Association of peptic ulcer disease and pulmonary disease with risk of periprosthetic fracture after primary total knee arthroplasty

Abstract

Objective

To assess the association of specific comorbidities with periprosthetic fractures after primary total knee replacement (TKR).

Methods

We used the prospectively collected data in the Mayo Clinic total joint registry from 1989–2008 on all patients who had undergone primary TKR. The outcome of interest was postoperative periprosthetic fractures during followup. The main predictors of interest were comorbidities grouped from the validated Deyo-Charlson index. Multivariable-adjusted Cox regression analyses were adjusted for sex, age, body mass index (BMI), American Society of Anesthesiology (ASA) class, operative diagnosis, and implant fixation. Hazard ratios (HRs) and 95% confidence intervals (95% CIs) were calculated.

Results

We included 17,633 primary TKRs with a mean followup of 6.3 years. The mean age was 68 years, 55% were women, and the mean BMI was 31 kg/m². There were 188 postoperative periprosthetic fractures on postoperative day 1 or later; 162 fractures (86%) occurred on postoperative day 90 or later. In multivariable analyses that simultaneously adjusted for all comorbidities and other variables (age, sex, BMI, ASA, operative diagnosis, and cement status), the following 2 conditions were significantly associated with an increased hazard of postoperative periprosthetic fractures: peptic ulcer disease (HR 1.87, 95% CI 1.28–2.75; P = 0.0014) and chronic obstructive pulmonary disease (HR 1.62, 95% CI 1.10–2.40; P = 0.02).

Conclusion

Peptic ulcer disease and chronic obstructive pulmonary disease are associated with a higher risk of periprosthetic fractures after primary TKR. This may be related to the disease or its treatments, which need further study. Identification of specific risk factors may allow for implementation of intervention strategies to reduce this risk.

Introduction

Total knee replacement (TKR) is the most common joint replacement procedure. Approximately 0.5 million procedures are performed annually in the US, and the procedure volume is projected to increase 6-fold by 2030 (1). Periprosthetic fracture is associated with higher morbidity (2) and mortality (3, 4). A better understanding of risk factors for periprosthetic fractures after TKR can provide guidance regarding potential intervention for risk modification targeting modifiable risk factors to improve outcomes.

Studies of periprosthetic fractures after TKR have identified risk factors. Risk factors assessed previously and found to be associated with prosthetic fracture include corticosteroid use (5–7), demographics such as female sex and age >70 years (8), and underlying diagnosis such as inflammatory arthritis (5–7, 9) and previous revision arthroplasty (9, 10). In 2 small studies of 26 and 14 fractures, osteoporosis was suspected to be a risk factor for periprosthetic fractures (9, 10); a cadaveric biomechanical study had similar findings (11). Most of the previous studies focused on demographics and the underlying diagnosis leading to TKR, which are unmodifiable. To our knowledge, only 2 small studies focused on osteoporosis (9, 10). None of the previously published studies have examined any other comorbidity as a (potentially modifiable) risk factor for periprosthetic fractures after TKR.

Our objective was to examine whether medical comorbidities were associated with a higher risk of periprosthetic fractures after primary TKR in a large cohort of patients who were prospectively followed in an institutional joint registry.

Significance & Innovations

• The presence of peptic ulcer disease is associated with twice the risk of periprosthetic fractures after primary total knee replacement (TKR).

• Chronic obstructive pulmonary disease leads to a 1.6-times higher risk of periprosthetic fractures after primary TKR.

• Intervention strategies targeting better preoperative comorbidity management may lead to reduction in risk of periprosthetic fractures after primary TKR.

Materials and Methods

Study cohort.

