah Ponniah et al. found, however, that dysplasia alone was not a contraindication to resurfacing once sex and component size were accounted for [5]. However, most previous studies have grouped DDH together with primary osteoarthritis (OA) as a single “OA” category, potentially obscuring the contribution of dysplasia to implant failure. These risk factors are, however, strongly inter-related: dysplastic hips are predominantly female and receive smaller components, so series of this size cannot readily separate them. Furthermore, the long-term outcomes of HRA for non-OA etiologies such as avascular necrosis (AVN) remain less well-characterized, with concerns regarding femoral component loosening in necrotic bone [8].

Concurrent with survivorship concerns, the safety profile of metal-on-metal (MoM) bearings has been scrutinized due to adverse reactions to metal debris (ARMD), including pseudotumor formation and elevated systemic metal ion levels [9, 10]. Long-term surveillance data incorporating serum metal ions and MARS-MRI are essential for a comprehensive assessment of HRA outcomes.

The purpose of this study was to report the long-term outcomes of HRA at a minimum 14-year follow-up, with hips stratified into three etiological groups: Primary OA, DDH, and Non-OA, and with AVN additionally examined as an independent diagnostic group. We hypothesized that dysplastic hips would show inferior survivorship compared with primary OA and non-OA etiologies. We additionally performed post hoc exploratory subgroup analyses to examine whether any such difference could be attributed to dysplasia independently of female sex and small component size.

Materials and methods

Study design and patient selection

This retrospective cohort study was approved by the institutional review board. The requirement for written informed consent was waived owing to the retrospective nature of the study. We reviewed a consecutive series of 104 metal-on-metal HRAs performed in 101 patients by a single senior surgeon between January 2006 and November 2011 at a tertiary referral center. Three patients underwent staged bilateral HRA (six hips); all remaining patients contributed a single hip. The unit of analysis was the hip for survivorship, clinical and radiographic outcomes, and the patient for serum metal ion levels. All procedures utilized the Durom Hip Resurfacing System (Zimmer, Warsaw, IN, USA) via a posterolateral approach.

Inclusion criteria were: (1) symptomatic hip arthritis refractory to conservative management, (2) age 18 to 65 years, (3) adequate bone quality assessed by dual-energy X-ray absorptiometry (DEXA T-score > -1.5), (4) a femoral head considered suitable for resurfacing on preoperative templating; a diameter of 42 mm or more was preferred, although smaller components were accepted where bone quality and head morphology were judged adequate (implanted femoral components ranged from 36 to 54 mm), and (5) absence of large femoral head cysts (> 1 cm). Exclusion criteria included: (1) known metal hypersensitivity, (2) severe renal impairment (eGFR < 60 mL/min), (3) severe osteoporosis, and (4) active infection.

Completeness of follow-up

Implant status was ascertained for every hip in the series. Eight patients (8 hips) agreed to be contacted but declined to attend for further clinical, radiographic and laboratory assessment; each was interviewed by telephone in February 2026, confirmed that the resurfaced hip had not been revised, and reported no hip-related symptoms. Their status was additionally cross-referenced against the national administrative health and mortality databases, which confirmed that none had undergone revision at another institution. These eight hips therefore contribute to the survivorship analysis, censored at the date of the February 2026 contact, but have no final Harris Hip Score, UCLA Activity Score, radiographic measurement, serum metal ion result or MARS-MRI, and are excluded from those analyses. Because implant status was known for all 104 hips, no worst-case imputation for hips lost to follow-up was required.

Surgical technique and postoperative care

All procedures were performed using a standard posterolateral approach without navigation. The acetabular component was implanted press-fit with a target inclination of 40° to 45° and anteversion of 15° to 20°. The femoral component was cemented using high-viscosity cement. Postoperatively, patients were allowed full weight-bearing as tolerated on the first postoperative day. Deep vein thrombosis prophylaxis consisted of low-molecular-weight heparin or aspirin for 14 days, and intravenous antibiotics (cefazolin) were administered for 24 h. Patients were advised to avoid high-impact sports for the first 6 months to allow for complete osteointegration.

Group classification

Hips were stratified into three groups based on the primary underlying diagnosis. The Primary OA Group (n = 50) comprised hips with primary (idiopathic) osteoarthritis without evidence of underlying structural abnormality. The DDH Group (n = 29) comprised hips with osteoarthritis secondary to developmental dysplasia, classified according to the Crowe classification (Crowe type I, n = 16; type II, n = 13). The Non-OA Group (n = 25) comprised non-degenerative etiologies, including avascular necrosis (AVN, n = 15), inflammatory arthritis (n = 5), post-traumatic arthritis (n = 3), and Legg-Calvé-Perthes disease (n = 2).

Because the Non-OA group combines diagnoses with different pathologies and different indications for HRA, a secondary analysis was performed in which AVN (n = 15) was treated as an independent diagnostic group and the remaining ten non-OA hips were reported descriptively as “other non-OA”. This four-group classification is reported alongside the three-group classification throughout.

For AVN hips, the extent of femoral head involvement was assessed using the Kerboul combined necrotic angle on preoperative anteroposterior and lateral radiographs. Only hips with a Kerboul angle ≤ 230° were considered candidates for HRA. Ficat staging was also documented.

Clinical and radiographic assessment

Patients were evaluated clinically at 6 weeks, 3 months, 6 months, 1 year, and annually thereafter. Clinical outcomes were assessed using the Harris Hip Score (HHS) and the University of California Los Angeles (UCLA) Activity Score. Return to sporting activities was documented, categorized as low-impact (swimming, cycling), moderate-impact (tennis, skiing), or high-impact (running, martial arts, team sports). For the five hips that underwent revision, the functional scores reported are those recorded at the last assessment before revision surgery.

