Peripheral arterial disease and risk of hip fracture: A systematic review and meta-analysis of cohort studies.

Background: Previous studies have suggested an increased risk of hip fracture among patients with peripheral arterial disease (PAD), however, the results have been inconsistent. This meta-analysis was conducted with the aim to summarize all available evidence to better characterize the risk of incident hip fracture among these patients. Materials and
Methods: A comprehensive literature review was conducted using the MEDLINE and EMBASE databases through October 2017 to identify all cohort and case-control studies that compared the risk of subsequent hip fracture between patients with PAD and individuals without PAD. Effect estimates of the included studies were extracted and combined using the random-effect, generic inverse variance method of DerSimonian and Laird.
Results: The systematic review process yielded six eligible cohort studies comprising 15,895 patients with PAD. There was a significant association between incident hip fracture and PAD with the pooled relative risk (RR) of 1.64 (95% CI, 1.17-2.29; I2, 80%), comparing patients with PAD and individuals without PAD. Subgroup analysis by study design revealed significant results for both prospective studies (pooled RR 1.60; 95% CI, 1.12-2.28; I2, 0%) and retrospective studies (pooled RR 1.72; 95% CI, 1.07-2.77; I2, 92%). The funnel plot is relatively asymmetric suggesting publication bias.
Conclusion: This study found a significant association between PAD and hip fracture with the pooled RR of 1.64 (95% CI, 1.17-2.29) on comparing patients with PAD and individuals without PAD. Major limitations include high between-study heterogeneity, possibility of publication bias, and lack of data on the characteristics and type of hip fracture which may limit the clinical significance of the observations.

Introduction

Peripheral arterial disease (PAD) is an atherosclerotic disease that leads to the obstruction of blood flow in the peripheral arterial system. It occurs most commonly in the arteries of the lower extremities.[1] The reported prevalence of PAD is approximately 1% among individuals aged between 55 and 60 years and over 5% among individual aged between 80 and 85 years.[23] It is one of the leading causes of limited ambulation, decreased muscle strength, and increased tendency to fall among the elderly.[4] Previous studies have demonstrated that patients with PAD have lower bone mineral density, particularly in the femur, compared with the general population.[56] Taking the lower bone mineral density into consideration, it is possible that patients with PAD may be at a higher risk of hip fracture. Nonetheless, epidemiologic studies of this association have shown inconsistent results.[78910111213] This systematic review and meta-analysis was conducted to summarize all available evidence to better characterize the risk of incident hip fracture among these patients.

Materials and Methods

Information sources and search strategy

A systematic literature search was conducted using EMBASE and MEDLINE databases from inception to October 2017 to identify all cohort studies that investigated the risk of incident hip fracture among patients with PAD compared with individuals without PAD. The systematic literature review was independently conducted by the first two investigators using the search strategy that included the terms for “peripheral arterial disease” and “hip fracture,” as described in online Supplementary Data 1. A manual search for additional potentially relevant studies using references of the included articles and selected review articles was also performed. No language limitation was applied. This study was conducted in accordance with the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analysis) guidelines, which are provided as online Supplementary Data 2.

Click here for additional data file.

PRISMA 2009 Checklist

Click here for additional data file.

Selection criteria

Eligible studies were cohort studies or case-control studies investigating the risk of incident hip fracture among patients with PAD. For cohort studies, cases included patients with PAD whereas controls included individuals without PAD. Cohort studies were either prospective (the event of interest does not happen yet – investigators follow the participants with and without PAD prospectively until the event of interest occurs) or retrospective (both PAD and the event of interest already occurred before investigators conducted the study – the study was conducted based on existing medical records). For case-control studies, cases included patients with hip fracture whereas controls were individuals without hip fracture. Eligible studies mentioned the effect estimates [relative risks (RR), hazard ratios (HR), or odds ratio (OR)] with 95% confidence intervals (CI) for calculating pooled-effect estimates. If more than one article using the same database/cohort were identified, only one study with the most comprehensive data was included in the meta-analysis.

