771
Journal of Alzheimer’s Disease 46 (2015) 771–776 DOI 10.3233/JAD-141551 IOS Press
SORL1 Gene is Associated with the Conversion from Mild Cognitive Impairment to Alzheimer’s Disease Paola Piscopoa , Giuseppe Tostob , Chiara Bellia , Giuseppina Talaricob , Daniela Galimbertic , Marina Gasparinib , Marco Canevellib , Anna Poleggia , Alessio Crestinia , Diego Albanid , Gianluigi Forlonid , Ugo Luccad , Pierluigi Quadrif , Mauro Tettamantid , Chiara Fenoglioc , Elio Scarpinic , Giuseppe Brunob , Nicola Vanacoree and Annamaria Confalonia,∗ a Department
of Cell Biology and Neuroscience, Istituto Superiore di Sanit`a, Rome, Italy Clinic, Department of Neurology and Psychiatry, University of Rome “Sapienza”, Italy c Department of Pathophysiology and Transplantation, “Dino Ferrari” Center, University of Milan, Fondazione C`a Granda, IRCCS Ospedale Maggiore Policlinico, Milan, Italy d IRCCS – Istituto di Ricerche Farmacologiche “Mario Negri”, Milan, Italy e Department of National Centre of Epidemiology, Surveillance and Health Promotion, Istituto Superiore di Sanit` a, Rome, Italy f Geriatric Division, Ospedali Regionali of Lugano and Mendrisio, Switzerland b Memory
Handling Associate Editor: Emilio Di Maria
Accepted 18 March 2015
Abstract. Several studies have established the sortilin-related receptor gene (SORL1) as a susceptibility locus for Alzheimer’s disease (AD). Single nucleotide polymorphisms of SORL1 reported in literature as being associated with AD were investigated in an Italian case-control data set, and their role as a risk factor of conversion to AD was studied in an independent sample of subjects diagnosed with mild cognitive impairment (MCI) at baseline. rs641120, rs2070045, and rs1010159 were genotyped in 734 subjects diagnosed with AD (n = 338) and MCI (n = 181) and in healthy controls (n = 215). Our results confirmed the association between rs641120 and AD (p = 0.01). In the MCI cohort, rs1010159 was associated with conversion to AD (HR = 1.56, p = 0.002). Taken together, these findings confirm that SORL1 is associated with AD and might be a potential tool for identifying MCI subjects at high risk of conversion to AD. Keywords: Alzheimer’s disease, mild cognitive impairment, single nucleotide polymorphisms, SORL1
INTRODUCTION The pathogenic process of Alzheimer’s disease (AD) starts decades before the clinical onset of the disease [1]. During this preclinical phase, multiple brain structures and functions are affected long before the ∗ Correspondence to: Dr. Annamaria Confaloni, Department of Cell Biology and Neuroscience, Istituto Superiore di Sanit`a, Viale Regina Elena, 299, 00161 Rome, Italy. Tel.: +39 06 49902930; Fax: +39 06 49902040; E-mail: [email protected].
diagnosis of AD, and at a certain threshold, the first symptoms appear. At this stage, patients, who do not yet fulfill the criteria for dementia, may be diagnosed with mild cognitive impairment (MCI). Percentages of conversion from MCI to AD over 1 year range from 10.2 to 33.6% and from 9.8 to 36.3% over 2 years [2]. SORL1 is a sorting receptor that regulates trafficking and processing of amyloid- protein precursor (APP), acting as a retention factor for APP in
ISSN 1387-2877/15/$35.00 © 2015 – IOS Press and the authors. All rights reserved
772
P. Piscopo et al. / SORL1 Gene Polymorphisms and MCI
trans-Golgi compartments/trans-Golgi network and preventing the release of the precursor into regular processing pathways [3, 4]. According to the “amyloid cascade” model, APP is synthesized in the endoplasmic reticulum and Golgi, and is the target of ␣- or -secretase first and, subsequently, of the ␥-secretase complex. This - and ␥-secretase cleavage ultimately produces A fragments. Alternatively, SORL1 binds APP holoprotein acting as a sorting receptor. Absence of SORL1 switches APP away from the retromer recycling pathway, directing it into the -secretase pathway and ultimately increasing the production of toxic A peptide [5]. To date, SORL1 gene has been investigated in a several papers, as reported at https://www.alzgene.org where 19 case-control or family-based studies are currently listed and 14 additional association-studies are reported in the literature. Moreover, SORL1 has been associated with AD in several genome-wide association studies [5–7]. In particular, two clusters of different single-nucleotide polymorphisms (SNPs) located at the 3’ and 5’ terminus of SORL1 have been validated [5]. Interestingly, Sager and colleagues described a reduction of neuronal SORL1 levels in prodromal AD [8]; a correlation between its expression and cognitive performances suggests that SORL1 might predict conversion to AD in a subgroup of individuals with MCI. In the present study we investigated the relationship between the SORL1 polymorphisms and MCI in order to evaluate their role as risk factors for the conversion from MCI status to AD.