Standard anteroposterior pelvis and lateral hip radiographs were obtained at each visit. Radiographic assessment included: (1) component loosening, defined as progressive radiolucent lines > 2 mm in any zone or component migration > 3 mm, with acetabular radiolucent lines recorded according to the three zones described by DeLee and Charnley [11]; the > 2 mm and > 3 mm thresholds were selected because they exceed the reported measurement error of plain-radiographic assessment; (2) osteolysis, defined as a well-defined lytic area ≥ 5 mm not present on immediate postoperative radiographs; (3) femoral neck narrowing, measured as a percentage decrease from the immediate postoperative neck width; and (4) heterotopic ossification, graded according to the Brooker classification.

Cup inclination and anteversion were measured on the immediate postoperative radiograph and at the latest available radiograph, with correction for radiographic magnification and projection. Serial cup orientation was not re-measured at every interval visit, and change in orientation over time was therefore not used as a criterion for loosening; this is stated as a limitation. Component position was additionally classified as inside or outside the Lewinnek safe zone (inclination 40 ± 10°, anteversion 15 ± 10°).

Metal ion and MRI surveillance

Serum cobalt (Co) and chromium (Cr) levels were measured using inductively coupled plasma mass spectrometry (ICP-MS) at the most recent follow-up visit. A threshold of 7 µg/L was applied to either ion, by analogy with the 7 ppb alert level of MHRA Medical Device Alert MDA/2012/036 [12]; because that alert specifies whole-blood sampling whereas our assays were performed on serum, the two are not directly interchangeable, and this is acknowledged as a limitation. Because serum metal ions are a patient-level rather than a hip-level measurement, all ion results are reported per patient (n = 93); a single value is assigned to the patient and is shared by both hips of a bilateral patient in any hip-level display, and bilateral and unilateral patients were compared directly. The Co/Cr ratio was calculated for each patient as an index of bearing wear characteristics.

Metal artifact reduction sequence MRI (MARS-MRI) was performed in patients with: (1) elevated metal ion levels (Co or Cr > 7 µg/L), (2) unexplained hip pain or symptoms, or (3) as part of routine surveillance in consenting patients. Pseudotumours were classified on MARS-MRI according to the system of Hauptfleisch et al. [13], as Type I (thin-walled cystic mass, wall < 3 mm), Type II (thick-walled cystic mass) or Type III (predominantly solid mass). Abductor muscle integrity was also assessed.

Statistical analysis

Analyses were performed in Python 3.9 (pandas 2.3, scipy 1.13, statsmodels 0.14, lifelines 0.30). Continuous variables were compared using one-way analysis of variance with Tukey’s honestly significant difference test for pairwise comparisons; because several distributions were skewed, Kruskal-Wallis tests are reported alongside and any discrepancy between the two is stated explicitly. Categorical variables were compared using the chi-square test, with Fisher’s exact test (Freeman-Halton extension for 2 × 3 tables) used for the sparsest tables, in which almost all expected cell counts fell below five; where a chi-square test is reported for a table containing an expected count below five, the corresponding exact test was also computed and did not alter any interpretation. Implant survivorship was estimated by the Kaplan-Meier method with revision for any reason as the endpoint and 95% confidence intervals computed on the complementary log-log scale; groups were compared by the log-rank test. Hips that had not been revised were censored at the date of last documented contact — the last clinical assessment for the 96 assessed hips, and the February 2026 telephone contact for the eight hips that declined further assessment. Competing-risks analysis was not performed, as there were no deaths during the study period.

Three post hoc analyses were also performed. First, survivorship and outcomes were recalculated with AVN as an independent group. Second, exploratory survivorship analyses were restricted to female patients and, separately, to hips receiving a cup ≤ 50 mm, using a clinically motivated threshold rather than a cut-point derived from these data. Third, the sensitivity of the results to bilateral cases was assessed by repeating the analyses at the patient level. Multivariable Cox regression was attempted but was not estimable: because no revision occurred in any group other than DDH, the models exhibited monotone likelihood, and Firth-penalised estimation returned hazard ratios with confidence intervals spanning several orders of magnitude. No adjusted estimate is therefore reported, and all subgroup comparisons are presented as hypothesis-generating rather than confirmatory. Between-group differences in HHS were interpreted against the published minimal clinically important improvement threshold of 15.9 to 18 points [14]. A p-value of < 0.05 was considered statistically significant. The reporting of this observational study follows the STROBE statement (Supplementary File 1).

Results

Patient demographics

Implant status was ascertained for all 104 hips in 101 patients. Ninety-six hips (92.3%) in 93 patients underwent full clinical and radiographic assessment; the 91 that had not been revised were assessed at a mean of 16.2 ± 1.5 years (range, 14.2 to 19.4 years), so every surviving hip satisfied the minimum 14-year criterion, while the five revised hips were followed to revision at 4.2 to 12.1 years. The remaining eight hips belonged to patients who declined further assessment and were interviewed by telephone in February 2026 (3 Primary OA, 4 DDH, 1 Non-OA); none had undergone revision (Fig. 1).

Fig. 1.

Patient Flow Diagram. Selection, stratification into three etiological groups and follow-up of 104 hips in 101 patients, showing the three staged bilateral procedures, the avascular necrosis subgroup (n = 15) analysed separately, and the eight hips whose implant status was confirmed by telephone in February 2026 without further assessment

The demographic characteristics are summarized in Table 1. The DDH group had a significantly higher proportion of female patients (82.8%) compared to the Primary OA (30.0%) and Non-OA (28.0%) groups (p < 0.001). The Non-OA group was significantly younger (42.3 ± 7.3 years) than both the Primary OA (58.2 ± 4.5 years) and DDH (55.4 ± 7.2 years) groups (p < 0.001). The DDH group had significantly smaller mean acetabular cup sizes (48.6 ± 3.6 mm) compared to Primary OA (54.2 ± 4.8 mm) and Non-OA (52.8 ± 4.2 mm) groups (p < 0.001), and a significantly higher proportion of hips receiving a cup ≤ 50 mm (58.6% vs. 26.0% and 40.0%, p = 0.016). Perioperative metrics showed no significant differences among the groups; the mean operating time was 84.6 ± 15.9 min and the mean estimated blood loss 250 ± 81 mL. There were no deaths during the follow-up period.

Table 1.

