Haptoglobin Genotype Does Not Confer a Risk of Stroke in Type 1 Diabetes

Type 1 diabetes markedly increases the risk of stroke. This increased risk is observed already at a young age (1) and is mostly driven by the presence and severity of diabetic kidney disease (DKD) (2). Additionally, individuals with type 1 diabetes have a worse prognosis after a stroke (3).

Haptoglobin (Hp) is an abundant plasma protein that binds free hemoglobin released from lysed red blood cells. Thus, Hp both prevents iron loss and protects the cardiovascular system from oxidative heme-bound iron (4). A copy number variant (CNV), rs72294371, in the Hp gene on chromosome 16q22 affects the Hp protein multimerization; Hp circulates as a dimer in individuals homozygous for the minor Hp1 allele, whereas the major Hp2 allele enables the formation of trimers, tetramers, and even hexamers (5). The Hp-hemoglobin complex formed from the Hp2-2 genotype is cleared slowly from the bloodstream; thus, Hp2-2 has weaker antioxidative properties (5). Hp2-2 has also been linked to cholesterol transport defects (6) and HDL dysfunction (7). Most importantly, Hp2-2 is associated with an increased risk of cardiovascular disease (8,9), kidney disease (10), and mortality (11) in individuals with diabetes and in the general population (12). However, some studies have showed less convincing or contradictory results (13,14).

Regarding stroke, it remains inconclusive whether the Hp alleles are associated with stroke and which allele increases the risk. For example, a study of 124 lacunar individuals with stroke (i.e., type of cerebral small-vessel disease) and 918 control subjects found an association between the Hp1-1 genotype and stroke in the general population (15), but a study in 378 Finnish individuals with stroke and 1,426 population control subjects found no excess risk of stroke related to Hp1-1 (16). On the other hand, stroke survivors with the Hp2-1 or Hp2-2 genotypes had increased cardiovascular mortality (16). In the Finnish 2000 Health survey cohort, Hp2-2 was associated with the pooled cardiovascular disease phenotype, including stroke and transient ischemic attack (12). Furthermore, a meta-analysis of studies in individuals with diabetes suggested Hp2-2 as a stroke-risk genotype (17). In contrast, a recent study in 316 individuals with and without diabetes showed no association between Hp genotypes and lacunar infarctions (18). The only study conducted in individuals with type 1 diabetes included 607 subjects from the Epidemiology of Diabetes Complications (EDC) Study and suggested Hp1-1 as a risk factor for incident stroke (n = 33), but only in individuals with hypertension (19).

Here, we aimed to assess the relationship between the Hp genotypes and stroke, taking advantage of our well-characterized cohort of 194 incident individuals with stroke (2). Additionally, we imputed the Hp alleles to extend the analyses to a larger cohort of 4,345 individuals with type 1 diabetes, including 505 individuals with stroke.

All study participants are part of the prospective nationwide Finnish Diabetic Nephropathy (FinnDiane) Study, initiated in 1997. We analyzed two partially overlapping cohorts: a smaller Hp genotyping cohort and a larger Hp imputation cohort. At baseline, data collection occurred at a regular clinical visit that included, for example, anthropometric measurements and a thorough review of the individuals’ medical history. Additionally, blood and urine samples were collected. For the Hp1/2 imputation cohort, we identified strokes from the Finnish Care Register for Health Care and the Finnish Cause of Death Register (International Classification of Diseases codes in Supplementary Table 1). Additional information was gathered from death certificates, FinnDiane visits, and medical files, and all individuals with transient ischemic attack were excluded. For the Hp1/2 genotyping cohort, the DKD status was assessed based on albumin measurements from two of three timed overnight or 24-h urine collections. For the Hp1/2 imputation cohort, we retrieved data on kidney failure from the Finnish Care Register for Health Care until the end of 2017, and the data were further complemented with the data from FinnDiane baseline or prospective visit questionnaires or medical files. Data on coronary artery disease (CAD) were gathered from the Finnish Care Register for Health Care and the Finnish Cause of Death Register (Supplementary Table 1) or from the FinnDiane study visit questionnaire (Hp1/2 genotyping cohort). Mortality data by the end of 2017 were obtained from the Finnish Cause of Death Register. The Helsinki and Uusimaa Hospital District Ethics Committee approved the FinnDiane study protocol, which followed the Declaration of Helsinki. All study participants signed an informed consent.