Retrieved articles were independently screened and reviewed for their eligibility by the same two investigators. Difference in the determination of study's eligibility was resolved by conference with all investigators. Newcastle–Ottawa quality assessment scale was used for appraisal of the quality of the included studies. This scale evaluates the quality of the included studies in three areas including the recruitment of participants, the comparability between the groups, and the ascertainment of the outcome of interest.[14]

Data abstraction

A standard data collection form was used to extract the following data from each study – title of the study, year of publication, name of the first author, year when the study was conducted, country of origin, number of cases and controls, baseline demographic data of cases and controls, methods used to identify and verify diagnosis of PAD and hip fratcure, as well as adjusted-effect estimates and their associated 95% CI. This process was independently performed by the same two investigators to ensure the accuracy of data extraction and minimize errors. Any discrepancies were resolved by referring back to the original study.

Statistical analysis

Review Manager 5.3 software from the Cochrane Collaboration (London, United Kingdom) was used for all data analysis. Adjusted point estimates from each study were combined using the generic inverse variance method described by DerSimonian and Laird, which assigned the weight of each study inversely to its variance.[15] Given that the included studies were conducted in different populations and settings and the primary aim of the meta-analysis was to estimate the risk of PAD that could be generalized to various populations, a random-effect model was used rather than a fixed-effect model. Cochran's Q-test and I2 statistic were used to determine the between-study heterogeneity. A value of I2 of 0–25% represented insignificant heterogeneity, 26–50% represented low heterogeneity, 51–75% represented moderate heterogeneity, and more than 75% represented high heterogeneity.[16] Funnel plot was used to assess publication bias.

Results

A total of 8,464 potentially eligible articles were identified using our search strategy (3,876 articles from Medline and 4,588 articles from EMBASE). After exclusion of 3,633 duplicate articles, 4,831 articles underwent title and abstract review. At this stage, 4,125 articles were excluded as they were neither cohort studies nor case-control studies (they were case reports, letter to editor, review articles, animal studies, or randomized controlled trials etc.), leaving 706 articles for full-text review. The majority of these studies were excluded after full-length review for the following reasons; 120 articles were excluded because they were not cohort studies or case-control studies (the reason for exclusion was the same as the 4,125 articles excluded after title and abstract review but it was not clear from just the title and abstract review for these 120 articles; full-text review was needed before it was certain that they were neither cohort studies nor case-control studies); 86 articles were case-control studies that did not include PAD as one of the exposures of interest; 384 articles were cohort studies that were descriptive in nature (i.e., only one cohort of patients with PAD was in the study without controls); 109 articles were cohort studies that did not report the outcome of interest (i.e., hip fracture). A total of 7 studies met the eligibility criteria.[78910111213] However, two studies[1213] utilized the same database (the database of the national health insurance of Taiwan). To avoid duplication of patients, only the study by Liu et al.[12] was included in this meta-analysis as their analysis was more comprehensive than the study by Lai et al.[13] Therefore, six cohort studies (three prospective cohort studies and three retrospective cohort studies) involving 15,895 patients with PAD and 233,835 comparators without PAD met the eligibility criteria and were included in the meta-analysis. The literature review and study selection process are shown in Figure 1. The characteristics and Newcastle–Ottawa assessment scales of the included studies are presented in Table 1. It should be mentioned that all studies eligible to be included in this meta-analysis were cohort studies. No case-control study fulfilled the eligibility criteria.