Patient Demographics and Baseline Characteristics

Characteristic	Primary OA (n = 50 hips)	DDH (n = 29 hips)	Non-OA (n = 25 hips)	p-value
Patients / hips	49 / 50	28 / 29	24 / 25	—
Bilateral HRA, n patients	1	1	1	—
Age at surgery, years (mean ± SD)	58.2 ± 4.5	55.4 ± 7.2	42.3 ± 7.3	< 0.001
Sex, n (%)				< 0.001
Male	35 (70.0%)	5 (17.2%)	18 (72.0%)
Female	15 (30.0%)	24 (82.8%)	7 (28.0%)
Body Mass Index, kg/m² (mean ± SD)	26.1 ± 3.0	24.8 ± 2.6	25.4 ± 2.8	0.17
Follow-up, years (mean ± SD), unrevised hips	16.1 ± 1.6	17.0 ± 1.8	15.7 ± 0.9	—
Component size, mm (mean ± SD)				< 0.001
Acetabular cup	54.2 ± 4.8	48.6 ± 3.6	52.8 ± 4.2
Femoral head	48.2 ± 4.8	42.6 ± 3.6	46.8 ± 4.2
Cup ≤ 50 mm, n (%)	13 (26.0%)	17 (58.6%)	10 (40.0%)	0.016
Operating time, min (mean ± SD)	86.3 ± 14.6	85.2 ± 17.1	80.4 ± 16.7	0.31
Estimated blood loss, mL (mean ± SD)	236 ± 80	258 ± 77	267 ± 85	0.25
Etiology-specific features
Kerboul angle (AVN only; n = 15)	N/A	N/A	185° ± 24°	N/A
Ficat stage III (AVN only; n = 15)	N/A	N/A	9 (60.0%)	N/A
Crowe type I / II (DDH only)	N/A	16 (55.2%) / 13 (44.8%)	N/A	N/A

OA, osteoarthritis; DDH, developmental dysplasia of the hip; SD, standard deviation; N/A, not applicable. Kerboul angle and Ficat stage were assessed only in the 15 hips with avascular necrosis (AVN) within the Non-OA group. Three patients underwent staged bilateral HRA (one in each group), so 104 hips were performed in 101 patients

Among the 15 AVN hips, the mean Kerboul combined necrotic angle was 185° ± 24° (range, 139° to 222°). Nine hips (60%) were classified as Ficat stage III and six (40%) as Ficat stage II.

Three patients underwent staged bilateral HRA (one in each group); none of these six hips was revised, and repeating the analyses after retaining only the first-operated hip did not change any conclusion.

Implant survivorship

The overall 15-year Kaplan-Meier survivorship with revision for any reason as the endpoint was 95.2% (95% CI: 88.8–98.0) (Fig. 2). Both the Primary OA group and the Non-OA group demonstrated 100% survivorship at 15 years. In contrast, the DDH group had a significantly lower 15-year survivorship of 82.8% (95% CI: 63.4–92.4) (log-rank: DDH vs. Primary OA, p = 0.002; DDH vs. Non-OA, p = 0.031; overall three-group comparison, p = 0.001) (Table 2). Corresponding DDH survivorship was 96.6% (95% CI 77.9–99.5) at 5 years and 89.7% (95% CI 71.3–96.5) at 10 years.

Fig. 2.

Kaplan–Meier Survivorship Analysis. Survival curves with revision for any reason as the endpoint. The DDH group demonstrated significantly lower survivorship compared to both the Primary OA (p = 0.002) and Non-OA (p = 0.031) groups. Numbers at risk are shown beneath the curves. Panel B shows the four-group analysis with avascular necrosis as an independent group

Table 2.

Kaplan-Meier Survivorship by Etiological Group (endpoint: revision for any reason) — Panel A, three etiological groups; Panel B, avascular necrosis analysed as an independent group

Follow-up	Overall (n = 104)	Primary OA (n = 50)	DDH (n = 29)	Non-OA (n = 25)
Panel A — Three etiological groups
5-year survivorship (95% CI)	99.0% (93.4–99.9)	100%	96.6% (77.9–99.5)	100%
Number at risk	103	50	28	25
10-year survivorship (95% CI)	97.1% (91.3–99.1)	100%	89.7% (71.3–96.5)	100%
Number at risk	101	50	26	25
15-year survivorship (95% CI)	95.2% (88.8–98.0)	100%	82.8% (63.4–92.4)	100%
Number at risk	80	40	18	22
Revisions, n	5	0	5	0
Revision rate / 100 person-years	0.30	0.00	1.12	0.00

Follow-up	Primary OA (n = 50)	DDH (n = 29)	AVN (n = 15)	Other non-OA (n = 10)
Panel B — Avascular necrosis as an independent group
5-year survivorship	100%	96.6% (77.9–99.5)	100%	100%
10-year survivorship	100%	89.7% (71.3–96.5)	100%	100%
15-year survivorship	100%	82.8% (63.4–92.4)	100%	100%
Number at risk at 15 years	40	18	12	10
Revisions, n	0	5	0	0

CI, confidence interval. Log-rank (Panel A): Primary OA vs. Non-OA, p = 1.00; DDH vs. Primary OA, p = 0.002; DDH vs. Non-OA, p = 0.031; overall three-group comparison, p = 0.001. Log-rank (Panel B): overall four-group comparison, p = 0.003; DDH versus AVN, p = 0.095, reflecting the small size of the AVN group. No revision occurred in the AVN or other non-OA groups; with zero events in these strata no hazard ratio is estimable and the comparison with DDH is underpowered. “Other non-OA” comprises inflammatory arthritis (n = 5), post-traumatic arthritis (n = 3) and Legg-Calvé-Perthes disease (n = 2). Numbers at risk are the number of hips whose implant survival time reaches the stated time point. All 104 hips contribute to the survivorship analysis: unrevised hips were censored at the date of last documented contact, which for the eight hips that declined further assessment was the February 2026 telephone contact. Because implant status was known for every hip, no worst-case imputation was required

Five hips required revision surgery, all within the DDH group (5/29 = 17.2%). The overall revision rate was 0.30 per 100 person-years; group-specific rates were 0.00 for both Primary OA and Non-OA and 1.12 for DDH. The reasons for revision were aseptic acetabular component loosening (n = 3, at 4.2, 7.8 and 12.1 years), adverse reaction to metal debris (n = 1, at 9.8 years) and periprosthetic femoral fracture (n = 1, at 10.5 years). All five failures occurred in female patients with cup sizes ≤ 50 mm (42, 44, 48, 48 and 50 mm) and Crowe type I or II dysplasia (type I, n = 2; type II, n = 3). Case-level detail is given in Table 3.