SUBJECTS AND METHODS Study population 734 subjects were enrolled in the present study, including subjects diagnosed at baseline with AD (n = 338) and MCI (n = 181), as well as healthy agematched controls (n = 215). Subjects were recruited at the Memory Clinic of University “Sapienza” (Rome, Italy), at the Alzheimer’s Units of Ospedale Maggiore Policlinico (Milan, Italy) and at the Geriatric Division of the Ospedali Riuniti of Lugano and Mendrisio (Canton Ticino, Switzerland; Italian language). The study was approved by the Institutional Review Board of all centers and a written informed consent approved by the ethics committees was obtained from all participants involved in this study.
The clinical diagnosis of probable AD was defined according to the DSM-IV and Alzheimer’s disease and Related Disorders Association (NINCDS-ADRDA) criteria [9] in a consensus diagnostic conference by a panel of neurologists and neuropsychologists. A detailed clinical history (with special focus to comorbidities and concomitant pharmacological treatments) was recorded from patients and/or first-degree relatives. All the participants underwent general physical and neurological examination, laboratory tests to exclude secondary cause of dementia, and to assess the following clinical assessments: 1) global cognitive functioning by the Mini-Mental State Examination [10] and neuropsychological assessment, with raw scores adjusted for demographic variables and compared with cut-off scores according to normative data for the Italian population [11, 12]; 2) functional independence, measured by the Activities of Daily Living and Instrumental Activities of Daily Living scales; 3) neuropsychiatric symptoms, evaluated through the Neuropsychiatric Inventory or 9-item scale Spontaneous Behavior Interview - Behavioral Problems. MCI subjects met the criteria described by Petersen [13] with a mean period of clinical observation of 5 years. The healthy control groups were enrolled among healthy volunteers: family members or caregivers unrelated to the AD patients, all undergoing a clinical interview with a neurologist with a completely normal cognitive and functional assessment. Genetic study SNPs rs641120, rs2070045, and rs1010159 were selected for the present study. According to previous reports and to the HapMap project, rs641120 and rs1010159 belong to two LD block (located at the 5’ and 3’) within the SORL1 gene previously associated with AD [5]. We also included the SNP rs2070045, which was reported in a recent meta-analysis as the most statistically significant [14]. A complete description of the selected SNPs, their location within the gene and the gene’s LD pattern can be obtained at http://www.broadinstitute.org/mpg/snap/ldplot.php. Genomic DNA was extracted from blood samples; genotyping of the SORL1 variants was carried out by polymerase chain reaction and high-resolution melting analysis [15] using the real time cycler (Rotor-Gene 6000; Corbett Life Science, Mortlake, Australia). To obtain best results in HRM analysis, we designed a primer set for each SORL1 SNP using Primer3 software (available free: http://bioinfo.ut.ee/primer3-0.4.0/)
P. Piscopo et al. / SORL1 Gene Polymorphisms and MCI
[16] to create amplicons smaller than 250 bp. Then, DNA was amplified in 20 l reaction volumes containing: 100 ng of genomic DNA, 1× buffer, 2.5 mM MgCl2 , 0.2 mM of each dNTP, 1 pmol/ml of each primer, 0.2 l AmpliTaq Gold (Applied Biosystems, Carlsbad, California, USA) and 1 M of the fluorescent dye EvaGreen. The PCR cycling conditions were: one cycle at 94◦ C for 10 min; 35 cycles at 94◦ C for 30 s 60◦ C for 45 s, 72◦ C for 45 s; and a final step at 72◦ C for 10 min. Thereafter the products underwent a melt with a range of temperature specific for each SNP rising by 0.1◦ C each step on a Rotor-Gene 6000 analyzer. APOE was genotyped by using polymerase chain reaction and HhaI restriction enzyme digestion, as previously reported [17].