Figure 1

Literature review and selection

Table 1

Characteristics of the included studies

	Sennerby et al.[7]	Collins et al.[8]	Hyde et al.[9]	Reyes et al.[10]	Palaez et al.[11]	Liu et al.[12]
Country of origin	Sweden	United States	Australia	Spain	Spain	Taiwan
Study design	Retrospective cohort	Prospective cohort	Prospective cohort	Retrospective cohort	Prospective cohort	Retrospective cohort
Year of publication	2009	2009	2012	2014	2015	2017
Cases	All patients in the Swedish twin Registry who were diagnosed with PAD. This registry included 15,968 pairs of twins born in Sweden between 1914 and 1944.	All subjects in the MrOS study who were diagnosed with PAD. This study recruited patients from 6 major metropolitan areas across the United States from March 2000 to April 2002. All patients underwent ABI testing to look for evidence of PAD.	All patients in the screening for vascular health study of the University of Western Australia who were diagnosed with PAD. This study recruited patients from April 1996 to June 1998. All patients underwent ABI testing to look for evidence of PAD.	All patients who were diagnosed with PAD before January 1, 2007. Cases were identified from the electronic database of the general practitioner of Catalonia, Spain which comprised over 2 million people.	Cases with PAD were recruited from residences of a nursing home in Valladolid, Spain. All residents underwent ABI testing to look for evidence of PAD.	All patients who were first diagnosed with PAD between January 1, 2000 and December 31, 2004. Cases were identified from the electronic database of the Taiwan’s National Health Insurance Program.
Diagnosis of PAD	Presence of diagnostic codes for PAD in the database plus telephone interview for confirmation.	ABI of <0.9.	ABI of <0.9.	Presence of diagnostic codes for PAD in the database.	ABI of<0.9.	Presence of diagnostic codes for PAD in the database.
Controls	The remaining patients without PAD and other cardiovascular diseases in the database served as controls.	The remaining patients without PAD in the cohort served as controls.	The remaining patients without PAD in the cohort served as controls.	The remaining patients without PAD in the database served as controls.	The remaining residences without PAD served as controls.	Sex and age-matched individuals without PAD who were randomly selected from the same database served as controls
Diagnosis of hip fracture	Presence of diagnostic codes for hip fracture in the database plus telephone interview for confirmation.	Periodic review of the medical records plus mail/telephone interview.	Periodic review of the medical records.	Presence of diagnostic codes for hip fracture in the database.	Periodic review of the medical records of the nursing home.	Presence of diagnostic codes for hip fracture in the database.
Follow-up	Until occurrence of study endpoint, death or December 31, 2005.	Until occurrence of study endpoint, death or August 1, 2007.	Until December 31, 2008.	Until December 31, 2009.	20 months after the survey date.	Until occurrence of the study endpoint, death, emigration out of system, or December 31, 2008.
Number of cases	1,061	358	506	7,305	18	6,647
Number of controls	19,089	5,423	3,815	178,866	54	26,588
Mean age for cases/controls (years)	69.0/62.0	77.0/73.3	N.A.	N.A.	88.0/82.5	N.A.
Average duration of follow-up for cases/controls (years)	N.A.	5.6 (overall)	11.1 (overall)	2.6/2.8	1.7/1.7	9.9/10.6
Confounder adjusted for	Sex and comorbidities	Age, ethnicity, and study site	Age, BMI, smoking, alcohol use, and vascular comorbidities	Age, BMI, smoking, alcohol use, and comorbidities	None	Age, sex, comorbidities, medication, and socioeconomic status
Quality assessment (Newcastle-Ottawa scale)	Selection: 4 stars Comparability: 2 stars Outcome: 3 stars	Selection: 4 stars Comparability: 2 stars Outcome: 3 stars	Selection: 3 stars Comparability: 2 stars Outcome: 3 stars	Selection: 4 stars Comparability: 1 star Outcome: 3 stars	Selection: 3 stars Comparability: 1 star Outcome: 2 stars	Selection: 4 stars Comparability: 2 stars Outcome: 3 stars

ABI: Ankle-brachial index, BMI: Body mass index, N.A: Not available, PAD: Peripheral arterial disease

There was a significant association between incident hip fracture and PAD with the pooled RR of 1.64 (95% CI, 1.17–2.29), comparing patients with PAD with individuals without PAD. The statistical heterogeneity was high with an I2 of 80% (Q statistic, 25.02; P - value, 0.0001). Subgroup analysis by study design showed a significant association between incident hip fracture and PAD for both prospective studies (pooled RR, 1.60; 95% CI, 1.12–2.28; I2, 0%; Q statistic, 1.41; P - value, 0.49) and retrospective studies (pooled RR, 1.72; 95% CI, 1.07–2.77; I2, 92%; Q statistic, 23.57; P - value, <0.0001). The forest plot of the meta-analysis is shown in Figure 2.