To select individuals for the Hp1/2 genotyping, we identified 4,173 FinnDiane participants who had type 1 diabetes defined as diabetes diagnosis before the age 40 and insulin treatment initiated within a year from diagnosis, were free of a history of stroke at the baseline visit (between years 1997 and 2011), and had their DKD status defined (Fig. 1). We identified from the Finnish Care Register for Health Care and death certificates 194 individuals with an incident stroke during follow-up between years 1997 and 2013. All strokes were further verified from medical files and brain imaging by a stroke neurologist (J.P.). Altogether, 137 (71%) were ischemic and 57 (29%) were hemorrhagic strokes. These individuals had been part of a study described in detail previously (2). We then matched 540 individuals without stroke (control subjects) for age, sex, DKD status, and presence of CAD at baseline (Fig. 1). Supplementary Table 2 presents the characteristics of the matched control subjects compared with those not included in the current study. The group of individuals not included were younger, had lower blood pressure, fewer diabetes micro- and macrovascular complications, and included fewer men, when compared with the matched control subjects.

We required quality-controlled genome-wide genotyping data, genotyped with the Illumina HumanCoreExome chips and preprocessed at the University of Virginia (20), to be available for the individuals selected for the cohort for Hp1/2 imputation. In addition to the individuals with stroke available from the incident stroke cohort, we identified more individuals with stroke among those who had a stroke either before the FinnDiane baseline visit or until the year 2017 according to the Finnish Care Register for Health Care. We verified 69.9% of the individuals with stroke from medical files, 8.4% from the study visit questionnaires, and register data were used for the remaining 21.7%. The familial relationships in the cohort were calculated with KING (Kinship-based INference for Gwas) v.1.3 software. We included only one individual from the first-degree relatives, prioritizing the relative with a stroke, the parent from the parent-offspring pair, or randomly, one of the siblings. The control subjects were required to have diabetes duration >10 years, age >35 years, and no history of stroke. Individuals with traumatic brain hemorrhages reported in the FinnDiane visit questionnaires or medical files were excluded from the case subjects and control subjects. Altogether, the full phenotypic data were available for 505 individuals with stroke and 3,840 individuals without stroke (Fig. 1). Additionally, we analyzed four subgroups: individuals with 1) no CAD or kidney failure (n = 3,192, including 269 individuals with stroke), 2) kidney failure (n = 404, including 121 individuals with stroke), 3) CAD (n = 444, including 65 individuals with stroke), or 4) individuals with both CAD and kidney failure (n = 266, including 45 individuals with stroke).

The Hp CNV significantly affects the Hp gene length. The Hp1 allele contains five exons versus seven exons in the Hp2 allele. This difference is detectable with two PCRs. Thus, we genotyped the CNV with a PCR method similar to Ijäs et al. (12) (Supplementary Table 3). The PCR amplicons were analyzed on an agarose gel or with the Caliper LabChip GX instrument (PerkinElmer, Waltham, MA) at the Finnish Institute of Molecular Medicine to read the Hp1-1, Hp2-1, or Hp2-2 genotypes from the gel figures or Caliper results.

Boettger et al. (21) developed an imputation protocol for the Hp CNV. This imputation method can also separate the “S” and “F” alleles (Hp1S, Hp1F, Hp2FF, Hp2FS, and Hp2SS) that differ in a few amino acid residues and cause the Hp protein to run slower (S) or faster (F) in the gel electrophoresis (22). The imputation reference panel comprised 274 unrelated individuals of European origin from the 1000 Genomes Utah residents (CEPH) with Northern and Western European ancestry (CEU), Iberian populations in Spain (IBS), and Tuscans from Italy (TSI) populations. The panel included 1,277 variants (chr16:71,088,193–73,097,663; GRCh37/hg19) genotyped with the Illumina OMNI 2.5 single nucleotide polymorphism (SNP) array, and the variants in the Hp CNV region (chr16:72,090,310–72,093,744) had been replaced with Hp1/2 alleles determined with droplet-digital PCR. We renamed some variants with the corresponding rs number (e.g., SNP16–69,664,602 to rs144319423). Thereafter, we used PLINK v.1.9 to extract genotyped variants located at chr16:72,090,310–72,093,744, with Hardy-Weinberg equilibrium (HWE) P < 0.001 and >0.90 genotyping success rate for variants and individuals from our GWAS data. We converted the 146 variants overlapping with the imputation reference panel to Beagle format (.bgl) with PLINK and imputed the Hp1/2 alleles with Beagle 3.2 software similar to Boettger et al. (21) (nsamples = 15, niterations = 15, maxwindow = 2,000). The median imputation info score was >0.99 for all alleles Hp1S, Hp1F, Hp2FS, and Hp2SS, whereas the rare Hp2FF allele was absent. The Hp1/2 genotypes (Hp1-1 = Hp1S-Hp1S, Hp1S-Hp1F, or Hp1F-Hp1F; Hp2-2 = Hp2SS-Hp2SS, Hp2SS-Hp2FS, or Hp2FS-Hp2FS; Hp2-1 = Hp1S-Hp2SS, Hp1S-Hp2FS, Hp1F-Hp2SS, or Hp1F-Hp2FS) were coded for 500 of 505 individuals with stroke and 3,806 of 3,840 control subjects (total 99.1% success rate), with both Hp1/2 alleles imputed with ≥0.70 probability. With an expected Hp1 allele frequency of 38% in the Finnish population (16) and P = 0.05, we can detect an odds ratio of 1.54 between the case subjects and control subjects or a hazard ratio (HR) of 1.20 in a prospective analysis, with 80% power.