773
RESULTS Sample’s characteristics Demographic and clinical features of patients and controls are summarized in Table 1. No differences in terms of age at baseline and gender were observed between AD subjects, MCI subjects, and healthy controls. Genotyping We genotyped three SNPs in the SORL1 locus: rs641120, rs2070045, and rs1010159. The genotype and allelic frequencies are shown in Table 2. None of the variants deviated from Hardy–Weinberg equilibrium [20] in the healthy control group.
Statistical analysis First, SNPs were assessed for deviations from Hardy-Weinberg equilibrium. Multivariate logistic regression analyses in PLINK (http://pngu.mgh. harvard.edu/∼purcell/plink/) were run with an additive genetic model (i.e., according to the number of reference alleles SNPs were coded as 0/1/2), adjusting for gender and age (Model 1) and gender, age, and APOE genotype (modelled as having at least one copy of the 4 allele). The False Discovery Rate was used to correct for multiple testing [18]. Extent of linkage disequilibrium (LD) between markers was assessed using the R2 in Haploview [19]. The SNPs were also examined for associations with conversion to AD in the MCI group employing multiple Cox’s proportional hazard models in R (http://www.r-project.org/), adjusting for gender, age at baseline (Model 1) and gender, age at baseline, and APOE 4 (Model 2).
SORL1 and AD SNP rs641120 was significantly associated with AD in both models investigated (Model 1: OR = 1.39, p = 0.01; Model 2: OR = 1.38, p = 0.01) (Table 2). SORL1 and MCI At the end of the 5-year follow-up, 108 out of 182 subjects diagnosed with MCI at baseline converted to AD (60% conversion rate; median 45.5, interquartile range 22–60). rs1010159 was significantly associated with higher hazardous of converting to AD (HR = 1.56; p = 0.002) in Model 1. Results were further confirmed in Model 2 (HR = 1.74; p < 0.001) (Table 3). Survival curves for the MCI subjects according to their rs1010159 genotype are plotted in Fig. 1.
Table 1 Demographic and clinical features of study subjects, stratified by affection status Groups Healthy controls AD MCI Non converter MCI converter Total
n
Gender (M/F)
Age (mean ± SD)
APOE 4 (n)
215 338 181 74 108 734
76/138 108/230 75/107 29/45 46/62 259/475
74.19 ± 9.19 74.64 ± 8.03 75.60 ± 7.53 75.96 ± 7.23 75.50 ± 7.83 74.68 ± 8.27
29 136 63 45 18 228
Table 2 Multivariate logistic regression model for the case-control data set SNP rs641120 rs2070045 rs1010159
BP
N
A1
A2
MAF
121510256 121577381 121612692
551 552 545
T G G
C C A
0.36 0.19 0.30
MODEL 1a
MODEL 2b
OR (CI 95%)
p
OR (CI 95%)
p
1.39 (1.08–1.79) 0.95 (0.70–1.28) 1.25 (0.97–1.63)
0.01 0.74 0.09
1.38 (1.06–2.33) 1.03 (0.74–1.41) 1.28 (0.97–1.68)
0.01 0.86 0.08
a adjusted for age and gender. b adjusted foe age, gender, APOE 4. BP, base pair; A1, minor allele; A2, major allele; MAF, minor allele frequency;
OR, odd ratio; CI, confidence interval. In bold p < 0.05 values.
