593190
research-article2015
IJLXXX10.1177/1534734615593190The International Journal of Lower Extremity WoundsZhou et al
Clinical and Translational Research
HbA1c and Lower Extremity Amputation Risk in Patients With Diabetes: A MetaAnalysis
The International Journal of Lower Extremity Wounds 2015, Vol. 14(2) 168–177 © The Author(s) 2015 Reprints and permissions: sagepub.com/journalsPermissions.nav DOI: 10.1177/1534734615593190 ijl.sagepub.com
Zhen-Yu Zhou, MD1,2, Ya-Ke Liu, MD2, Hong-Lin Chen, MD3, Hui-Lin Yang, MD1, and Fan Liu, MD2
Abstract In this meta-analysis, we aimed to assess glycosylated hemoglobin (HbA1c) level and lower extremity amputation (LEA) risk in patients with diabetes. Systematic computerized searches of the PubMed and Web of Knowledge were performed. We compared HbA1c level between groups with LEA and without LEA by meta-analysis; we also examined the dose–response relationship between HbA1c level and LEA risk. Sixteen studies were included in the meta-analysis. Eleven studies with 43 566 patients compared HbA1c between groups with and without LEA. The mean HbA1c (%) ranged from 8.3 to 12.5 in the group with LEA and from 7.4 to 11.3 in the group without LEA. The pooled weighted mean difference was 1.110 (95% confidence interval = 0.510-1.709; Z = 3.63, P = .008). The funnel plot was symmetrical, and Begg’s test (z = 0.00, P = 1.000) and Egger’s test (t = −0.02, P = .984) suggested no significant publication bias. Six studies with 109 933 patients included in the dose–response meta-analysis. The LEA incidence ranged from 0.3% to 14.6% between different HbA1c levels. Dose–response meta-analysis showed statistically significant association between HbA1c and LEA risk (χ2 = 65.51, P = .000). In linear model, the odds ratio for LEA incidence was 1.229 (95% confidence interval = 1.169-1.292) for every 1% HbA1c increase. In the spline model, the odds ratio of LEA risk increased with HbA1c levels, especially when HbA1c ranged from 5% to 9%. Our meta-analysis indicates that high level of HbA1c is an important risk factor for LEA in patients with diabetes. This evidence supports the strategy for lowering glucose levels to reduce amputation in patients with diabetes. Keywords diabetes mellitus, lower extremity amputation, risk factor, glycosylated hemoglobin, meta-analysis
Foot ulcers are one of the most serious complications of diabetes mellitus, with the prevalence of 4% to 10% among persons diagnosed as having diabetes mellitus.1 Foot ulcers frequently become infected, cause morbidity, engender financial costs, and lead to lower extremity amputation (LEA).1 Studies showed the incidence of all forms of LEA ranges from 46.1 to 9600 per 100 000 in diabetes patients, and major amputation ranges from 5.6 to 600 per 100 000.2 Most important, LEA is companied with a significantly elevated mortality at follow-up, ranging from 13% to 40% at 1 year, 35% to 65% at 3 years, and 39% to 80% at 5 years—worse than for most malignancies.1 LEA also severely reduces quality of life and has major economic consequences. The mean cost per inpatient stay was $17 103 according to the US Agency for Healthcare Research and Quality Healthcare Cost and Utilization Project.3 A study in Germany showed costs up to 24 weeks after LEA were 36 686 Euros and 115 676 Euros for 3 years after LEA.4 Many studies have investigated the risk factors for LEA in diabetes patients, and the results showed LEAs
were independently associated with higher age, male gender, longer diabetes duration, diabetic polyneuropathy, and poor glycemic control.5,6 There is a significant increase in apoptosis in diabetic wounds with poorly controlled blood sugar, which contributes to delayed wound healing.7 Studies also showed patients with poor glycemic control had higher percentage of minor and major LEA.8 Glycosylated hemoglobin (HbA1c) was an important serum marker for monitoring the degree of control of glucose in diabetic patients. In diabetes mellitus, higher amounts of HbA1c, indicating poorer 1
First Affiliated Hospital of Soochow University, Suzhou, Jiangsu, PR China Affiliated Hospital of Nantong University, Nantong, Jiangsu, PR China 3 Nantong University, Nantong, Jiangsu, PR China 2
Corresponding Author: Fan Liu, Department of Orthopaedics, Affiliated Hospital of Nantong University, Nantong 226001, Jiangsu Province, People’s Republic of China. Email: [email protected]
Downloaded from ijl.sagepub.com at University of New England on July 22, 2015
169
Zhou et al control of blood glucose levels, have been associated with diabetes complications, such as cardiovascular disease, microalbuminuria, and the progression of renal status.9 However, the relationship between HbA1c and LEA in patients with diabetes is still unknown. The aim of this meta-analysis was to determine whether HbA1c is associated with LEA in patients with diabetes, and to establish the dose–response relationship of this impact.