Table 3.

Details of the Five Revised Hips

Case	Sex	Crowe	Cup / head (mm)	Inclination / anteversion (°)	Lewinnek zone	Time to revision (y)	Reason
1	F	II	50 / 44	47.9 / 18.2	Inside	4.2	Aseptic acetabular loosening
2	F	II	48 / 42	43.6 / 23.5	Inside	7.8	Aseptic acetabular loosening
3	F	I	44 / 38	43.1 / 15.1	Inside	10.5	Periprosthetic femoral fracture
4	F	I	48 / 42	40.8 / 14.3	Inside	9.8	ARMD
5	F	II	42 / 36	45.2 / 17.5	Inside	12.1	Aseptic acetabular loosening

F, female; ARMD, adverse reaction to metal debris. The Lewinnek safe zone is defined as inclination 40 ± 10° and anteversion 15 ± 10°. All five revised hips were in female patients with Crowe type I or II dysplasia and received a cup of 50 mm or smaller

When AVN was analysed as an independent group, no revision occurred in either the AVN group (n = 15) or the remaining non-OA diagnoses (n = 10); 15-year survivorship was 100% in both (four-group log-rank p = 0.003). The direct comparison of DDH with AVN alone did not reach significance (82.8% vs. 100%, p = 0.095), reflecting the small size of the AVN group; this comparison is underpowered (Table 2, Panel B).

Exploratory subgroup analyses are summarised in Table 4. Among female patients only (46 hips), 15-year survivorship was 79.2% in DDH versus 100% in non-dysplastic female hips (p = 0.025). Among hips receiving a cup ≤ 50 mm (40 hips), it was 70.6% versus 100% (p = 0.006). The 23 hips that were both in female patients and received a cup ≤ 50 mm accounted for all five revisions and had a 15-year survivorship of 78.3%, whereas the remaining 81 hips had none. Because no comparator subgroup contained an event, hazard ratios were not estimable and the independent contributions of dysplasia, female sex and component size cannot be separated in this cohort.

Table 4.

Exploratory Subgroup Survivorship Analyses

Subgroup	Hips (n)	Revisions (n)	15-year survivorship (95% CI)	Log-rank p
Female patients only	46	5	89.1% (75.8–95.3)	—
Female, DDH	24	5	79.2% (57.0–90.8)	0.025
Female, non-DDH	22	0	100%
Cup ≤ 50 mm only	40	5	87.5% (72.5–94.6)	—
Cup ≤ 50 mm, DDH	17	5	70.6% (43.1–86.6)	0.006
Cup ≤ 50 mm, non-DDH	23	0	100%
Female AND cup ≤ 50 mm	23	5	78.3% (55.4–90.3)	< 0.001
All other hips	81	0	100%

Every comparator subgroup contains zero revision events; consequently hazard ratios are not estimable, multivariable Cox regression exhibits complete separation, and these comparisons are hypothesis-generating only

Radiographic outcomes

Radiographic analysis revealed that the mean cup inclination was 43.5° ± 4.2° and mean anteversion 18.2° ± 3.5°, with 91.7% (88/96) of components positioned within the Lewinnek safe zone, with no significant differences among the three groups (inclination p = 0.87, anteversion p = 0.49, proportion within the safe zone p = 0.22). All five revised hips had cups within the Lewinnek safe zone, and inclination and anteversion in revised hips (44.1 ± 2.6° and 17.7 ± 3.6°) did not differ from unrevised hips (43.5 ± 4.2° and 18.3 ± 3.6°; p = 0.75 and p = 0.74). Given that 91.7% of the whole series lay within the zone, however, the observation that all five revised hips did so is what would be expected by chance (Fisher’s exact p = 1.00) and therefore excludes nothing. Radiolucent lines > 2 mm were observed in 5.2% of hips (5/96), predominantly in the DDH group (16.0% [4/25] vs. 2.1% [1/47] in Primary OA and 0% [0/24] in Non-OA; Freeman-Halton exact p = 0.029 across the three groups). Osteolysis was rare, occurring in 2 hips in the DDH group (8.0%, Fisher’s exact p = 0.13). Heterotopic ossification (Brooker grade I or II) was observed in 12.5% of hips (12/96), with no cases of grade III or IV. Significant femoral neck narrowing (> 10% decrease) was observed in 8.3% of hips (8/96), predominantly asymptomatic, and did not differ significantly between groups (p = 0.65).

Functional outcomes

The mean HHS improved substantially from preoperative values in all three groups (Table 5); preoperative scores did not differ between groups (Primary OA 42.2 ± 8.2, DDH 44.4 ± 7.3, Non-OA 44.6 ± 9.6; p = 0.42). At final follow-up, the Primary OA group achieved a mean HHS of 93.9 ± 6.8, the Non-OA group 95.0 ± 4.9 and the DDH group 89.5 ± 8.5 (p = 0.012 between groups; Kruskal-Wallis p = 0.021) (Fig. 3). Post-hoc Tukey analysis revealed that the DDH group had significantly lower final HHS compared to both Primary OA (p = 0.032) and Non-OA (p = 0.016) groups. The absolute differences were 4.4 points versus Primary OA and 5.6 points versus Non-OA, both well below the minimal clinically important improvement threshold of 15.9 to 18 points for the HHS [14]; the difference is therefore statistically significant but of limited clinical importance.

Table 5.

