Predictors of periprosthetic fracture after total knee replacement

Figures & data

Table 1. Demographic features of study cohort as mean (SD) or n (%), unless specified otherwise

	Primary TKA	Revision TKA
	(n = 17,633)	(n = 4,090)
	Mean (SD)	Mean (SD)
	or n (%)	or n (%)
Mean follow-up, years (SD)	6.3 (4.7)	5.1 (4.2)
Female	9,781 (55%)	2,084 (51%)
Bilateral	4,719 (27%)	804 (20%)
Mean age at surgery, years (SD)	68.4 (10.0)	67.9 (11.3)
Age category
≤ 60 years	3,352 (19%)	915 (22.4%)
61–70 years	6,206 (35.2%)	1,286 (31.4%)
71–80 years	6,493 (36.8%)	1,476 (36.1%)
>80 years	1,582 (9%)	413 (10.1%)
Body mass index (BMI), kg/m^2 (SD)	31.2 (6.2)	31.1 (6.5)
BMI category, kg/m^2
Missing	67 (0.4%)	36 (0.9%)
Normal, < 25.0	2,362 (13.4%)	584 (14.4%)
Overweight, 25–29.9	5,961 (33.9%)	1,363 (33.6%)
Obese, 30–39.9	7,710 (43.9%)	1,731 (42.7%)
Morbidly obese, ≥ 40.0	1,533 (8.7%)	376 (9.3%)
ASA class ^a
1	291 (1.7%)	50 (1.2%)
2	9,614 (54.7%)	1,869 (45.8%)
3	7,544 (42.9%)	2,098 (51.4%)
4	117 (0.7%)	61 (1.5%)
Mean Deyo-Charlson index (SD)	1.5 (2.2)	1.3 (2.1)
Deyo-Charlson index
0	7,985 (45.3%)	2,086 (51%)
1	3,598 (20.4%)	836 (20.4%)
2	2,291 (13%)	470 (11.5%)
3+	3,759 (21.3%)	698 (17.1%)
Previous cardiac event
No	14,404 (81.7%)	3,490 (85.3%)
Yes	3,229 (18.3%)	600 (14.7%)
Previous thromboembolic event
No	16,869 (95.7%)	3,873 (94.7%)
Yes	764 (4.3%)	217 (5.3%)
Operative diagnosis (primary TKR)
Osteoarthritis	16,372 (92.8%)
Rheumatoid arthritis	657 (3.7%)
Avascular necrosis	152 (0.8%)
Other ^b	604 (3.4%)
Operative diagnosis (revision TKR)
Failure: loose/wear/osteolysis		2,154 (52.7%)
Failure: previous surgery with
components removed		801 (19.6%)
Failure: fracture, dislocation		877 (21.4%)
Failure: non-union		27 (0.7%)
Failure: infection		228 (5.6%)
Implant fixation		Not applicable
Uncemented	1,555 (8.8%)
Cemented or hybrid*	16,078 (91.2%)

^a American Society of Anesthesiologists (ASA) score was missing in 67 (0.4%) of primary TKR patients and 12 (0.3%) of revision TKR patients

^b Other diagnoses for primary TKA included: genu varum, genu valgum, hemophilia, Paget’s disease, failed previous disease including arthrodesis, failed previous osteotomy, failed previous patellectomy, Chacot arthropathy, chondromalacia, pigmented villonodular synovitis; data on operative diagnosis were missing for 3 patients with revision TKR.

Table 2. Univariate and multivariable-adjusted risk of periprosthetic fracture following primary total knee arthroplasty

