Results

After data extraction, 128 TKAs and 64 THAs were eligible for inclusion. One patient (with 2 TKAs and 1 THA) was lost to follow-up after being transferred to another hospital in 2018 and was censored at that moment. On average, for TKA, 5% of data was missing across 14 variables, and for THA, 13% was missing across 9 variables. Patient and perioperative characteristics are presented in Table 1.

Characteristic | TKA | THA
Patient characteristics
Number of joints (n) | 128 | 64
Hemophilia |  | 
Severe A | 100 (78%) | 49 (77%)
Severe B | 13 (10%) | 6 (9%)
Moderate A | 12 (9%) | 9 (14%)
VWD | 3 (2%) | 0 (0%)
Age at time of surgery (y) | 51 (26-77) | 57 (27-77)
BMI (kg/m2) | 25.7 (18.3-35.4)a(n = 69) | 24.7 (18.9-42.2)a(n = 41)
Had surgery after 2000 | 82 (64%) | 45 (70%)
Treatment regimen |  | 
On demand | 47 (37%)a(n = 126) | 20 (31%)
Factor prophylaxis | 77 (61%)a(n = 126) | 40 (63%)
Emicizumab | 2 (6%)a(n = 126) | 4 (6%)
Hepatitis C positive | 78 (62%)a(n = 126) | 34 (53%)
HIV positive | 12 (10%)a(n = 126) | 3 (5%)
Inhibitor presentb | 3 (3%)a(n = 108) | 3 (5%)a(n = 57)
Petterson score before first surgery | 48 (3-78) | 49 (0-78)
Multiple joint procedure |  | 
With ankle arthrodesis | 15 (12%) | 1 (2%)
Bilateral | 24 (19%) | 4 (6%)
Bilateral and ankle arthrodesis | 12 (9%) | 0 (0%)
TKA and THA (unilateral) | 6 (5%) | 6 (9%)
TKA, THA, and ankle arthrodesis | 1 (1%) | 1 (2%)
Perioperative characteristics
Δ Hb (preop vs postop, mmol/L) | 2.1 (0.5-5.9) | 2.2 (0.7-4.9)
Length of hospital stay (d) | 17 (4-148) | 12 (7-111)
Preoperative angiography | 85 (67%)a(n = 126) | NA
Cemented femoral stem | NA | 26 (43%)a(n = 60)
Weightbearing protocol |  | 
Full | 54 (43%)a(n = 126) | 41 (75%)a(n = 55)
Partial | 43 (34%)a(n = 126) | 13 (24%)a(n = 55)
No weightbearing | 29 (23%)a(n = 126) | 1 (2%)a(n = 55)

Most patients had severe hemophilia A (TKA 78%, THA 77%) and were treated with factor prophylaxis (TKA, 61%, THA 63%). Median age at surgery was 51 years for TKA and 57 years for THA. Positivity for hepatitis C was present in 62% and 53%, HIV in 10% and 5%, and inhibitors in 3% and 5%, respectively. The degree of arthropathy was already severe before the first surgery, with median Pettersson scores of 48 (TKA) and 49 (THA). In 36% (n = 71) of the cases, the surgery was part of a multiple joint procedure, with bilateral procedures (TKA 19%, THA 6%) and combinations with ankle arthrodesis (TKA 12%, THA 2%) performed most often. Half of the patients who underwent TKA were allowed full weightbearing on the prosthesis (43%), although this was more common after THA (75%). In 67% (n = 85) of the TKAs, preoperative angiography was performed. The femoral stem was cemented in 40% (n = 26) of the THAs.

Postoperative bleeding occurred in 32% of the TKAs (n = 40); furthermore, TKA was complicated by both PJI (6%) and loosening (10%). In THAs, postoperative bleeding occurred in 18% (n = 10), a PJI was present in 3% of the cases, while loosening occurred in 16%. Complication rates did not differ significantly before and after 2000 (P > .05) and are presented in Table 2.

Complication | TKA | THA
n = 128 | n = 64
Postoperative bleeding | 32% (24%-41%)a(n=126) | 18% (9%-30%)a(n = 55)
Prosthetic joint infection | 6% (3%-12%)a(n = 126) | 3% (0.4%-11%)a(n = 61)
Loosening | 10% (6%-17%)a(n=127) | 16% (8%-28%)a(n = 62)
 | nB = 46, nA = 82 | nB = 19, nA = 45
Postoperative bleeding |  | 
Before 2000 | 28% (16%-44%) | 38% (10%-51%)
After 2000 | 33% (23%-44%) | 11% (4%-25%)
Fisher’s exact P value | .69 | .15
Prosthetic joint infection |  | 
Before 2000 | 11% (4%-24%) | 5% (0%-28%)
After 2000 | 6% (2%-14%) | 2% (0%-13%)
Fisher’s exact P value | .49 | .51
Loosening or liner wear on X-ray |  | 
Before 2000 | 7% (2%-19%) | 26% (10%-51%)
After 2000 | 12% (6%-22%) | 11% (4%-25%)
Fisher’s exact P value | .38 | .15

Prothesis survival rates

The PSR for TKA in people with bleeding disorders was 92.2% (nR = 71; 95% CI, 87.5-97.3) after 15 years and 90.9% (nR = 28; 95% CI, 85.5-96.6) after 25 years. The PSR for THA in people with bleeding disorders was 91.0% (nR = 27; 95% CI, 82.8-100) after 15 years and 78.6% (nR = 14; 95% CI, 64.8-95.5) after 25 years. The survival curves are presented in Figures 1 and 2. Individual reasons for failure are presented in Table 3.