Functional and Radiographic Outcomes

Outcome Measure	Primary OA	DDH	Non-OA	p-value
Hips assessed, n (patients)	47 (46)	25 (24)	24 (23)	—
Hips with functional scores, n	47	25	24	—
Harris Hip Score — pre-operative	42.2 ± 8.2	44.4 ± 7.3	44.6 ± 9.6	0.42
Harris Hip Score — final follow-up	93.9 ± 6.8	89.5 ± 8.5	95.0 ± 4.9	0.012*
UCLA Activity Score (final)	7.8 ± 1.1	6.5 ± 1.9	8.1 ± 1.0	< 0.001*
Return to high-impact sports, n (%)	22/47 (46.8%)	4/25 (16.0%)	13/24 (54.2%)	0.012*
Radiolucent lines > 2 mm, n (%)	1/47 (2.1%)	4/25 (16.0%)	0/24 (0%)	0.029*
Osteolysis, n (%)	0/47 (0%)	2/25 (8.0%)	0/24 (0%)	0.13
Femoral neck narrowing > 10%, n (%)	5/47 (10.6%)	2/25 (8.0%)	1/24 (4.2%)	0.65
Heterotopic ossification (Brooker I–II), n (%)	6/47 (12.8%)	3/25 (12.0%)	3/24 (12.5%)	1.00
Cup inclination, ° (mean ± SD)	43.5 ± 4.2	43.8 ± 2.9	43.2 ± 5.2	0.87
Cup anteversion, ° (mean ± SD)	18.4 ± 3.8	17.5 ± 3.5	18.6 ± 3.0	0.49
Within Lewinnek safe zone, n (%)	44/47 (93.6%)	24/25 (96.0%)	20/24 (83.3%)	0.22
Revisions for any reason, n (%) of hips implanted	0/50 (0%)	5/29 (17.2%)	0/25 (0%)	0.002*
Aseptic acetabular loosening	0	3	0
Periprosthetic femoral fracture	0	1	0
Adverse reaction to metal debris	0	1	0

*p < 0.05 across three groups. The comparison of revision proportions between groups uses Fisher’s exact test (Freeman–Halton), p = 0.002; the chi-square value is p = 0.001. Functional, radiographic and safety outcomes use the number of hips actually assessed (47/25/24) rather than the number implanted (50/29/25), while the crude revision rate correctly uses the number implanted. For the five revised hips the last pre-revision functional score is used. Post-hoc Tukey comparisons of final Harris Hip Score: DDH vs. Primary OA p = 0.032, DDH vs. Non-OA p = 0.016, Primary OA vs. Non-OA p = 0.77; corresponding Kruskal-Wallis p = 0.021. Osteolysis and radiolucent-line comparisons use Fisher’s exact test (Freeman-Halton) because expected cell counts were below five. Survivorship is reported in Table 2. Corresponding four-group values (AVN separate): final HHS 94.2 ± 5.3 (AVN, n = 14) and 96.2 ± 4.3 (other non-OA, n = 10); UCLA 8.1 ± 0.9 and 8.0 ± 1.2

Fig. 3.

Harris Hip Score Comparison. Pre-operative and final follow-up HHS by group, for the 96 assessed hips; for the five revised hips the last pre-revision score is used. Error bars are one standard deviation

The Non-OA group demonstrated the highest UCLA Activity Scores (8.1 ± 1.0) compared to Primary OA (7.8 ± 1.1) and DDH (6.5 ± 1.9) (p < 0.001; Kruskal-Wallis p = 0.006; Tukey DDH vs. Primary OA p < 0.001, DDH vs. Non-OA p < 0.001). Participation in high-impact sports was 54.2% (13/24) in Non-OA, 46.8% (22/47) in Primary OA and 16.0% (4/25) in DDH (p = 0.012) (Table 5).

In the four-group analysis, final HHS was 94.2 ± 5.3 in AVN (n = 14) and 96.2 ± 4.3 in the other non-OA diagnoses (n = 10), with UCLA Activity Scores of 8.1 ± 0.9 and 8.0 ± 1.2 respectively.

Metal ion levels and MRI findings

Serum metal ion levels were obtained in all 93 assessed patients (Table 6). The DDH group had higher mean serum cobalt levels (3.0 ± 2.3 µg/L) compared to Primary OA (2.2 ± 1.3 µg/L) and Non-OA (1.8 ± 0.9 µg/L) (p = 0.023); this difference was not robust to a rank-based test (Kruskal-Wallis p = 0.16) and should therefore be interpreted cautiously. Three patients (3.2%) had cobalt levels exceeding 7 µg/L (two DDH, one Primary OA) and one (DDH) exceeded the threshold for chromium. Mean chromium did not differ significantly between groups (2.1 ± 1.8, 1.6 ± 0.8 and 1.4 ± 0.5 µg/L for DDH, Primary OA and Non-OA; p = 0.061). The Co/Cr ratio was 2.4 in DDH, 1.9 in Primary OA and 1.6 in Non-OA, with no significant difference (ANOVA p = 0.37; Kruskal-Wallis p = 0.61), indicating that the higher absolute cobalt concentrations in the DDH group were not accompanied by a disproportionate shift in the cobalt-to-chromium relationship.

Table 6.