We aimed to conduct both time-to-event analyses and case-control analyses in our Hp genotyping and imputation cohorts, as presented in Fig. 2. We analyzed normally distributed continuous variables with the Student t test or ANOVA, and the nonnormally distributed continuous variables with the nonparametric Mann–Whitney U test or Kruskal–Wallis test. The difference in categorical variables between groups was tested with the χ² test. The association between Hp genotype and stroke was analyzed with χ² test, logistic regression, and with the Cox proportional hazards model. In the genotyped Hp cohort, the follow-up until stroke started at the individual’s baseline study visit and at the time of diabetes diagnosis for the imputation cohort. Analyses were performed with R versions 3.6–4.0, and a P value of <0.05 was considered statistically significant.

The data sets analyzed in the current study are not publicly available because the FinnDiane study protocol does not allow the sharing of individual-level data. The scripts used in the data analysis are available from the corresponding author upon reasonable request.

Individuals with or without stroke were matched for age, DKD stage, and CAD at baseline, but case subjects had higher systolic blood pressure (SBP), lower BMI, poorer glycemic control, and a higher proportion of current smokers compared with the control subjects (Supplementary Table 4). The genotyping success rate of the Hp polymorphism was 0.95; thus, 179 individuals with stroke (92%) and 517 control subjects (96%) remained in further analyses. The Hp1-1 genotype was observed in 14.9% (n = 104), Hp2-1 in 45.5% (n = 317), and Hp2-2 in 39.5% (n = 275) of the individuals, and the frequencies followed the HWE (P > 0.05). The Hp genotype frequencies did not differ between individuals with an incident stroke (Hp1-1: 13.4%, Hp2-1: 48.6%, and Hp2-2: 38.0%) and individuals without stroke (Hp1-1: 15.5%, Hp2-1: 44.5%, Hp2-2: 40.0%; χ² test P = 0.602), and neither did the allele frequencies (Hp1: 37.7% in both groups). The baseline clinical characteristics of individuals with different Hp genotypes were similar (Table 1).

The median follow-up time in the genotyped Hp cohort was 10.7 years (interquartile range [IQR] 5.5–13.4). Time-to-event analysis showed no difference between the Hp genotypes (number of Hp2-alleles, additive model) and the risk of stroke in an unadjusted analysis (HR 1.01 [95% CI 0.82, 1.25], P = 0.905) (Fig. 3A) or in a model with baseline diabetes duration, SBP, BMI, HbA_1c, and retinal photocoagulation as covariates (HR 0.98 [0.79, 1.22], P = 0.872). Furthermore, the Hp genotype was not a risk factor in separate time-to-event analyses for ischemic or hemorrhagic stroke, adjusted for the same covariates (ischemic stroke: HR 0.92 [0.72, 1.18], P = 0.517; all hemorrhagic strokes: HR 1.25 [0.82, 1.90], P = 0.298) or when analyzing lacunar infarcts or intracranial and subarachnoid hemorrhage separately (Supplementary Table 5). A fully adjusted model with baseline diabetes duration, SBP, BMI, HbA_1c, and retinal photocoagulation in addition to LDL cholesterol, triglycerides, estimated glomerular filtration rate (eGFR), hs-CRP concentration, and current smoking as covariates gave similar nonsignificant findings for the Hp1/2 genotype (data not shown).