Methods Databases and Literature Search We searched MEDLINE and Web of Science up to December 12, 2014. The search terms included of “Diabetes,” “Amputation,” “Glycosylated Hemoglobin,” and “HbA1c.” The search detail in MEDLINE was ((Diabete*[Title/ Abstract]) AND Amputation [Title/Abstract]) AND (Glycosylated hemoglobin [Tw] OR HbA1c [Tw]). In Web of Science, the search detail used was the following: (Ts= Diabete*) and (Ts= Amputation) and (Ts= Glycosylated hemoglobin or Ts= HbA1c). We supplemented our searches by manually reviewing the references of all relevant studies.
Statistical Analysis We analyzed the data by 2 stages. First, we assessed HbA1c level difference between groups with and without LEA. Fixed effect model or random effect model was chose according to heterogeneity. Weighted mean difference (WMD) was used for assessing difference, and overall effects were determined using the Z test. Visual inspection of a funnel plot, the Egger test, and Begg test were performed to assess publication bias. Sensitivity analyses were performed by excluding the study with largest sample. Second, we examined the dose–response relationship between different HbA1c categories and LEA risk. The dose–response meta-analysis was carried out using the method proposed by Greenland and Longnecker11 and Orsini et al.12,13 We examined linear and spline dose– response relationships. The spline relationship was examined by HbA1c category using restricted cubic splines with 3 knots. All statistical analyses were performed with Stata software, version 13.0 (Stata Corp, College Station, TX). Two-sided P < .050 was considered statistically significant.
Results
Eligibility Criteria The following inclusion criteria had to be fulfilled: (a) observational studies (cross-sectional, case–control, or cohort study) that permitted assessment of a causal association between HbA1c and LEA risk in patients with diabetes; (b) LEA included not only major amputation but also minor amputation (major amputation was defined as any resection proximal to midtarsal level, and minor amputation was defined as midtarsal or distal amputations); (c) compared HbA1c levels between groups with and without LEA; or (d) reported number of cases (LEA) and controls (no LEA) for each HbA1c category, which included at least 3 HbA1c categorized levels.
Data Extraction and Critical Appraisal The following data were abstracted onto standardized forms: first author, publication year, country, study design, enrollment year, age and gender of participants, diabetes type, diabetes duration, follow-up years, the HbA1c levels in groups with and without LEA, and number of patients in case or control group in each HbA1c category. The quality of case–control and cohort study was assessed according to the Newcastle–Ottawa Scale (NOS).10 The NOS contains 8 items, categorized into 3 dimensions: Selection (4), Comparability (1), and Exposure (3). The NOS ranges between 0 and 9. Data extraction and article quality assessment were carried out independently by 2 reviewers. Disagreements were resolved by discussion.