Safety Outcomes: Serum Metal Ions and MARS-MRI Findings

Safety Measure	Primary OA	DDH	Non-OA	p-value
Patients tested, n	46	24	23	—
Hips represented, n	47	25	24	—
Serum cobalt, µg/L (mean ± SD)	2.2 ± 1.3	3.0 ± 2.3	1.8 ± 0.9	0.023*
Serum chromium, µg/L (mean ± SD)	1.6 ± 0.8	2.1 ± 1.8	1.4 ± 0.5	0.061
Co/Cr ratio (mean)	1.9	2.4	1.6	0.37
Co/Cr ratio (median)	1.37	1.06	1.15	0.61 (Kruskal-Wallis)
Patients with Co > 7 µg/L, n (%)	1/46 (2.2%)	2/24 (8.3%)	0/23 (0%)	0.23
Patients with Cr > 7 µg/L, n (%)	0/46 (0%)	1/24 (4.2%)	0/23 (0%)	0.23
MARS-MRI screened hips, n	42	25	20	—
Pseudotumor, n (% of screened)	2 (4.8%)	4 (16.0%)	0 (0%)	0.08
Hauptfleisch Type I / II / III	1 / 1 / 0	2 / 1 / 1	0 / 0 / 0
Abductor muscle atrophy, n (% of screened)	4 (9.5%)	6 (24.0%)	1 (5.0%)	0.11
Serum cobalt, bilateral vs. unilateral patients	Bilateral (n = 3): 5.6 ± 3.2 µg/L	Unilateral (n = 90): 2.2 ± 1.4 µg/L		0.033*

*p < 0.05. Serum metal ions are a patient-level measurement and are reported per patient (n = 93 assessed patients). Group ANOVA for cobalt is p = 0.023 at patient level, but the difference is not robust to a rank-based test (Kruskal-Wallis p = 0.16). Four-group values (AVN separate): mean cobalt 1.6 ± 0.7 µg/L in AVN; no pseudotumour in any of the 12 screened AVN hips

The three patients who underwent bilateral HRA had markedly higher serum cobalt concentrations (mean 5.6 µg/L) than unilateral patients (2.2 µg/L; Mann-Whitney p = 0.033), consistent with a cumulative systemic burden from two bearings, and supporting the decision to report ion concentrations at the patient rather than the hip level.

MARS-MRI was performed in 87 hips (90.6%; Primary OA n = 42, DDH n = 25, Non-OA n = 20). Pseudotumors were identified in 6 hips (6.9% of those screened): 4 in the DDH group (16.0%), 2 in the Primary OA group (4.8%) and none in the Non-OA group (p = 0.08). Half of the pseudotumours were Hauptfleisch Type I and remained asymptomatic. Abductor muscle atrophy was noted in 11 hips (12.6% of those screened), with a higher prevalence in the DDH group (24.0%) compared to Primary OA (9.5%) and Non-OA (5.0%) (p = 0.11) (Table 6). No pseudotumor was detected in any of the 12 screened AVN hips.

Discussion

This study provides long-term evidence that HRA outcomes differ according to the underlying etiology, with all revisions in this series occurring in dysplastic hips. By separating DDH from primary OA—a distinction not consistently made in prior literature—we demonstrate that both primary OA and non-OA etiologies (including AVN analysed separately) achieve excellent 15-year survivorship of 100%, while DDH carries a significantly elevated failure risk (15-year survivorship 82.8%, p = 0.002). It is important to state what these data can and cannot support. All five revisions occurred in female patients with dysplasia who received a cup ≤ 50 mm, and no revision occurred anywhere else in the cohort. Because these three characteristics were almost completely collinear, no multivariable model was estimable and dysplasia cannot be identified as an independent risk factor from these data. Our exploratory analyses do show that dysplastic hips fared worse than non-dysplastic hips both within female patients (p = 0.025) and within hips receiving small cups (p = 0.006), which is consistent with a contribution of dysplasia beyond sex and size alone; however, the comparator subgroups contained no events at all, and these comparisons must be treated as hypothesis-generating. Furthermore, to our knowledge, this is among the first long-term (≥ 14 years) HRA series reported in an Asian cohort. Asian patients tend to have smaller native anatomy and a higher prevalence of DDH compared to Caucasian populations, making these findings particularly relevant. These observations are consistent with a large international multicentre HRA cohort of 11,382 hips from 27 centres in 13 countries, in which dysplasia-related diagnoses had the worst long-term implant survival (78.3%, 95% CI 76.5–80.1, at 20 years, versus a 22-year survivorship for osteoarthritis whose 95% CI was 92.1–93.3; p < 0.001) [7] (Table 7).

Table 7.

Comparison of Long-Term Survivorship in Selected Hip Resurfacing Cohorts

Study	Implant	N (hips)	Mean follow-up	Etiology / cohort	Survivorship	Pseudotumour / ARMD
Current study	Durom	104 implanted; 96 assessed	16.2 y	Primary OA / DDH / Non-OA (AVN n = 15 analysed separately)	15-y: 100% (OA) / 82.8% (DDH) / 100% (Non-OA) / 100% (AVN)	4.8% (OA) / 16.0% (DDH) / 0% (Non-OA), screened hips
Dhawan et al. [3]	BHR	234 (217 patients)	20.9 y	Mixed indications	20-y: 96.5% male / 87.0% female	0% pseudotumour on ultrasound at 20 y
Nandra et al. [4]	BHR	144 (130 patients)	25 y	Mixed (designer-surgeon, all indications)	25-y: 83.5% overall / 89.5% male / 66.9% female	ARMD 4/144 (2.8%) at 25 y
Subbiah Ponniah et al. [5]	BHR	243 (205 patients)	11.2 y	Non-designer surgeon (93.8% OA)	10-y: 94.8%; 17-y aseptic: 100% male / 89.6% female	Elevated ions implicated in 50% of female revisions
Van Der Straeten et al. [7]	Mixed (18 designs)	11,382 (≤ 50 y at surgery)	7.6 y	International multicentre	22-y: 88.9% overall; OA 22-y 95% CI 92.1–93.3%; 20-y dysplasia 78.3%	Adverse local tissue reaction a common revision cause
Amstutz & Le Duff [8]	Conserve Plus	99 (82 patients)	Mean 10.8 y (2–18)	Osteonecrosis only	15-y: 90.3%	No wear-related failures
Hartmann et al. [21]	Durom	134 (121 patients)	8.5 y	Mixed consecutive Durom cohort	Mid-term survival evaluated	16.5% (17/103 hips screened)

Survivorship endpoints vary by study (revision for any reason unless otherwise stated). The “current study” row reports the AVN subgroup separately. All cited values were cross-checked against the published abstracts; items not exposed by the abstract are referred to the full text rather than stated as specific figures