During follow-up, 209 control subjects (40%) and 120 individuals with stroke (67%) died. In case subjects, the Hp genotype frequencies were similar in those who died (Hp1-1: 12.5%; Hp2-1: 45.8%, Hp2-2: 41.7%) and those who stayed alive (Hp1-1: 15.3%; Hp2-1: 54.2%, Hp2-2: 30.5%, P = 0.351). The median follow-up time after an incident stroke was 5.4 years (IQR 1.2–9.1). In the 35 individuals who died after a hemorrhagic stroke, the time to death was significantly shorter (median, 0.05 years [IQR 0.00–3.06]) compared with the 85 individuals who died after an ischemic stroke (3.7 years [1.2–7.1], P < 0.001). Altogether 66% (n = 23) of the deaths after a hemorrhagic stroke occurred within 30 days from the stroke, whereas the same proportion for the ischemic strokes was 21% (n = 18).

Survival analyses did not show Hp1/2 genotype (coded as a number of Hp2 alleles) as a risk factor for death after a hemorrhagic stroke (HR 1.30 [95% CI 0.79, 2.17], P = 0.303) or an ischemic stroke (HR 1.16 [0.84, 1.61], P = 0.368), or in a combined analysis of all strokes together (HR 1.24 [0.95, 1.63], P = 0.118) (Fig. 3B). Similarly, Hp genotype was not a risk factor for short-term (<30 days) or long-term (≥30 days) mortality after an ischemic or hemorrhagic stroke (data not shown).

Individuals with stroke were younger age at diagnosis of type 1 diabetes, had more kidney failure and CAD at the time of the stroke, and included fewer women compared with individuals without stroke (all P < 0.0001) (Supplementary Table 6). On the other hand, the case subjects were younger and had a shorter duration of diabetes at the time of stroke compared with the age and duration of the control subjects at the end of the follow-up (death or 31 December 2017; P < 0.003). We conducted the following analyses in 500 individuals with stroke and in 3,806 individuals without stroke with successfully imputed Hp1/2 genotypes.

In the imputation cohort, the frequency of the minor Hp1 allele was 36.1%, and the genotype frequencies followed the HWE (P > 0.05) both in individuals with stroke (Hp1-1: 13.8%, Hp2-1: 47.8%, and Hp2-2: 38.4%) and in control subjects (Hp1-1: 13.4%, Hp2-1: 45.0%, and Hp2-2: 41.6%).

Hp1/2 genotypes were successfully determined with both PCR genotyping and imputation in 609 individuals (Supplementary Fig. 1). In this group, the concordance between the imputed genotypes and the genotyped was high: 92.6% for Hp1-1, 96.7% for Hp2-1, and 97.5% for Hp2-2 (Supplementary Table 7).

The Hp1/2 genotype frequencies were similar in individuals with and without stroke (P = 0.237, sex- and age-adjusted) (Table 2). Additionally, the Hp1/2 genotype frequencies were similar in 263 individuals with an ischemic stroke, 83 people with hemorrhagic stroke, and in 149 individuals with unclassified stroke compared together or separately with the control group without stroke (all P > 0.05, sex- and age-adjusted logistic regression) (Table 2). The Hp2-2 genotype carriers had similar age at stroke in individuals with an ischemic stroke (Hp1-1: 52.3 years, Hp2-1: 51.5 years, and Hp2-2: 50.1 years; ANOVA P = 0.225), a hemorrhagic stroke (Hp1-1: 50.6 years, Hp2-1: 49.9 years, and Hp2-2: 48.0 years; P = 0.342), or among all case subjects, including the 149 nonclassified strokes as well (n = 500; Hp1-1: 51.4 years, Hp2-1 years: 51.0, and Hp2-2: 49.3 years; P = 0.084) (Table 3). Additionally, the Hp2 allele was not a significant risk factor for stroke in the Cox proportional hazards models, when analyzing all strokes, or ischemic or hemorrhagic strokes with follow-up starting at diabetes diagnosis and sex and age at diabetes diagnosis as covariates (e.g., risk of any stroke, increasing number of Hp2 alleles; HR 0.93 [95% CI 0.82, 1.06], P = 0.260).