774
P. Piscopo et al. / SORL1 Gene Polymorphisms and MCI
Table 3 Multivariate Cox regression analysis. Separate models were run for each SORL1 variants as main predictors and conversion from MCI status to AD as the selected outcome MODEL 1a Gender Age APOE 4 rs641120 rs2070045 rs1010159
MODEL 2b
HR
CI 95%
p
HR
1.1 0.99 – 1.32 0.96 1.56
0.73–1.65 0.96–1.02 – 0.99–1.75 0.69–1.33 1.18–2.07
0.65 0.72 – 0.053 0.80 0.002
1.1 0.99 1.54 1.24 1.03 1.74
CI 95%
p
0.76–1.70 0.52 0.96–1.02 0.56 1.03–2.3 0.03 0.93–1.66 0.14 1.00–1.43 0.88 1.29–2.34
Journal of Alzheimer’s Disease 46 (2015) 771–776 DOI 10.3233/JAD-141551 IOS Press
SORL1 Gene is Associated with the Conversion from Mild Cognitive Impairment to Alzheimer’s Disease Paola Piscopoa , Giuseppe Tostob , Chiara Bellia , Giuseppina Talaricob , Daniela Galimbertic , Marina Gasparinib , Marco Canevellib , Anna Poleggia , Alessio Crestinia , Diego Albanid , Gianluigi Forlonid , Ugo Luccad , Pierluigi Quadrif , Mauro Tettamantid , Chiara Fenoglioc , Elio Scarpinic , Giuseppe Brunob , Nicola Vanacoree and Annamaria Confalonia,∗ a Department
of Cell Biology and Neuroscience, Istituto Superiore di Sanit`a, Rome, Italy Clinic, Department of Neurology and Psychiatry, University of Rome “Sapienza”, Italy c Department of Pathophysiology and Transplantation, “Dino Ferrari” Center, University of Milan, Fondazione C`a Granda, IRCCS Ospedale Maggiore Policlinico, Milan, Italy d IRCCS – Istituto di Ricerche Farmacologiche “Mario Negri”, Milan, Italy e Department of National Centre of Epidemiology, Surveillance and Health Promotion, Istituto Superiore di Sanit` a, Rome, Italy f Geriatric Division, Ospedali Regionali of Lugano and Mendrisio, Switzerland b Memory
Handling Associate Editor: Emilio Di Maria
Accepted 18 March 2015
Abstract. Several studies have established the sortilin-related receptor gene (SORL1) as a susceptibility locus for Alzheimer’s disease (AD). Single nucleotide polymorphisms of SORL1 reported in literature as being associated with AD were investigated in an Italian case-control data set, and their role as a risk factor of conversion to AD was studied in an independent sample of subjects diagnosed with mild cognitive impairment (MCI) at baseline. rs641120, rs2070045, and rs1010159 were genotyped in 734 subjects diagnosed with AD (n = 338) and MCI (n = 181) and in healthy controls (n = 215). Our results confirmed the association between rs641120 and AD (p = 0.01). In the MCI cohort, rs1010159 was associated with conversion to AD (HR = 1.56, p = 0.002). Taken together, these findings confirm that SORL1 is associated with AD and might be a potential tool for identifying MCI subjects at high risk of conversion to AD. Keywords: Alzheimer’s disease, mild cognitive impairment, single nucleotide polymorphisms, SORL1
INTRODUCTION The pathogenic process of Alzheimer’s disease (AD) starts decades before the clinical onset of the disease [1]. During this preclinical phase, multiple brain structures and functions are affected long before the ∗ Correspondence to: Dr. Annamaria Confaloni, Department of Cell Biology and Neuroscience, Istituto Superiore di Sanit`a, Viale Regina Elena, 299, 00161 Rome, Italy. Tel.: +39 06 49902930; Fax: +39 06 49902040; E-mail: [email protected].
diagnosis of AD, and at a certain threshold, the first symptoms appear. At this stage, patients, who do not yet fulfill the criteria for dementia, may be diagnosed with mild cognitive impairment (MCI). Percentages of conversion from MCI to AD over 1 year range from 10.2 to 33.6% and from 9.8 to 36.3% over 2 years [2]. SORL1 is a sorting receptor that regulates trafficking and processing of amyloid- protein precursor (APP), acting as a retention factor for APP in
ISSN 1387-2877/15/$35.00 © 2015 – IOS Press and the authors. All rights reserved
772
P. Piscopo et al. / SORL1 Gene Polymorphisms and MCI
trans-Golgi compartments/trans-Golgi network and preventing the release of the precursor into regular processing pathways [3, 4]. According to the “amyloid cascade” model, APP is synthesized in the endoplasmic reticulum and Golgi, and is the target of ␣- or -secretase first and, subsequently, of the ␥-secretase complex. This - and ␥-secretase cleavage ultimately produces A fragments. Alternatively, SORL1 binds APP holoprotein acting as a sorting receptor. Absence of SORL1 switches APP away from the retromer recycling pathway, directing it into the -secretase pathway and ultimately increasing the production of toxic A peptide [5]. To date, SORL1 gene has been investigated in a several papers, as reported at https://www.alzgene.org where 19 case-control or family-based studies are currently listed and 14 additional association-studies are reported in the literature. Moreover, SORL1 has been associated with AD in several genome-wide association studies [5–7]. In particular, two clusters of different single-nucleotide polymorphisms (SNPs) located at the 3’ and 5’ terminus of SORL1 have been validated [5]. Interestingly, Sager and colleagues described a reduction of neuronal SORL1 levels in prodromal AD [8]; a correlation between its expression and cognitive performances suggests that SORL1 might predict conversion to AD in a subgroup of individuals with MCI. In the present study we investigated the relationship between the SORL1 polymorphisms and MCI in order to evaluate their role as risk factors for the conversion from MCI status to AD.