Our findings are consistent with evidence identifying DDH as a specific risk factor in HRA. Gaillard and Gross [6] studied 363 HRAs performed for developmental dysplasia and showed that failure in dysplastic hips is modifiable by technique rather than inevitable, reporting substantially improved survivorship after a series of technical interventions including an intraoperative radiographic alignment guideline. Shanaa et al. [15], in a systematic review and meta-analysis pooling 1,006 dysplastic hips, reported 93% pooled survivorship and a 13% complication rate; the difference in revision risk between dysplastic and osteoarthritic HRA cohorts did not reach statistical significance, although the trend favoured primary OA. Our data, with longer follow-up than most of the pooled studies, are compatible with that trend. The anatomical challenges of DDH—a shallow acetabulum, increased anteversion and smaller femoral head size—make optimal component positioning more technically demanding [6], and dysplastic cohorts show poorer survivorship in both large multicentre series and pooled analyses [7, 15]. In the present series, however, we found no radiographic support for a malposition mechanism: cup inclination and anteversion did not differ between groups (p = 0.87 and p = 0.49) or between revised and unrevised hips (p = 0.75 and p = 0.74). All five revised hips lay within the Lewinnek safe zone, but since 91.7% of the whole series did so, this is uninformative rather than exculpatory (Fisher’s exact p = 1.00). Two considerations bear on this. First, the Lewinnek zone was derived for conventional total hip arthroplasty and may not define an acceptable position for a large-diameter metal-on-metal bearing, in which edge loading can occur despite nominally acceptable inclination and anteversion. Second, we did not assess acetabular version relative to the native dysplastic anatomy, contact-patch-to-rim distance, or three-dimensional component position. Component orientation as an explanation is therefore neither supported nor excluded by our data, and we present it only as a hypothesis. Smaller component diameter, itself a consequence of dysplastic anatomy and female sex, remains the more parsimonious mechanical explanation (Tables 3 and 4).

The 100% survivorship in the Non-OA group, despite 60% of AVN hips being Ficat stage III, is noteworthy. Analysed as an independent group, AVN retained 100% 15-year survivorship, a final HHS of 94.2 ± 5.3 and a UCLA Activity Score of 8.1 ± 0.9, confirming that the favourable outcome of the Non-OA group is not an artefact of pooling. No pseudotumour was identified in any screened AVN hip. This addresses earlier concerns about femoral component loosening in necrotic bone and is consistent with Amstutz and Le Duff [8], who reported the first series of hip resurfacing for osteonecrosis with 15-year survivorship data (99 hips in 82 patients; 15-year survivorship 90.3%) and concluded that patients with advanced osteonecrosis are excellent candidates for HRA — notably having resurfaced hips regardless of the size of the necrotic lesion. We attribute our favorable outcome to rigorous patient selection: all AVN hips had a Kerboul combined necrotic angle ≤ 230° (observed range 139° to 222°), ensuring adequate viable bone stock. With only 15 AVN hips and no events, however, the direct comparison of AVN with DDH did not reach significance (p = 0.095); the 100% figure is precise only to the limits of a small denominator.

The higher activity levels in the Non-OA group (UCLA 8.1, high-impact sports 54.2%) reflect the younger age and activity demands of this population. Previous studies have shown that HRA allows a high rate of return to sports, including high-impact activities such as triathlon [16, 17], and a recent systematic review of physical activity and implant integrity concluded that patients after either resurfacing or total hip arthroplasty should be counselled towards moderate activity [18]. The preservation of proximal femoral bone stock likely contributes to these outcomes [19, 20].

The lower final Harris Hip Score in the dysplasia group deserves careful interpretation. The difference is real and statistically significant (89.5 versus 93.9 and 95.0 points; p = 0.012), but its magnitude — 4.4 to 5.6 points — is roughly a quarter to a third of the minimal clinically important improvement threshold of 15.9 to 18 points reported for the HHS in a primary total hip arthroplasty cohort of 2,667 patients [14], and all three groups achieved mean scores of 89 points or above, placing them in the highest (81 to 100) HHS category of that cohort — the stratum against which scores below 55 carried a significantly increased revision hazard. The clinically important difference between the groups is therefore the risk of revision rather than the function of the surviving implant. Activity level showed a larger separation (UCLA 6.5 in dysplasia versus 7.8 and 8.1; p < 0.001), which we attribute principally to age and sex distribution. Because the five revised hips contribute their last pre-revision scores, this comparison also incorporates the functional decline that preceded revision.

Regarding metal ion surveillance, the DDH group demonstrated higher cobalt levels and a greater prevalence of pseudotumors than the other groups. This is plausibly attributable to the smaller component sizes in that group (mean cup 48.6 mm), since a smaller bearing generates a greater volumetric wear burden for a given linear wear rate; we did not measure wear directly and offer this only as a mechanism consistent with the observation. This inference should be made cautiously: the cobalt difference did not survive a rank-based test (p = 0.16), and the Co/Cr ratio did not differ between groups (p = 0.37). A disproportionate elevation of cobalt relative to chromium has been proposed as an indicator of accelerated bearing wear or of a taper-related source; its absence here suggests that the higher absolute cobalt concentrations reflect greater overall wear volume from a smaller bearing rather than a qualitatively different wear mechanism. We also observed that the three patients with bilateral resurfacings had substantially higher serum cobalt than unilateral patients (5.6 vs. 2.2 µg/L, p = 0.033), reinforcing that ion concentrations are a patient-level measurement. The Durom implant used in this study, although no longer commercially available, provides valuable long-term data on MoM bearing performance. Hartmann et al. [21] reported pseudotumours in 17 of 103 screened hips (16.5%) in a consecutive Durom series at a mean of 8.5 years. Our overall rate of 6.9% is considerably lower, although our DDH subgroup rate of 16.0% approaches theirs; our MARS-MRI screening was selective rather than universal, which would if anything bias our apparent rate upward, so the difference is unlikely to be an artefact of ascertainment. A clinical review of ARMD revision surgery reports that outcomes appear to have improved with time, in part because of regular patient surveillance and a lowered threshold for performing revision [10].