Other diabetes complications were common among individuals with stroke: 48.2% had kidney failure (33.2% prior to stroke) and 44.0% had CAD (22.0% prior to stroke) (Supplementary Table 6). The Hp2-2 genotype frequencies were similar in individuals with kidney failure (Hp2-2: 43.1% vs. 40.9%) compared with individuals without kidney failure and in individuals with CAD (Hp2-2: 42.8% vs. 40.9%) compared with individuals without CAD (P > 0.05 in χ² test comparing the Hp genotype frequencies, or in logistic regression models adjusted for sex and age at the time of the event [case subjects] or at the end of the follow-up [control subjects]). When the subgroups of individuals stratified by the presence of CAD and kidney failure were analyzed, there was a trend of lower age at stroke in individuals with the Hp2-2 genotype (Table 3). This difference was significant within 121 individuals with kidney failure prior to stroke (Hp1-1: 51.8 years, Hp2-1: 49.3 years, Hp2-2: 46.6 years; P = 0.012). However, time-to-event analysis with follow-up starting at diabetes diagnosis did not support Hp2 as a stroke risk allele (data not shown). Furthermore, we did not observe any significant differences in Hp genotype frequencies between the individuals with stroke and control subjects in these four comorbidities subgroups (Table 3). For example, the 45 individuals who had both CAD and kidney failure prior to stroke had similar Hp1/2 frequencies compared with the 221 individuals who had CAD and kidney failure but no stroke (unadjusted P = 0.638; sex- and age-adjusted P = 0.551).

In the imputation cohort, the frequencies of different S and F Hp-alleles were Hp1S: 19.5%, Hp1F: 16.6%, Hp2FS: 60.7%, and Hp2SS: 1.3%. These frequencies were similar to those reported for the 119 individuals from the European CEU population (21): Hp1S: 22.7%, Hp1F: 13.9%, Hp2FS: 60.5%, and Hp2SS: 2.9% (χ² test P = 0.083). With respect to stroke, the exact Hp genotype frequencies in individuals with stroke and control subjects were similar within each Hp1/2 genotype group (P > 0.05 both unadjusted and sex- and age-adjusted) (Supplementary Table 8).

In the imputation cohort, the median follow-up time after a suffered stroke was 6.0 years (IQR 1.6–12.8), during which altogether 305 individuals (61%) died. Those who died had more often kidney failure (73.9% vs. 49.0%, P < 0.0001) and CAD (70.0% vs. 53.9%, P < 0.0001) compared with individuals who stayed alive. Of note, 68.9% of the kidney failures but only 48.2% of the CAD events had occurred prior to stroke. Dying from cardiovascular causes accounted for 76.1% of all deaths in individuals with stroke.

The Hp1/2 genotype frequencies were similar in individuals who died after stroke compared with individuals who stayed alive, and the number of Hp2-alleles was not a risk factor for death in a Cox proportional hazards model adjusted for sex and age at stroke (HR 1.03 [95% CI 0.87, 1.22], P = 0.722). Similarly, Hp genotype did not affect the survival after an ischemic stroke or a hemorrhagic stroke when analyzed separately (data not shown). Since prior comorbidities, especially kidney failure, affects the risk of death, we added the presence of kidney failure (prior to stroke) as a predictor variable to the Cox model. In this analysis, kidney failure almost tripled the risk of dying after stroke (HR 2.99 [2.35, 3.80], P < 0.0001), while the Hp1/2 genotype did not affect the risk of death (HR 0.97 [0.82, 1.15], P = 0.755). Additionally, prior CAD was a risk factor for dying after stroke (HR 1.46 [1.10, 1.94], P = 0.008) in a model with sex, age at stroke, and the Hp1/2 genotype, which was nonsignificant. Cardiovascular mortality was similar after any stroke, ischemic stroke, or hemorrhagic stroke in the different Hp1/2 genotype carriers.

Our study in Finnish individuals with type 1 diabetes found no association between Hp gene CNV and stroke. Both case-control and time-to-event analyses showed a similar risk of stroke and stroke subtypes in the carriers of Hp1-1, Hp2-1, and Hp2-2 genotypes. Additionally, survival after a stroke was not affected by the Hp1/2 genotype.

In type 1 diabetes, our study analyzing 179 individuals with stroke with genotyped Hp alleles and 500 individuals with stroke with imputed Hp genotypes is thus far the largest conducted. While some earlier studies in the general population and diabetes cohorts have shown increased stroke risk associated either with Hp2-2 or Hp1-1 genotypes, our results follow those studies showing no association between Hp CNV and stroke (16,18). Similar to our results, Costacou et al. (19) found no association between Hp1/2 genotypes and the risk of stroke in the EDC cohort. They detected, however, a higher incidence of stroke in individuals with the Hp1-1 genotype compared with the Hp2 allele carriers (Hp2-1 and Hp2-2) in a subgroup of individuals with more recent disease onset (diagnosis year ≥1965) and hypertension (<10 individuals with stroke) (19). We failed to replicate this finding in our cohort (258 individuals, of which 75 had a stroke, fulfilling these criteria in our cohort; data not shown).