We used the prospectively collected data in the Total Joint Registry at the Mayo Clinic, Rochester, Minnesota. The Mayo Clinic Total Joint Registry, details described elsewhere (12, 13), captures demographic, clinical, and implant-related information for each patient undergoing joint replacement surgery at the Mayo Clinic. Each patient is followed prospectively clinically in the total joint registry and important postoperative outcomes, including revision, infection, fracture, and patient-reported outcomes, are captured. Several previous studies have examined periprosthetic fractures using this prospective registry (14–18), which collects data from local medical records. Standardized rigorous followup protocols are followed for the total joint registry for collection of these and other patient-reported outcomes. For patients failing to return for regular clinic followup visits, a letter is sent to patients inquiring about complications related to the index arthroplasty, including periprosthetic fractures, infections, and additional surgeries or procedures, in addition to other questions. In some cases where both approaches (clinic followup and mailed questionnaire) are not successful, trained staff personnel contact patients by telephone and inquire about postarthroplasty complications (and other outcomes). In addition, medical records from other health care facilities (including radiographs) are obtained for patients with report of complications, including fractures, infection, and any other surgery on the index joint, who do not return for followup to the Mayo Clinic. For this study, we included a cohort of all patients who had undergone primary TKR from 1989–2008. We chose this time period since key variables are available for this entire period, allowing us to adjust the analyses for important factors.

Outcome of interest.

The outcome of interest was periprosthetic fracture at postoperative day 1 or later. We did not include intraoperative fractures since their etiology is different from postoperative fractures. We chose not to include periprosthetic fractures on postoperative day 0, since these could not be definitively distinguished from intraoperative fractures due to a lack of availability of the exact time of the fracture and performance of multiple surgeries related to complications on postoperative day 0 in some patients with fractures, as well as for other complications. Therefore, our cohort selection approach was specific and conservative. The documentation of periprosthetic fracture in the Mayo Clinic Total Joint Registry came from 2 key components of patient's medical records: the operating room report and/or the orthopedic surgeon's clinical note that indicated the occurrence of fracture post-TKR in the patient's medical records. In an occasional instance, fractures that were patient reported (on mailed questionnaire or during telephone conversation) and subsequently confirmed by an orthopedic surgeon's documentation of a periprosthetic fracture in the letter of correspondence to the patient (based on review of outside medical records, operating room reports, and radiographic studies, obtained as a clinical protocol for all postarthroplasty complications) was considered consistent with an occurrence of periprosthetic fracture. Therefore, the data regarding periprosthetic fractures in the registry came from patient medical records. Patients were followed from the day of their index primary TKR to the time of occurrence of postoperative periprosthetic fracture or death, whichever occurred earlier.

Predictor variables and definitions.

The main predictors of interest in our study were medical comorbidities, grouped from the well-known comorbidity measure, the Deyo-Charlson index. Data regarding the presence of Deyo-Charlson comorbidities were obtained from the Mayo Clinic electronic health care records, where data on Deyo-Charlson comorbidities are available for this study period. It is a valid measure of comorbidity, assessed as a cumulative weighted score of several comorbidities (19, 20). The comorbidities were grouped into 10 categories as follows: cardiac disease (myocardial infarction, congestive heart failure); peripheral vascular disease; cerebrovascular disease, hemiplegia, or paraplegia; moderate to severe renal disease; peptic ulcer disease; chronic obstructive pulmonary disease (COPD); diabetes mellitus (with or without organ damage); connective tissue disease; cancer (leukemia, lymphoma, any other tumor, metastatic solid tumor); and other (dementia, liver disease, acquired immunodeficiency syndrome [AIDS]/human immunodeficiency virus [HIV]).

Statistical analyses.

We used univariate Cox proportional hazards regression analyses to examine the association of each of the Deyo-Charlson comorbidities and periprosthetic fracture. We performed multivariable-adjusted Cox regression analyses that simultaneously adjusted for the following important variables (known or suspected to be associated with periprosthetic fracture risk). These included sex, age (≤60, 61–70, 71–80, and >80 years), body mass index (BMI) as per the World Health Organization classification (<25, 25–29.9, 30–39.9, and ≥40 kg/m²) (21), American Society of Anesthesiology (ASA) physical status score class (1, 2, 3, and 4) (22, 23), operative diagnosis (osteoarthritis, rheumatoid arthritis, other), and implant fixation (cemented/hybrid, uncemented). Hazard ratios (HRs) and 95% confidence interval (95% CIs) are shown.

Results

Demographic and clinical characteristics of the patients are shown in Table 1. A total of 17,633 primary TKRs were included in the study, with a mean followup of 6.3 years. The mean age was 68 years, the mean BMI was 31 kg/m², 27% had bilateral procedures, and 55% were women. There were 188 postoperative periprosthetic fractures on postoperative day 1 or later, of which 162 fractures (86%) occurred 90 days or later after the index TKR.