Figure 1: Survival curve for primary total knee arthroplasty in people with bleeding disorders (red line), with 5-, 10-, and 15-year general population survival rates shown as gray dots.

Figure 2: Survival curve for primary total hip arthroplasty in people with bleeding disorders (red line), with 5-, 10-, and 15-year general population survival rates shown as gray dots.

Patient No. | Total knee arthroplasty (n = 128) | Total hip arthroplasty (n = 64)
Failure reason | Year of surgery | Years to failure | Failure reason | Year of surgery | Years to failure
1 | Liner wear | 1990 | 31 | Loosening | 1998 | 31
2 | Liner wear | 1992 | 29 | Loosening | 1992 | 16
3 | Liner wear | 1994 | 29 | Liner wear | 1993 | 26
4 | Infection | 1995 | 2 | Loosening | 1995 | 19
5 | Loosening | 1996 | 11 | Loosening | 2001 | 14
6 | Infection | 1996 | 16 | Fracture | 2003 | 22
7 | Infection | 1999 | 3 | Dislocation | 2008 | 0
8 | Disabling functional impairment | 2001 | 4 | Loosening | 2013 | 10
9 | Infection | 2009 | 4 | Infection | 2013 | 1
10 | Infection | 2012 | 1 |  |  | 
11 | Infection | 2017 | 5 |  |  | 
12 | Disabling functional impairment | 2017 | 6 |  |  | 
13 | Infection | 2017 | 0 |  |  | 

Among failed TKAs (n = 13), early failure (within 5 years, n = 8) was caused by PJI in 86% of the cases, while in cases of late failure (after 5 years, n = 5), this was only in 17%. Among failed THAs (n = 9), early failure was very low (n = 2), while almost every late failure (n = 7) was due to loosening (n = 6, 86%).

Predictors for prosthesis failure

Univariable Cox PH regression analysis for TKA showed that patients with an inhibitor at time of surgery had 16.3 times higher risk for prosthesis failure (P = .03) , and patients with PJI at any time during follow-up had 35.9 times higher risk (P < .001) (¹). A trend was observed such that patients with moderate hemophilia and those with VWD had 4.0 times (P = .07) higher risk for prosthesis failure, and patients receiving on-demand treatment had 3.1 times (P = .09) higher risk. In the multivariable analysis, PJI (P = .03) was the only significant predictor (Table 4). In a post hoc multivariable analysis with only preoperative variables available, inhibitor status (P = .08) was the sole variable showing a trend (Table 5). Therefore, PJI is assumed to mediate the association between inhibitor status and PSR; patients with an inhibitor have an increased risk of PJI, and PJI increases the risk of prosthetic failure. In addition, on-demand therapy, nonsevere hemophilia, and inhibitor status are known to be interrelated.

Variable | Exponent | 95% CI | P
Total knee arthroplasty (n = 128) |  |  | 
Severity of disorder0 = severe1 = nonsevere | 0.28 | 0.045-1.79 | .23
Treatment regimen0 = on demand1 = prophylaxis | 0.65 | 0.16-2.58 | .57
Inhibitor status0 = absent1 = present | 2.90 | 0.25-33.57 | .47
Prosthetic joint infections0 = no1 = yes | 14.39 | 2.59-79.98 | .03a
Total hip arthroplasty (n = 64)
Age | 1.06 | 0.99-1.14 | .14
Continuous
Cemented femoral stem0 = yes1 = no | 3.57 | 0.78-16.67 | .16

Variable | Exponent | 95% CI | P
Severity of disorder | 0.39 | 0.08-1.92 | .29
0 = severe
1 = nonsevere
Treatment regimen | 0.46 | 0.13-1.62 | .27
0 = on demand
1 = prophylaxis
Inhibitor status | 14.83 | 0.56-141.3 | .08
0 = absent
1 = present

Univariable Cox PH regression analysis for THA showed a trend toward higher risk of prosthesis failure in older people than in younger people (P = .07; ¹). Each additional year of age at the time of surgery increased the risk by 1.08. A trend toward higher risk for prosthesis failure was also observed for patients with a cemented femoral stem (P = .09). These patients had 4.5 times higher risk than patients with an uncemented femoral stem. In multivariable Cox PH analysis, no significant predictors were observed (Table 4). Age and cementation were moderately correlated (r = 0.6), showing that older patients more often received a cemented femoral stem. Because neither variable remained significant in the multivariable model, this may indicate a mediating relationship: age influences the choice for cementation, which is subsequently associated with risk of failure. This is supported by age not being a significant predictor in a subset with only uncemented THA (P = .74).

Sensitivity analysis

After selecting every first TJR in the dataset, all patient and perioperative values and long-term complication rates in the sensitivity analysis were within 5% difference of those in the primary analysis, even all PSRs. For both TKA and THA, the same predictors were observed (data not shown).

Footnotes

1. The online version contains supplementary material available at https://doi.org/10.1016/j.rpth.2026.106887. ↩ ↩²