Figure 2

Forest plot of this meta-analysis

Evaluation for publication bias

The funnel plot is relatively asymmetric and suggesting publication bias [Figure 3].

Figure 3

Funnel plot of this meta-analysis

Sensitivity analysis

Since the study by Palaez et al.[11] was the only study with a negative result, we conducted a sensitivity analysis by excluding this study from the our analysis to investigate if the pooled result and/or heterogeneity would be significantly altered. The result of this sensitivity analysis was not significantly different from the complete analysis with the pooled RR of 1.69 (95% CI, 1.21–2.37; I2, 83%; Q statistic, 23.74; P value, <0.0001).

Discussion

This is the first systematic review and meta-analysis that summarizes all available studies on the risk of subsequent hip fracture among patients with PAD. The results of individual studies ranged from insignificantly lower risk to significantly higher risk. Taking all results together, we could demonstrate a significant association between PAD and incident hip fracture with the relative risk of approximately 1.6-fold, comparing patients with PAD versus individuals without PAD. There are few possible explanations for this increased risk.

First, decreased blood flow to the lower extremities, as a result of PAD, may affect bone homeostasis. Poor blood supply could result in accelerated demineralization and osteoporosis which are the prime risk factors for hip fracture. In fact, several studies have demonstrated that bone mineral density of the lower extremities of patients with compromised arterial flow is lower than sex and age-matched controls.[681718] A study conducted in the United States also found a significantly higher prevalence of PAD among individuals with osteopenia and osteoporosis (4.5% and 10.9% of men with osteopenia and osteoporosis, respectively, compared with 3.0% of men with normal BMD; 4.8% and 11.8% of women with osteopenia and osteoporosis, respectively, compared with 3.3% of women with normal BMD).[19]

Although the lower bone mineral density among patients with PAD is probably the main mechanism leading to the increased risk of hip fracture, other factors may also play a role. The increased risk of hip fracture could be partly attributable to the increased tendency to fall among patients with PAD as their impaired lower extremity circulation is associated with reduced muscle strength, increased prevalence of peripheral neuropathy, and functional impairment.[4202122] The limited ambulation because of PAD may also lead to sedentary lifestyle with less exposure to sunlight, and thus, lower level of vitamin D, which could be another contributing factor for the decreased bone mineral density.[23]

Although the literature review was comprehensive and the quality of the included cohort studies was high, we acknowledge that this systematic review and meta-analysis has some limitations. Therefore, the results should be interpreted with caution.

First, the statistical heterogeneity of this meta-analysis was high. We suspected that the difference in study designs was partly responsible as subgroup analysis showed an I2 of 0% for prospective studies. Second, funnel plot was relatively asymmetric and may suggest that publication bias was present. Third, some included studies were medical registry-based studies. Those studies relied on diagnostic codes in the registry to identify and verify diagnosis of PAD and hip fracture, which may raise a concern over incomplete coding and misclassification as well as lack of data on the characteristics and type of hip fracture which may limit the clinical significance of the observations.

Conclusion

In summary, this study demonstrated a significant association between PAD and hip fracture with the RR of approximately 1.6-fold, comparing patients with PAD versus individuals without PAD. Major limitations include high between-study heterogeneity and possibility of publication bias. How this significant association should be addressed in clinical practice requires further investigation.

Financial support and sponsorship

Nil.

Conflicts of interest

There are no conflicts of interest.