Table 1. Demographic features of the study cohort*

	Primary TKR (n = 17,633)
* TKR = total knee replacement; BMI = body mass index; ASA = American Society of Anesthesiologists; AIDS = acquired immunodeficiency syndrome; HIV = human immunodeficiency virus. † Other diseases include genu varum, genu valgum, hemophilia, Paget's disease, failed previous disease, including arthrodesis, failed previous osteotomy, failed previous patellectomy, Charcot's arthropathy, chondromalacia, pigmented villonodular synovitis, etc.
Followup, mean ± SD years	6.3 ± 4.7
Male/female, no. (%)	7,852 (45)/9,781 (55)
Unilateral/bilateral, no. (%)	12,914 (73)/4,719 (27)
Age at surgery, mean ± SD years	68.4 ± 10.0
Age category, no. (%)
≤60 years	3,352 (19)
61–70 years	6,206 (35.2)
71–80 years	6,493 (36.8)
>80 years	1,582 (9)
BMI, mean ± SD kg/m^2	31.2 ± 6.15
BMI category, no. (%)
Missing	67 (0.4)
Normal (<25.0 kg/m^2)	2,362 (13.4)
Overweight (25–29.9 kg/m^2)	5,961 (33.9)
Obese (30–39.9 kg/m^2)	7,710 (43.9)
Morbidly obese (≥40.0 kg/m^2)	1,533 (8.7)
ASA score, no. (%)
Missing	67 (0.4)
1	291 (1.7)
2	9,614 (54.7)
3	7,544 (42.9)
4	117 (0.7)
Deyo-Charlson Index group, no. (%)
Heart disease	1,794 (10.2)
Peripheral vascular disease	1,106 (6.3)
Cerebrovascular disease, hemiplegia, or paraplegia	1,678 (9.5)
Moderate to severe renal disease	1,320 (7.5)
Peptic ulcer disease	1,721 (9.8)
Chronic obstructive pulmonary disease	2,228 (12.6)
Diabetes mellitus	2,217 (12.6)
Connective tissue disease	1,596 (9.1)
Cancer	2,925 (16.6)
Other (dementia, liver disease, AIDS/HIV)	1,173 (6.7)
Operative diagnosis (primary), no. (%)
Osteoarthritis	16,372 (92.8)
Rheumatoid/inflammatory arthritis	657 (3.7)
Other†	604 (3.4)
Implant fixation, no. (%)
Uncemented	1,555 (8.8)
Cemented or hybrid	16,078 (91.2)

In univariate analyses, peptic ulcer disease, COPD, and other diagnoses (dementia, liver disease, AIDS/HIV) were each associated with a higher risk of periprosthetic fractures (Table 2).

Table 2. Univariate and multivariable-adjusted hazards of postoperative periprosthetic fracture following primary TKR*