Characteristics of Included Studies We identified 16 studies14-29 that met our inclusion criteria for meta-analysis. The detailed steps of our literature search are shown in Figure 1. The studies included 10 prospective cohorts, 4 retrospective cohorts, and the remaining 2 studies did not report details of the the study design. Of these studies, 6 were conducted in the United States, 2 in the United Kingdom, 2 in China; Turkey, Australia, Finland, Germany, Spain, and India contributed 1 study each. The enrollment year ranged from 1982 to 2011. All studies investigated the relationship in adult, and did not include children. The mean age ranged from 26 to 65 years. The male patients accounted for 36.9% to 62.9% of the proportion. Type 2 diabetes was assessed in 3 studies, type 1 in 1 study, type 1 and type 2 in 3 studies, and the remaining 9 studies did not report the diabetes type. The quality rating of the included studies ranged from 6 to 8 on a scale of 9. Tables 1 and 2 show the characteristics of the 16 identified studies.
Meta-Analysis for HbA1c Difference Between Groups With and Without LEA Eleven studies14,16-23,28,29 with 43 566 patients included in the meta-analysis compared HbA1c level differences between the 2 groups. The HbA1c level (%) ranged from 8.3 to 12.5 in the group with LEA, and from 7.4 to 11.3 in the group without LEA. Significant heterogeneity was found between the studies (I2 = 97.4%, P = .000), and the
Downloaded from ijl.sagepub.com at University of New England on July 22, 2015
170
The International Journal of Lower Extremity Wounds 14(2)
Records identified through Pubmed searching (n=83) Records identified through Web of Science searching (n=67)
Additional records identified through meeting abstract (n =3)
Records duplicates removed (n = 61)
Records screened (n =92)
Records excluded by title and abstract review Animal experiments and biological research (n=4) Reviews, case reports and case series (n=14) Re-amputation risk (n=2) No available data (n=35)
Full-text articles assessed (n=37)
Full-text articles excluded No available data (n=21)
Studies included in quantitative synthesis (meta-analysis) (n=16) Figure 1. Flow diagram showing selection of studies.
random effect model was used. The pooled WMD was 1.110 (95% confidence interval [CI] = 0.510-1.709; Z = 3.63, P = .008; Figure 2). The funnel plot was symmetrical (Figure 3), and Begg’s test (z = 0.00, P = 1.000) and Egger’s test (t = −0.02, P = .984) suggested there was no significant publication bias. Sensitivity analyses by excluding the study with largest sample showed the pooled WMD was 1.124 (95% CI = 0.268-1.981; Z = 2.57, P = .010).
In the linear model, the LEA incidence OR was 1.229 (95% CI = 1.169-1.292) for every 1% HbA1c increase. In the spline model, the OR of LEA risk increased with HbA1c levels, especially when HbA1c ranged from 5% to 9%. Figure 4 shows the results of the dose–response meta-analysis for the relationship between HbA1c level and LEA risk.
Dose–Response Meta-Analysis for HbA1c Level and LEA Risk
Our meta-analysis showed the LEA risk increased with HbA1c level. Some studies also used the logistic regression method to investigate the relationship between HbA1c level and LEA in patients with diabetes. Davis reported the LEA OR as 1.30 [1.10-1.54] for every 1% HbA1c increase,21 and Moss reported the LEA OR as 1.39 [1.22-1.59] for every 1% HbA1c increase.27 Jbour reported the LEA OR as 1.38 [1.06-1.79] for every 1% HbA1c increase.30 The spline model of our dose–response meta-analysis showed the LEA OR as 1.229 (95% CI = 1.169-1.292) for every 1% HbA1c
Six studies15,20,24-27 (10 cohorts) with 109 933 patients included in the dose–response meta-analysis for HbA1c and LEA risk [10-15, 18-19]. The LEA incidence ranged from 0.3% to 14.6% between different HbA1c levels. Table 2 showed characteristics of the included studies. Dose– response meta-analysis showed statistically significant association between HbA1c and LEA risk (χ2 = 65.51, P = .000).