	Total	Periprosthetic fractures	Univariate hazard	Multivariable ^a model 1 hazard	Multivariable ^b model 2 hazard	Multivariable ^c model 3 hazard
Variable	(n = 17,633)	(n = 188)	ratio (95% CI)	ratio (95% CI)	ratio (95% CI)	ratio (95% CI)
Sex			p = 0.2			p = 0.2
Male	7,852	74 (1%)	1.00 (ref)			1.00 (ref)
Female	9,781	114 (1%)	1.23 (0.91–1.64)			1.24 (0.92–1.68)
Age category			p < 0.001	p < 0.001	p = 0.001	p = 0.002
≤ 60	3,352	52 (2%)	1.00 (ref)	1.00 (ref)	1.00 (ref)	1.00 (ref)
61–70	6,206	59 (1%)	0.52 (0.36–0.76)	0.51 (0.35–0.74)	0.52 (0.36–0.76)	0.52 (0.36–0.77)
71–80	6,493	58 (1%)	0.56 (0.38–0.81)	0.55 (0.38–0.80)	0.55 (0.37–0.80)	0.54 (0.36–0.80)
> 80	1,582	19 (1%)	0.97 (0.57–1.65)	0.97 (0.57–1.64)	0.94 (0.55–1.61)	0.90 (0.52–1.58)
Body mass index, kg/m^2			p = 0.4			p = 0.5
Normal, < 25.0	2,362	29 (1%)	1.00 (ref)			1.00 (ref)
Overweight, 25–29.9	5,961	75 (1%)	0.77 (0.50–1.18)			1.09 (0.70–1.68)
Obese, 30–39.9	7,710	71 (1%)	0.94 (0.49–1.81)			0.83 (0.54–1.30)
Morbidly obese, ≥ 40.0	1,533	13 (1%)	0.99 (0.64–1.51)			0.87 (0.44–1.71)
Deyo-Charlson index			p = 0.06		p = 0.05	p = 0.04
0	7,985	75 (1%)	1.00 (ref)		1.00 (ref)	1.00 (ref)
1	3,598	42 (1%)	1.27 (0.87–1.86)		1.21 (0.82–1.79)	1.22 (0.82–1.81)
2	2,291	25 (1%)	1.35 (0.86–2.13)		1.39 (0.88–2.19)	1.42 (0.90–2.26)
3+	3,759	46 (1%)	1.66 (1.15–2.40)		1.68 (1.16–2.44)	1.75 (1.18–2.59)
Previous cardiac event			p = 0.2			p = 0.07
No	14,404	166 (1%)	1.00 (ref)			1.00 (ref)
Yes	3,229	22 (1%)	0.76 (0.49–1.19)			0.65 (0.40–1.04)
Previous thromboembolism			p = 0.6			p = 0.6
No	16,869	181 (1%)	1.00 (ref)			1.00 (ref)
Yes	764	7 (1%)	1.21 (0.57–2.58)			1.20 (0.56–2.57)
Operative diagnosis			p = 0.2		p = 0.2	p = 0.3
Osteoarthritis	16,372	159 (1%)	1.00 (ref)		1.00 (ref)	1.00 (ref)
Rheumatoid arthritis	657	13 (2%)	1.13 (0.28–4.57)		1.16 (0.64–2.11)	1.01 (0.55–1.87)
Avascular necrosis	152	2 (1%)	1.73 (0.99–3.01)		–	–
Other	452	14 (3%)	1.51 (0.85–2.65)		1.57 (0.94–2.65)	1.55 (0.92–2.61)
Implant fixation			p = 0.3			p = 0.6
Uncemented	1555	14 (1%)	1.00 (ref)			1.00 (ref)
Cemented or hybrid	16,078	174 (1%)	0.76 (0.44–1.31)			0.87 (0.50–1.51)
ASA class			p = 0.5			p = 0.6
1	291	5 (2%)	1.00 (ref)			1.00 (ref)
2	9,614	97 (1%)	0.64 (0.26–1.58)			0.75 (0.30–1.86)
3	7,544	83 (1%)	0.79 (0.32–1.94)			0.93 (0.36–2.37)
4	117	1 (1%)	0.66 (0.08–5.68)			0.80 (0.09–6.99)

^a Multivariable model 1 was a Cox regression model that considered all significant variables from the univariate analyses with a p-value of < 0.05–and retained only those significantly associated in the multivariable model by using a backward-selection process; the c-statistic of the multivariable model was 0.58.

^b Multivariable model 2 was a Cox regression model obtained by retaining all variables from the univariate analyses with a p-value of < 0.20 and forcing all of them in the model; the c-statistic of the multivariable model was 0.61.

^c Multivariable model 3 was a Cox regression model obtained by retaining all variables from the univariate analyses regardless of p-value and forcing all of them in the model; the c-statistic of the multivariable model was 0.62.

ASA: American Society of Anesthesiologists; CI: confidence interval; ref: reference category.

Table 3. Multivariable-adjusted hazard of periprosthetic fracture after revision total knee arthroplasty