	Total (n = 17,633), no.	Periprosthetic fractures (n = 188), no. (%)	Univariate HR (95% CI)	Multivariable HR (95% CI)†	Multivariable HR (95% CI)‡
* TKR = total knee replacement; HR = hazard ratio; 95% CI = 95% confidence interval; MI = myocardial infarction; CHF = congestive heart failure; COPD = chronic obstructive pulmonary disease; AIDS = acquired immunodeficiency syndrome; HIV = human immunodeficiency virus. † Multivariable model simultaneously adjusted for all variables, including age, sex, American Society of Anesthesiologists class, body mass index, operative diagnosis, implant fixation (cement status), and each Deyo-Charlson comorbidity category; 17,499 of the 17,633 TKRs were available for the multivariable analyses. ‡ Multivariable model adjusted only for comorbidities significant in univariate analyses with a P value <0.05.
Heart disease (MI, CHF), P			0.419	0.96
No	15,839	168 (1)	1.00 (ref.)	1.00 (ref.)
Yes	1,794	20 (1)	1.21 (0.76–1.93)	1.01 (0.61–1.67)
Peripheral vascular disease, P			0.16	0.48
No	16,527	174 (1)	1.00 (ref.)	1.00 (ref.)
Yes	1,106	14 (1)	1.48 (0.86–2.55)	1.23 (0.69–2.19)
Cerebrovascular disease (including hemiplegia, paraplegia), P			0.693	0.73
No	15,955	171 (1)	1.00 (ref.)	1.00 (ref.)
Yes	1,678	17 (1)	1.11 (0.67–1.82)	0.91 (0.53–1.55)
Moderate to severe renal disease, P			0.037	0.13	0.11
No	16,313	170 (1)	1.00 (ref.)	1.00 (ref.)	1.00 (ref.)
Yes	1,320	18 (1)	1.68 (1.03–2.73)	1.49 (0.89–2.49)	1.51 (0.92–2.49)
Peptic ulcer disease, P			< 0.001	0.0014	0.0011
No	15,912	154 (1)	1.00 (ref.)	1.00 (ref.)	1.00 (ref.)
Yes	1,721	34 (2)	2.08 (1.43–3.02)	1.87 (1.28–2.75)	1.88 (1.29–2.75)
COPD, P			0.002	0.02	0.01
No	15,405	154 (1)	1.00 (ref.)	1.00 (ref.)	1.00 (ref.)
Yes	2,228	34 (2)	1.78 (1.22–2.57)	1.62 (1.10–2.40)	1.62 (1.11–2.39)
Diabetes mellitus (with or without organ damage), P			0.161	0.24
No	15,416	163 (1)	1.00 (ref.)	1.00 (ref.)
Yes	2,217	25 (1)	1.35 (0.89–2.06)	1.30 (0.84–2.03)
Other (dementia, liver disease, AIDS/HIV), P			0.042	0.07	0.13
No	16,460	171 (1)	1.00 (ref.)	1.00 (ref.)	1.00 (ref.)
Yes	1,173	17 (1)	1.68 (1.02–2.77)	1.64 (0.96–2.81)	1.48 (0.89–2.46)
Connective tissue disease, P			0.056	0.55
No	16,037	161 (1)	1.00 (ref.)	1.00 (ref.)
Yes	1,596	27 (2)	1.49 (0.99–2.24)	1.17 (0.69–1.99)
Cancer, P			0.785	0.17
No	14,708	161 (1)	1.00 (ref.)	1.00 (ref.)
Yes	2,925	27 (1)	0.94 (0.63–1.42)	0.73 (0.47–1.15)

In multivariable analyses that simultaneously adjusted for all comorbidities and other variables (age, sex, BMI, ASA, operative diagnosis, cement status), only peptic ulcer disease and COPD were significantly independently associated with a higher risk of periprosthetic fracture, with respective HRs of 1.87 (95% CI 1.28–2.75, P = 0.0014) and 1.62 (95% CI 1.10–2.40, P = 0.02) (Table 2). Other diagnoses (dementia, liver disease, AIDS/HIV) were not significantly associated with risk of periprosthetic fractures after multivariable adjustment.

Sensitivity analyses were performed by entering only significant comorbidities from the univariate analyses into multivariable-adjusted analyses, which did not alter the significance of associations and only changed their magnitude minimally (Table 2). Another sensitivity analysis adjusting the main multivariable model additionally for bilaterality did not have any meaningful effect on the associations for peptic ulcer disease (odds ratio [OR] 1.87, 95% CI 1.27–2.75) and COPD (OR 1.63, 95% CI 1.10–2.40).

Discussion

In this study, we found that the presence of 2 diseases was associated with an increased risk of postoperative periprosthetic fractures. Peptic ulcer disease increased the risk of periprosthetic fracture by 87% and COPD increased the risk by 62%. This increased risk is significant both statistically and clinically. One key finding to be kept in mind is the robustness of these estimates. For both conditions, the estimates were attenuated only slightly from the univariate models to multivariable-adjusted regression models (adjusted for 16 variables simultaneously). Both conditions were also common in our patient cohort undergoing primary TKR (10% and 13% of patients, respectively). These findings are important and novel and deserve further discussion.