Discussion
Downloaded from ijl.sagepub.com at University of New England on July 22, 2015
171
Downloaded from ijl.sagepub.com at University of New England on July 22, 2015
Turkey Spain China India USA China Australia USA USA UK Finland
Tabur, 2014 Laclé, 2012 Fei, 2012 Zubair, 2012 Callaghan, 2011 Li, 2011 Davis, 2006 Olson, 2002 Watts, 2001 Coppini, 1998 Lehto, 1996
NR Prospective cohort Retrospective cohort Prospective cohort NR Retrospective cohort Prospective cohort Prospective cohort Retrospective cohort Prospective cohort Retrospective cohort
Study Design NR 2001-2007 2005-2011 2008-2011 1995-2006 2000-2009 1993-1996 1986-1988 NR 1982-1985 1982-1989
60.0 ± 9.4 60.6 ± 12.0 65 ± 12, 26-97 51.1 ± 11.4 Mean 59.4-60.8 65 64 ± 11 26 ± 8 67 ± 10.5 50 45-64
Age, Years 49.1 36.9 58.8 59.2 61 62.9 48 51 51 54 56
Type 2 NR NR Both Both NR Type 2 Type 1 NR Both Type 2
11.1 ± 8.2 12.9 ± 6.6 9 ± 7.3 13.5 ± 4.6 10 12 4 18 ± 7 NR 6 7.9-9.6
Diabetes Gender Diabetes Duration, (% Men) Type Years NR 7 NR NR 10 NR 9 10 NR 12 N7
55 572 685 162 28 701 450 1279 586 137 9895 1044
Follow- Number up of (Years) Patients
Abbreviations: HbA1c, glycated hemoglobin; LEA, lower extremity amputation; NOS, Newcastle–Ottawa Scale score; NR, not reported.
Country
Study
Enrollment Year
Table 1. Characteristics of the Included Studies Comparing HbA1c Levels Between Groups With and Without LEA.
10 21 78 46 981 79 44 70 23 20 58
n
n 9.7 ± 4.0 45 10.1 ± 2.3 528 10.2 ± 3.2 607 9.1 ± 2.16 116 9.05 ± 2.13 27 720 8.3 ± 2.3 310 11.1 ± 1.0 1235 10.9 ± 1.9 516 12.5 ± 2.5 110 12.5 ± 2.4 60 11.1 ± 0.3 986
With LEA
HbA1c (%)
10.5 ± 2.6 8.0 ± 2.2 8.8 ± 2.6 9.8 ± 1.9 8.36 ± 1.94 8.1 ± 2.2 7.4 ± 0.9 10.3 ± 1.8 11.3 ± 2.5 11.2 ± 2.7 9.8 ± 0.7
Without LEA
7 8 7 7 8 7 7 8 7 7 7
NOS
172
Downloaded from ijl.sagepub.com at University of New England on July 22, 2015
Prospective cohort; 8; 2000-2012
Prospective cohort; 8; 2000-2012
Prospective cohort; 8; 2000-2012
Prospective cohort; 8; 2000-2012
NR; 8; 19952006
NR; 7; 19781987
Zhao; 2013; USA (baseline African)
Zhao; 2013; USA (baseline White)
Zhao; 2013; USA (followup African)
Zhao; 2013; USA (followup White)
Callaghan; 2011; USA
Mühlhauser; 2000; Germany
First Author; Year; Country
Study Design; NOS; Enrollment Year
27.5 ± 9.5; 49.8% male
Mean 59.460.8; 61% male
Mean 51.153.7; 37.8% male
Mean 51.153.7; 37.8% male
Mean 51.153.7; 37.8% male
Mean 51.153.7; 37.8% male
Age; Gender
Type 1; 10.5 ± 9.5; 10.3 ± 3.4
Both; 10; 10
NR; NR; 6.83
NR; NR; 6.83
NR; NR; 6.83
NR; NR; 6.83
DM Type; DM Duration; Follow-up
research-article2015
IJLXXX10.1177/1534734615593190The International Journal of Lower Extremity WoundsZhou et al
Clinical and Translational Research
HbA1c and Lower Extremity Amputation Risk in Patients With Diabetes: A MetaAnalysis
The International Journal of Lower Extremity Wounds 2015, Vol. 14(2) 168–177 © The Author(s) 2015 Reprints and permissions: sagepub.com/journalsPermissions.nav DOI: 10.1177/1534734615593190 ijl.sagepub.com
Zhen-Yu Zhou, MD1,2, Ya-Ke Liu, MD2, Hong-Lin Chen, MD3, Hui-Lin Yang, MD1, and Fan Liu, MD2
Abstract In this meta-analysis, we aimed to assess glycosylated hemoglobin (HbA1c) level and lower extremity amputation (LEA) risk in patients with diabetes. Systematic computerized searches of the PubMed and Web of Knowledge were performed. We compared HbA1c level between groups with LEA and without LEA by meta-analysis; we also examined the dose–response relationship between HbA1c level and LEA risk. Sixteen studies were included in the meta-analysis. Eleven studies with 43 566 patients compared HbA1c between groups with and without LEA. The mean HbA1c (%) ranged from 8.3 to 12.5 in the group with LEA and from 7.4 to 11.3 in the group without LEA. The