	Total	Periprosthetic fractures	Univariate hazard	Multivariable ^a model 1 hazard	Multivariable ^b model 2 hazard	Multivariable ^c model 3 hazard
Variable	(n =4,090)	(n = 104)	ratio (95% CI)	ratio (95% CI)	ratio (95% CI)	ratio (95% CI)
Sex			p= 0.09		p = 0.06	p = 0.06
Male	2,006	42 (2%)	1.00 (ref)		1.00 (ref)	1.00 (ref)
Female	2,084	62 (3%)	1.40 (0.9–2.07)		1.47 (0.98–2.21)	1.48 (0.99–2.22)
Age category			p = 0.3			p = 0.3
≤60	915	27 (3%)	1.00 (ref)			1.00 (ref)
61-70	1,286	38 (3%)	0.84 (0.51–1.38)			0.82 (0.49–1.36)
71-80	1,476	34 (2%)	0.68 (0.41–1.14)			0.68 (0.40–1.16)
>80	413	5 (1%)	0.44 (0.17–1.15)			0.44 (0.16–1.18)
Body mass index, kg/m^2			p = 0.1		p = 0.2	p = 0.4
Normal, < 25.0	584	16 (3%)	1.00 (ref)		1.00 (ref)	1.00 (ref)
Overweight, 25-–29.9	1,363	35 (3%)	0.89 (0.50–1.60)		1.04 (0.58–1.87)	1.05 (0.58–1.91)
Obese, 30–39.9	1,731	39 (2%)	1.93 (0.92–4.02)		1.99 (0.95–4.19)	1.81 (0.85–3.86)
Morbidly obese, ≥4 0.0	376	13 (3%)	0.97 (0.54–1.76)		1.14 (0.62–2.08)	1.20 (0.65–2.22)
Deyo-Charlson index			p = 0.05		p = 0.08	p = 0.08
0	2,086	46 (2%)	1.00 (ref)		1.00 (ref)	1.00 (ref)
1	836	19 (2%)	1.07 (0.63–1.83)		1.07 (0.62–1.83)	1.08 (0.62–1.87)
2	470	15 (3%)	1.56 (0.87–2.80)		1.53 (0.85–2.75)	1.56 (0.85–2.85)
3+	698	24 (3%)	1.94 (1.18–3.18)		1.87 (1.13–3.12)	1.92 (1.13–3.26)
Previous cardiac event			p = 0.9			p = 0.5
No	3,490	91 (3%)	1.00 (ref)			1.00 (ref)
Yes	600	13 (2%)	1.05 (0.59–1.88)			0.82 (0.43–1.56)
Previous thromboembolism			p = 0.08		p = 0.3	p = 0.3
No	3,873	97 (3%)	1.00 (ref)		1.00 (ref)	1.00 (ref)
Yes	217	7 (3%)	1.97 (0.91–4.26)		1.50 (0.69–3.29)	1.51 (0.69–3.33)
Operative diagnosis (revision)			p = 0.001	p = 0.001	p = 0.003	p = 0.004
Failure: Loose/wear/osteolysis	2,154	48 (2%)	1.00 (ref)	1.00 (ref)	1.00 (ref)	1.00 (ref)
Failure: Previous surgery	801	29 (4%)	2.12 (1.34–3.37)	2.12 (1.34–3.37)	2.14 (1.33–3.42)	2.07 (1.29–3.32)
Failure: Fracture/dislocation	877	18 (2%)	1.14 (0.66–1.96)	1.14 (0.66–1.96)	1.14 (0.66–1.98)	1.13 (0.66–1.96)
Failure: Nonunion	27	2 (7%)	4.82 (1.17–19.9)	4.82 (1.17–19.9)	4.54 (1.09–18.9)	4.46 (1.06–18.7)
Failure: Infection	228	7 (3%)	2.88 (1.30–6.40)	2.88 (1.30–6.40)	2.67 (1.19–6.00)	2.73 (1.21–6.16)
ASA class			p = 0.4			p = 0.7
1	50	2 (4%)	1.00 (ref)			1.00 (ref)
2	1,869	41 (2%)	0.63 (0.15–2.62)			0.72 (0.17–3.05)
3	2,098	60 (3%)	0.91 (0.22–3.73)			0.91 (0.21–3.87)
4	61	1 (2%)	0.82 (0.07–9.06)			0.68 (0.06–7.89)

^a Multivariable model 1 was a Cox regression model that considered all significant variables from the univariate analyses with p-value <0.05 and retained only those significantly associated in the multivariable model by using a backward selection process; the c-statistic of the multivariable model was 0.60

^b Multivariable model 2 was a Cox regression model obtained by retaining all variables from the univariate analyses with p-value <0.20 and forcing all of them in the model; the c-statistic of the multivariable model was 0.63

^c Multivariable model 3 was a Cox regression model obtained by retaining all variables from the univariate analyses regardless of p-value and forcing all of them in the model; the c-statistic of the multivariable model was 0.64

ASA: American Society of Anesthesiologists; CI: confidence interval; ref: reference category.

Figure 1. Periprosthetic fracture rate after primary and revision TKA (top panels with error bars represent 95% confidence intervals), stratified further by age group (bottom panels). The number of patients at risk at baseline, 5-, 10-, 15- and 20-years is shown at the bottom of each panel.