The observation of association of peptic ulcer disease and periprosthetic fractures leads to additional interesting hypotheses that need to be tested in future studies. Is this association due to peptic ulcer disease? Or is it treatment related (proton-pump inhibitors versus other therapies)? Emerging evidence in cohorts of postmenopausal women suggests a consistently increased risk of fragility fractures in patients using proton-pump inhibitors (24–26), while it is unclear whether H₂ receptor blockers are associated with a higher or lower fracture risk (24, 26). It remains to be seen whether these observations made for postmenopausal fragility fractures are true for patients undergoing TKR. These hypotheses need to be tested in TKR patients with periprosthetic fractures in future studies.

The association of COPD with periprosthetic fracture risk is not surprising, considering that corticosteroids (inhaled, oral, and parenteral) are used not uncommonly in patients with COPD. Some evidence exists in the nonarthroplasty literature regarding similar risks. COPD duration and severity and the use of inhaled corticosteroids were associated with an increased risk of vertebral fractures (27–29). The lack of availability of medication data in the total joint registry limited us from performing additional analyses to investigate whether the use of corticosteroids was the underlying reason for the association of COPD with periprosthetic fractures. Studies to examine these hypotheses in a different data set that provides data on medication use in TKR cohorts are underway to determine the exact nature of this association.

How do these findings help surgeons and patients? Our study examined several diseases as potential predictors of periprosthetic fractures and found that only 2 comorbidities were associated independently with the risk. The knowledge gained from our study can be incorporated in patient–surgeon discussions regarding this increased risk in patients with peptic ulcer disease or COPD. These findings should also stimulate research into why these associations exist and further research into mechanisms of these associations.

These findings must be interpreted by considering study limitations. Misclassification bias for comorbidity is possible, since we used electronically captured diagnoses. This may have biased some estimates toward the null, meaning that we may have missed some significant associations. Therefore, our estimates are likely conservative. We may have missed periprosthetic fractures in some patients due to loss to followup despite intensive efforts by dedicated total joint registry staff, and thus the cumulative incidence estimates may be conservative. However, as described earlier, the dedicated registry staff makes every attempt to follow each patient prospectively for these outcomes. Another limitation is the inability to control for osteoporosis, an important confounder of the association of other comorbidities and fracture risk. The study was not designed to examine operation technique–related factors and postoperative physical activity, covariates that may impact the observed associations. Studies are needed to address if after adjustment for these and other factors, the associations noted in our study hold true. Both loss to followup (underreporting of fractures in some patients not reachable through clinic, mail, and phone contact) and misclassification of comorbidity and fractures due to capture from medical records (rather than a re-review and validation) likely biased our estimates toward the null, making these estimates conservative, where the real associations may be even more impressive. However, there is a rigorous protocol to follow all postarthroplasty patients for complications, and the diagnosis of periprosthetic fractures was based on medical records, including the operating room report, an orthopedic surgeon's clinical note, and/or review of outside records or radiographic films in all cases, and Deyo-Charlson comorbidities have been systematically captured in Mayo Clinic databases for this study period, a key reason for choosing this time period for study. These procedures minimize these biases. Selection bias is unlikely to have impacted these results, since the sample consisted of every patient who underwent TKR and provided permission for use of their medical records (>98% of all patients seen at Mayo Clinic provide permission for use of data for research).

Our study has several strengths, including the use of prospectively collected data from the total joint registry and a large sample size allowing for meaningful analyses. Another strength is the use of multivariable-adjusted analyses that control for multiple important demographic, clinical, and implant factors. Our estimates were robust, and showed minor attenuation when adjusted for several variables in multivariable analyses.

In summary, the presence of peptic ulcer disease and COPD was associated with a higher risk of postoperative periprosthetic fractures after primary TKR. The increase in risk was both statistically and clinically significant, indicated by almost doubling the risk. Surgeons should consider discussing this risk with patients prior to the surgery, considering that both conditions are common in patients undergoing TKR. Future studies should examine whether these associations are related to the disease (peptic ulcer disease or COPD) or the treatments used for these conditions (proton-pump inhibitors, H₂ blockers, corticosteroids, etc.).

AUTHOR CONTRIBUTIONS

All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Singh had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Singh, Lewallen.

Acquisition of data. Singh, Lewallen.

Analysis and interpretation of data. Singh, Lewallen.

Acknowledgements

We thank Youlonda Lochler for help with data acquisition and Matt Jensen and Scott Harmsen in helping with data analysis.