pooled weighted mean difference was 1.110 (95% confidence interval = 0.510-1.709; Z = 3.63, P = .008). The funnel plot was symmetrical, and Begg’s test (z = 0.00, P = 1.000) and Egger’s test (t = −0.02, P = .984) suggested no significant publication bias. Six studies with 109 933 patients included in the dose–response meta-analysis. The LEA incidence ranged from 0.3% to 14.6% between different HbA1c levels. Dose–response meta-analysis showed statistically significant association between HbA1c and LEA risk (χ2 = 65.51, P = .000). In linear model, the odds ratio for LEA incidence was 1.229 (95% confidence interval = 1.169-1.292) for every 1% HbA1c increase. In the spline model, the odds ratio of LEA risk increased with HbA1c levels, especially when HbA1c ranged from 5% to 9%. Our meta-analysis indicates that high level of HbA1c is an important risk factor for LEA in patients with diabetes. This evidence supports the strategy for lowering glucose levels to reduce amputation in patients with diabetes. Keywords diabetes mellitus, lower extremity amputation, risk factor, glycosylated hemoglobin, meta-analysis
Foot ulcers are one of the most serious complications of diabetes mellitus, with the prevalence of 4% to 10% among persons diagnosed as having diabetes mellitus.1 Foot ulcers frequently become infected, cause morbidity, engender financial costs, and lead to lower extremity amputation (LEA).1 Studies showed the incidence of all forms of LEA ranges from 46.1 to 9600 per 100 000 in diabetes patients, and major amputation ranges from 5.6 to 600 per 100 000.2 Most important, LEA is companied with a significantly elevated mortality at follow-up, ranging from 13% to 40% at 1 year, 35% to 65% at 3 years, and 39% to 80% at 5 years—worse than for most malignancies.1 LEA also severely reduces quality of life and has major economic consequences. The mean cost per inpatient stay was $17 103 according to the US Agency for Healthcare Research and Quality Healthcare Cost and Utilization Project.3 A study in Germany showed costs up to 24 weeks after LEA were 36 686 Euros and 115 676 Euros for 3 years after LEA.4 Many studies have investigated the risk factors for LEA in diabetes patients, and the results showed LEAs
were independently associated with higher age, male gender, longer diabetes duration, diabetic polyneuropathy, and poor glycemic control.5,6 There is a significant increase in apoptosis in diabetic wounds with poorly controlled blood sugar, which contributes to delayed wound healing.7 Studies also showed patients with poor glycemic control had higher percentage of minor and major LEA.8 Glycosylated hemoglobin (HbA1c) was an important serum marker for monitoring the degree of control of glucose in diabetic patients. In diabetes mellitus, higher amounts of HbA1c, indicating poorer 1
First Affiliated Hospital of Soochow University, Suzhou, Jiangsu, PR China Affiliated Hospital of Nantong University, Nantong, Jiangsu, PR China 3 Nantong University, Nantong, Jiangsu, PR China 2
Corresponding Author: Fan Liu, Department of Orthopaedics, Affiliated Hospital of Nantong University, Nantong 226001, Jiangsu Province, People’s Republic of China. Email: [email protected]
Downloaded from ijl.sagepub.com at University of New England on July 22, 2015
169
Zhou et al control of blood glucose levels, have been associated with diabetes complications, such as cardiovascular disease, microalbuminuria, and the progression of renal status.9 However, the relationship between HbA1c and LEA in patients with diabetes is still unknown. The aim of this meta-analysis was to determine whether HbA1c is associated with LEA in patients with diabetes, and to establish the dose–response relationship of this impact.