Long-term follow-up of the Birmingham Hip Resurfacing demonstrates that male patients can achieve survival approaching 90% at 25 years (89.5%, all indications), whereas females fare substantially worse (66.9%) [4]; our data showing 100% survival in male-predominant Primary OA and Non-OA groups versus 82.8% in the female-predominant DDH group supports the same pattern, and underscores that female sex, dysplasia and small component size act as overlapping rather than independent risk factors. This is also consistent with the long-term single-surgeon, non-designer series of Subbiah Ponniah et al. [5], in which aseptic survival was 100% in men and 89.6% in women at 17 years, albeit with only six female hips still at risk at that point. Based on these findings we recommend a risk-stratified surveillance approach: patients with DDH, female sex or cup sizes ≤ 50 mm should undergo annual clinical evaluation and metal ion testing, supplemented by MARS-MRI if ions rise or symptoms develop; male patients with primary OA or AVN and larger components may transition to less frequent surveillance after the first decade. In our cohort the highest-risk phenotype—female sex combined with a cup ≤ 50 mm—comprised 23 of 104 hips and contained every revision, which offers a practical way to target surveillance.

This study has several limitations. First, it is a retrospective, single-surgeon, single-implant (Durom) series, which limits generalizability to contemporary ceramic-on-ceramic resurfacing designs. Second, and most importantly, the study is limited by the number of events. Only five revisions occurred, all in one group. With zero events in the Primary OA, AVN and other non-OA groups, the reported 100% survivorship figures carry wide confidence limits, no hazard ratio is estimable and no multivariable adjustment is possible; the power to compare long-term outcomes between diagnostic groups is correspondingly low, and a genuine difference between the smaller groups could have gone undetected. Third, the DDH, AVN and other non-OA groups are small (29, 15 and 10 hips) and differ substantially in age, sex and component size; because these covariates are inherent to the etiologies and highly collinear, we could not disentangle the independent contribution of dysplasia, and have avoided any claim that we have done so. Fourth, cup orientation was measured on plain radiographs at two time points and serial change was not systematically re-measured, so the definition of loosening rests on progressive radiolucent lines and component migration alone; radiostereometric analysis and computed-tomography-based version measurement were not available. Fifth, MARS-MRI screening was selective, which may over-estimate pseudotumor prevalence in the screened subgroup. Sixth, metal ions were measured in serum, whereas the 7 µg/L MHRA alert level is defined for whole blood, so our threshold comparison is an analogy rather than a direct application of that guidance. Seventh, eight hips have implant status confirmed by telephone but no final functional, radiographic or laboratory data and no clinical assessment beyond 2.2 to 14.5 years, so their contribution is to the survivorship analysis alone, which rests on 104 hips, whereas the functional, radiographic and laboratory analyses rest on 96. Eighth, three patients underwent bilateral HRA, and their ion values are patient-level; hip-level ion comparisons should be interpreted with this dependency in mind. Ninth, the superior outcomes in the Non-OA group may be partially attributable to younger age and larger component sizes rather than etiology alone. Tenth, vital status was ascertained from the clinical record and the national mortality database, and no death was identified during the study period; we cannot exclude that a death was missed, and competing-risks analysis was therefore not performed. Eleventh, the minimal clinically important improvement threshold we use for interpretation was derived for within-patient change and is applied here to a between-group difference in final scores, which is an approximation. Twelfth, no adjustment was made for multiple comparisons; the subgroup and secondary analyses are unadjusted and are presented as hypothesis-generating. Finally, the Durom implant is no longer available; however, the principles of patient selection and the long-term behavior of MoM bearings remain relevant to current practice. As newer ceramic-on-ceramic resurfacing designs emerge, the lessons of the MoM era—specifically the importance of caution in dysplastic hips requiring small components—will be essential for bone-preserving arthroplasty.

Conclusion

HRA provides excellent long-term survivorship (100% at 15 years) in both primary OA and non-OA etiologies at a minimum 14-year follow-up, and this remains true when avascular necrosis is analysed as an independent group. Dysplasia may be associated with an increased risk of revision (15-year survivorship 82.8%), but all five revisions occurred in female patients with dysplasia who received a cup ≤ 50 mm, and dysplasia, female sex and small component size were too closely related in this cohort—with no revision occurring outside the dysplastic group—for their independent contributions to be separated. Final Harris Hip Scores were lower in the dysplasia group, but by only about a quarter to a third of the minimal clinically important improvement threshold, so the clinically important distinction between the groups is the risk of revision rather than the function of a surviving implant. Rigorous patient selection, particularly regarding AVN extent (Kerboul angle ≤ 230°) and dysplasia severity, together with risk-stratified surveillance of female patients receiving small components, is essential. These findings support the continued use of HRA in appropriately selected young, active patients, regardless of whether the underlying diagnosis is primary OA or AVN.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Yu-Chih Lin — Conceptualization; Methodology; Investigation; Data curation; Writing – original draft; Writing – review & editing. Chih-Hsiang Chang — Conceptualization; Methodology; Writing – review & editing; Supervision. Pang-Hsin Hsieh — Conceptualization; Methodology; Writing – review & editing; Supervision. Yu-Yi Huang — Conceptualization; Project administration; Writing – review & editing. Dave W. Chen — Formal analysis; Software; Visualization. Ngi-Chiong Lau — Conceptualization; Methodology; Project administration; Writing – review & editing; Supervision. All authors have read and approved the final manuscript.

Funding

One of the authors (Yu-Chih Lin) was supported by Chang Gung Memorial Hospital (Year 1: CORPG5N0021; Year 2: CORPG5N0022; Year 3: CMRPG5N0113) for study design, monitoring, and interpretation.

Data availability

The datasets generated and/or analysed during the current study are not publicly available due to patient privacy considerations but are available from the corresponding author on reasonable request.

Declarations

Conflict of interest

The authors declare no competing interests.

Ethical approval and consent to participate

This retrospective cohort study was approved by the Institutional Review Board of Chang Gung Memorial Hospital (IRB No. 202200675B0). The requirement for written informed consent was waived by the IRB owing to the retrospective nature of the study.