SUBJECTS AND METHODS Study population 734 subjects were enrolled in the present study, including subjects diagnosed at baseline with AD (n = 338) and MCI (n = 181), as well as healthy agematched controls (n = 215). Subjects were recruited at the Memory Clinic of University “Sapienza” (Rome, Italy), at the Alzheimer’s Units of Ospedale Maggiore Policlinico (Milan, Italy) and at the Geriatric Division of the Ospedali Riuniti of Lugano and Mendrisio (Canton Ticino, Switzerland; Italian language). The study was approved by the Institutional Review Board of all centers and a written informed consent approved by the ethics committees was obtained from all participants involved in this study.
The clinical diagnosis of probable AD was defined according to the DSM-IV and Alzheimer’s disease and Related Disorders Association (NINCDS-ADRDA) criteria [9] in a consensus diagnostic conference by a panel of neurologists and neuropsychologists. A detailed clinical history (with special focus to comorbidities and concomitant pharmacological treatments) was recorded from patients and/or first-degree relatives. All the participants underwent general physical and neurological examination, laboratory tests to exclude secondary cause of dementia, and to assess the following clinical assessments: 1) global cognitive functioning by the Mini-Mental State Examination [10] and neuropsychological assessment, with raw scores adjusted for demographic variables and compared with cut-off scores according to normative data for the Italian population [11, 12]; 2) functional independence, measured by the Activities of Daily Living and Instrumental Activities of Daily Living scales; 3) neuropsychiatric symptoms, evaluated through the Neuropsychiatric Inventory or 9-item scale Spontaneous Behavior Interview - Behavioral Problems. MCI subjects met the criteria described by Petersen [13] with a mean period of clinical observation of 5 years. The healthy control groups were enrolled among healthy volunteers: family members or caregivers unrelated to the AD patients, all undergoing a clinical interview with a neurologist with a completely normal cognitive and functional assessment. Genetic study SNPs rs641120, rs2070045, and rs1010159 were selected for the present study. According to previous reports and to the HapMap project, rs641120 and rs1010159 belong to two LD block (located at the 5’ and 3’) within the SORL1 gene previously associated with AD [5]. We also included the SNP rs2070045, which was reported in a recent meta-analysis as the most statistically significant [14]. A complete description of the selected SNPs, their location within the gene and the gene’s LD pattern can be obtained at http://www.broadinstitute.org/mpg/snap/ldplot.php. Genomic DNA was extracted from blood samples; genotyping of the SORL1 variants was carried out by polymerase chain reaction and high-resolution melting analysis [15] using the real time cycler (Rotor-Gene 6000; Corbett Life Science, Mortlake, Australia). To obtain best results in HRM analysis, we designed a primer set for each SORL1 SNP using Primer3 software (available free: http://bioinfo.ut.ee/primer3-0.4.0/)
P. Piscopo et al. / SORL1 Gene Polymorphisms and MCI
[16] to create amplicons smaller than 250 bp. Then, DNA was amplified in 20 l reaction volumes containing: 100 ng of genomic DNA, 1× buffer, 2.5 mM MgCl2 , 0.2 mM of each dNTP, 1 pmol/ml of each primer, 0.2 l AmpliTaq Gold (Applied Biosystems, Carlsbad, California, USA) and 1 M of the fluorescent dye EvaGreen. The PCR cycling conditions were: one cycle at 94◦ C for 10 min; 35 cycles at 94◦ C for 30 s 60◦ C for 45 s, 72◦ C for 45 s; and a final step at 72◦ C for 10 min. Thereafter the products underwent a melt with a range of temperature specific for each SNP rising by 0.1◦ C each step on a Rotor-Gene 6000 analyzer. APOE was genotyped by using polymerase chain reaction and HhaI restriction enzyme digestion, as previously reported [17].