Methods Databases and Literature Search We searched MEDLINE and Web of Science up to December 12, 2014. The search terms included of “Diabetes,” “Amputation,” “Glycosylated Hemoglobin,” and “HbA1c.” The search detail in MEDLINE was ((Diabete*[Title/ Abstract]) AND Amputation [Title/Abstract]) AND (Glycosylated hemoglobin [Tw] OR HbA1c [Tw]). In Web of Science, the search detail used was the following: (Ts= Diabete*) and (Ts= Amputation) and (Ts= Glycosylated hemoglobin or Ts= HbA1c). We supplemented our searches by manually reviewing the references of all relevant studies.
Statistical Analysis We analyzed the data by 2 stages. First, we assessed HbA1c level difference between groups with and without LEA. Fixed effect model or random effect model was chose according to heterogeneity. Weighted mean difference (WMD) was used for assessing difference, and overall effects were determined using the Z test. Visual inspection of a funnel plot, the Egger test, and Begg test were performed to assess publication bias. Sensitivity analyses were performed by excluding the study with largest sample. Second, we examined the dose–response relationship between different HbA1c categories and LEA risk. The dose–response meta-analysis was carried out using the method proposed by Greenland and Longnecker11 and Orsini et al.12,13 We examined linear and spline dose– response relationships. The spline relationship was examined by HbA1c category using restricted cubic splines with 3 knots. All statistical analyses were performed with Stata software, version 13.0 (Stata Corp, College Station, TX). Two-sided P < .050 was considered statistically significant.
Results
Eligibility Criteria The following inclusion criteria had to be fulfilled: (a) observational studies (cross-sectional, case–control, or cohort study) that permitted assessment of a causal association between HbA1c and LEA risk in patients with diabetes; (b) LEA included not only major amputation but also minor amputation (major amputation was defined as any resection proximal to midtarsal level, and minor amputation was defined as midtarsal or distal amputations); (c) compared HbA1c levels between groups with and without LEA; or (d) reported number of cases (LEA) and controls (no LEA) for each HbA1c category, which included at least 3 HbA1c categorized levels.
Data Extraction and Critical Appraisal The following data were abstracted onto standardized forms: first author, publication year, country, study design, enrollment year, age and gender of participants, diabetes type, diabetes duration, follow-up years, the HbA1c levels in groups with and without LEA, and number of patients in case or control group in each HbA1c category. The quality of case–control and cohort study was assessed according to the Newcastle–Ottawa Scale (NOS).10 The NOS contains 8 items, categorized into 3 dimensions: Selection (4), Comparability (1), and Exposure (3). The NOS ranges between 0 and 9. Data extraction and article quality assessment were carried out independently by 2 reviewers. Disagreements were resolved by discussion.