773
RESULTS Sample’s characteristics Demographic and clinical features of patients and controls are summarized in Table 1. No differences in terms of age at baseline and gender were observed between AD subjects, MCI subjects, and healthy controls. Genotyping We genotyped three SNPs in the SORL1 locus: rs641120, rs2070045, and rs1010159. The genotype and allelic frequencies are shown in Table 2. None of the variants deviated from Hardy–Weinberg equilibrium [20] in the healthy control group.
Statistical analysis First, SNPs were assessed for deviations from Hardy-Weinberg equilibrium. Multivariate logistic regression analyses in PLINK (http://pngu.mgh. harvard.edu/∼purcell/plink/) were run with an additive genetic model (i.e., according to the number of reference alleles SNPs were coded as 0/1/2), adjusting for gender and age (Model 1) and gender, age, and APOE genotype (modelled as having at least one copy of the 4 allele). The False Discovery Rate was used to correct for multiple testing [18]. Extent of linkage disequilibrium (LD) between markers was assessed using the R2 in Haploview [19]. The SNPs were also examined for associations with conversion to AD in the MCI group employing multiple Cox’s proportional hazard models in R (http://www.r-project.org/), adjusting for gender, age at baseline (Model 1) and gender, age at baseline, and APOE 4 (Model 2).
SORL1 and AD SNP rs641120 was significantly associated with AD in both models investigated (Model 1: OR = 1.39, p = 0.01; Model 2: OR = 1.38, p = 0.01) (Table 2). SORL1 and MCI At the end of the 5-year follow-up, 108 out of 182 subjects diagnosed with MCI at baseline converted to AD (60% conversion rate; median 45.5, interquartile range 22–60). rs1010159 was significantly associated with higher hazardous of converting to AD (HR = 1.56; p = 0.002) in Model 1. Results were further confirmed in Model 2 (HR = 1.74; p < 0.001) (Table 3). Survival curves for the MCI subjects according to their rs1010159 genotype are plotted in Fig. 1.
Table 1 Demographic and clinical features of study subjects, stratified by affection status Groups Healthy controls AD MCI Non converter MCI converter Total
n
Gender (M/F)
Age (mean ± SD)
APOE 4 (n)
215 338 181 74 108 734
76/138 108/230 75/107 29/45 46/62 259/475
74.19 ± 9.19 74.64 ± 8.03 75.60 ± 7.53 75.96 ± 7.23 75.50 ± 7.83 74.68 ± 8.27
29 136 63 45 18 228
Table 2 Multivariate logistic regression model for the case-control data set SNP rs641120 rs2070045 rs1010159
BP
N
A1
A2
MAF
121510256 121577381 121612692
551 552 545
T G G
C C A
0.36 0.19 0.30
MODEL 1a
MODEL 2b
OR (CI 95%)
p
OR (CI 95%)
p
1.39 (1.08–1.79) 0.95 (0.70–1.28) 1.25 (0.97–1.63)
0.01 0.74 0.09
1.38 (1.06–2.33) 1.03 (0.74–1.41) 1.28 (0.97–1.68)
0.01 0.86 0.08
a adjusted for age and gender. b adjusted foe age, gender, APOE 4. BP, base pair; A1, minor allele; A2, major allele; MAF, minor allele frequency;
OR, odd ratio; CI, confidence interval. In bold p < 0.05 values.
774
P. Piscopo et al. / SORL1 Gene Polymorphisms and MCI
Table 3 Multivariate Cox regression analysis. Separate models were run for each SORL1 variants as main predictors and conversion from MCI status to AD as the selected outcome MODEL 1a Gender Age APOE 4 rs641120 rs2070045 rs1010159
MODEL 2b
HR
CI 95%
p
HR
1.1 0.99 – 1.32 0.96 1.56
0.73–1.65 0.96–1.02 – 0.99–1.75 0.69–1.33 1.18–2.07
0.65 0.72 – 0.053 0.80 0.002
1.1 0.99 1.54 1.24 1.03 1.74
CI 95%
p
0.76–1.70 0.52 0.96–1.02 0.56 1.03–2.3 0.03 0.93–1.66 0.14 1.00–1.43 0.88 1.29–2.34