Characteristics of Included Studies We identified 16 studies14-29 that met our inclusion criteria for meta-analysis. The detailed steps of our literature search are shown in Figure 1. The studies included 10 prospective cohorts, 4 retrospective cohorts, and the remaining 2 studies did not report details of the the study design. Of these studies, 6 were conducted in the United States, 2 in the United Kingdom, 2 in China; Turkey, Australia, Finland, Germany, Spain, and India contributed 1 study each. The enrollment year ranged from 1982 to 2011. All studies investigated the relationship in adult, and did not include children. The mean age ranged from 26 to 65 years. The male patients accounted for 36.9% to 62.9% of the proportion. Type 2 diabetes was assessed in 3 studies, type 1 in 1 study, type 1 and type 2 in 3 studies, and the remaining 9 studies did not report the diabetes type. The quality rating of the included studies ranged from 6 to 8 on a scale of 9. Tables 1 and 2 show the characteristics of the 16 identified studies.
Meta-Analysis for HbA1c Difference Between Groups With and Without LEA Eleven studies14,16-23,28,29 with 43 566 patients included in the meta-analysis compared HbA1c level differences between the 2 groups. The HbA1c level (%) ranged from 8.3 to 12.5 in the group with LEA, and from 7.4 to 11.3 in the group without LEA. Significant heterogeneity was found between the studies (I2 = 97.4%, P = .000), and the
Downloaded from ijl.sagepub.com at University of New England on July 22, 2015
170
The International Journal of Lower Extremity Wounds 14(2)
Records identified through Pubmed searching (n=83) Records identified through Web of Science searching (n=67)
Additional records identified through meeting abstract (n =3)
Records duplicates removed (n = 61)
Records screened (n =92)
Records excluded by title and abstract review Animal experiments and biological research (n=4) Reviews, case reports and case series (n=14) Re-amputation risk (n=2) No available data (n=35)
Full-text articles assessed (n=37)
Full-text articles excluded No available data (n=21)
Studies included in quantitative synthesis (meta-analysis) (n=16) Figure 1. Flow diagram showing selection of studies.
random effect model was used. The pooled WMD was 1.110 (95% confidence interval [CI] = 0.510-1.709; Z = 3.63, P = .008; Figure 2). The funnel plot was symmetrical (Figure 3), and Begg’s test (z = 0.00, P = 1.000) and Egger’s test (t = −0.02, P = .984) suggested there was no significant publication bias. Sensitivity analyses by excluding the study with largest sample showed the pooled WMD was 1.124 (95% CI = 0.268-1.981; Z = 2.57, P = .010).
In the linear model, the LEA incidence OR was 1.229 (95% CI = 1.169-1.292) for every 1% HbA1c increase. In the spline model, the OR of LEA risk increased with HbA1c levels, especially when HbA1c ranged from 5% to 9%. Figure 4 shows the results of the dose–response meta-analysis for the relationship between HbA1c level and LEA risk.
Dose–Response Meta-Analysis for HbA1c Level and LEA Risk
Our meta-analysis showed the LEA risk increased with HbA1c level. Some studies also used the logistic regression method to investigate the relationship between HbA1c level and LEA in patients with diabetes. Davis reported the LEA OR as 1.30 [1.10-1.54] for every 1% HbA1c increase,21 and Moss reported the LEA OR as 1.39 [1.22-1.59] for every 1% HbA1c increase.27 Jbour reported the LEA OR as 1.38 [1.06-1.79] for every 1% HbA1c increase.30 The spline model of our dose–response meta-analysis showed the LEA OR as 1.229 (95% CI = 1.169-1.292) for every 1% HbA1c
Six studies15,20,24-27 (10 cohorts) with 109 933 patients included in the dose–response meta-analysis for HbA1c and LEA risk [10-15, 18-19]. The LEA incidence ranged from 0.3% to 14.6% between different HbA1c levels. Table 2 showed characteristics of the included studies. Dose– response meta-analysis showed statistically significant association between HbA1c and LEA risk (χ2 = 65.51, P = .000).
Discussion
Downloaded from ijl.sagepub.com at University of New England on July 22, 2015
171
Downloaded from ijl.sagepub.com at University of New England on July 22, 2015
Turkey Spain China India USA China Australia USA USA UK Finland
Tabur, 2014 Laclé, 2012 Fei, 2012 Zubair, 2012 Callaghan, 2011 Li, 2011 Davis, 2006 Olson, 2002 Watts, 2001 Coppini, 1998 Lehto, 1996
NR Prospective cohort Retrospective cohort Prospective cohort NR Retrospective cohort Prospective cohort Prospective cohort Retrospective cohort Prospective cohort Retrospective cohort
Study Design NR 2001-2007 2005-2011 2008-2011 1995-2006 2000-2009 1993-1996 1986-1988 NR 1982-1985 1982-1989
60.0 ± 9.4 60.6 ± 12.0 65 ± 12, 26-97 51.1 ± 11.4 Mean 59.4-60.8 65 64 ± 11 26 ± 8 67 ± 10.5 50 45-64
Age, Years 49.1 36.9 58.8 59.2 61 62.9 48 51 51 54 56
Type 2 NR NR Both Both NR Type 2 Type 1 NR Both Type 2
11.1 ± 8.2 12.9 ± 6.6 9 ± 7.3 13.5 ± 4.6 10 12 4 18 ± 7 NR 6 7.9-9.6
Diabetes Gender Diabetes Duration, (% Men) Type Years NR 7 NR NR 10 NR 9 10 NR 12 N7
55 572 685 162 28 701 450 1279 586 137 9895 1044
Follow- Number up of (Years) Patients
Abbreviations: HbA1c, glycated hemoglobin; LEA, lower extremity amputation; NOS, Newcastle–Ottawa Scale score; NR, not reported.
Country
Study
Enrollment Year
Table 1. Characteristics of the Included Studies Comparing HbA1c Levels Between Groups With and Without LEA.
10 21 78 46 981 79 44 70 23 20 58
n
n 9.7 ± 4.0 45 10.1 ± 2.3 528 10.2 ± 3.2 607 9.1 ± 2.16 116 9.05 ± 2.13 27 720 8.3 ± 2.3 310 11.1 ± 1.0 1235 10.9 ± 1.9 516 12.5 ± 2.5 110 12.5 ± 2.4 60 11.1 ± 0.3 986
With LEA
HbA1c (%)
10.5 ± 2.6 8.0 ± 2.2 8.8 ± 2.6 9.8 ± 1.9 8.36 ± 1.94 8.1 ± 2.2 7.4 ± 0.9 10.3 ± 1.8 11.3 ± 2.5 11.2 ± 2.7 9.8 ± 0.7
Without LEA
7 8 7 7 8 7 7 8 7 7 7
NOS
172
Downloaded from ijl.sagepub.com at University of New England on July 22, 2015
Prospective cohort; 8; 2000-2012
Prospective cohort; 8; 2000-2012
Prospective cohort; 8; 2000-2012
Prospective cohort; 8; 2000-2012
NR; 8; 19952006
NR; 7; 19781987
Zhao; 2013; USA (baseline African)
Zhao; 2013; USA (baseline White)
Zhao; 2013; USA (followup African)
Zhao; 2013; USA (followup White)
Callaghan; 2011; USA
Mühlhauser; 2000; Germany
First Author; Year; Country
Study Design; NOS; Enrollment Year
27.5 ± 9.5; 49.8% male
Mean 59.460.8; 61% male
Mean 51.153.7; 37.8% male
Mean 51.153.7; 37.8% male
Mean 51.153.7; 37.8% male
Mean 51.153.7; 37.8% male
Age; Gender
Type 1; 10.5 ± 9.5; 10.3 ± 3.4
Both; 10; 10
NR; NR; 6.83
NR; NR; 6.83
NR; NR; 6.83
NR; NR; 6.83
DM Type; DM Duration; Follow-up