We sought to stratify for comorbidities since the Hp2-2 genotype has been associated with CAD and kidney failure, while Hp1-1 is the previously suggested risk genotype for stroke incidence in type 1 diabetes (19). Nevertheless, the Hp genotype was not associated with stroke in individuals without comorbidities or in individuals with both CAD and kidney failure. We noticed, however, a significant trend toward younger age of stroke in the Hp2-2 genotype carriers in individuals with prior kidney failure. Time-to-event analyses in individuals with kidney failure showed, however, no excess risk of stroke in Hp2-2 genotype carriers.

Hp CNV might have several biological functions in stroke. An in vitro study suggested that Hp1-1 might be associated with poorer endothelial repair potential in individuals that suffered a lacunar stroke, the most common type of ischemic stroke (23). After a hemorrhagic stroke, the rapid clearance of free hemoglobin by Hp is essential for recovery (24). Therefore, possibly due to poorer free hemoglobin clearance capacity or some other vascular-related function, the Hp2-2 genotype seems to increase cardiovascular mortality after a stroke in the general population (16). However, in the type 1 diabetes context, we found no excess risk of death after a hemorrhagic stroke (or any stroke) related to the Hp2-2 genotype. As a limitation, however, we had only limited number of individuals with hemorrhagic stroke. Further, cardiovascular mortality after stroke was similar regardless of the Hp1/2 genotype.

In addition to having poorer antioxidant properties, the Hp2 allele has been associated with higher cholesterol concentrations. In our study, both total cholesterol and LDL cholesterol were the highest in the individuals with the Hp2-2 genotype. This nonsignificant trend was seen in both stroke case subjects and control subjects and in individuals with or without lipid-lowering medication. Similarly, in the Diabetes Control and Complication Trial (DCCT) type 1 diabetes cohort, Hp2-2 carriers showed a nonsignificant trend for higher LDL cholesterol and total cholesterol concentrations (14). As not being the main phenotype of interest of our study, we analyzed the lipid variables only in our smaller cohort with genotyped Hp alleles (n = 696) and comprehensive baseline phenotypic data available. Therefore, further studies in a larger cohort are needed to confirm the effect of Hp CNV on lipid concentrations in type 1 diabetes.

More interestingly, the same study (14) in the DCCT cohort found an association between the Hp2-2 genotype and CAD only in individuals belonging to the secondary cohort (diabetes duration 1–15 years at recruitment and early diabetic retinopathy) and intensive glucose control treatment arm, while their whole DCCT cohort analysis showed no excess CAD risk for Hp2-2 genotype carriers. Similarly, Hp2-2 was not a risk factor for combined CAD and stroke end points in individuals with elevated HbA_1c (13). Our study found no association between Hp1/2 genotype and CAD, but other studies have shown Hp2-2 as a CAD risk genotype in type 1 diabetes (8,9) or in individuals with elevated HbA_1c (25).

Similarly, individuals with kidney failure had similar Hp1/2 frequencies in our cohort. An earlier study in type 1 diabetes showed Hp2-2 to be a risk factor for kidney failure and kidney function decline (10). We could also see such a trend in our genotyped Hp1/2-alleles cohort, where the Hp2-2 genotype carriers tended to have a lower eGFR (P = 0.097), a measure of kidney function.

Since no single SNP correlates well with the Hp exon CNV (highest r² = 0.44), the European SNP reference panel for the Hp1/2 genotype imputation (21) is a great resource that enables Hp imputation in large cohorts with genome-wide SNP data. In our study, the imputation of Hp alleles performed well. The imputation method provides additional subdivision to S and F Hp1 and Hp2 alleles, which in our study did not associate with the risk of stroke, as the genotype frequencies were similar in case subjects and control subjects. Our study is among the first studies (21,26) to analyze the imputed S and F alleles.

Hp1/2 CNV primarily affects the Hp protein multimerization and function but is also modestly associated with Hp concentration in the blood—the Hp2 allele associated with lower Hp levels. Our main hypothesis was, however, that the structural effects of the Hp1/2 CNV are stronger in stroke susceptibility than the effect on the Hp concentration, which is more strongly regulated by other genetic variants (27–29).

An asymptomatic cerebral small-vessel disease characterized with, for example, cerebral microbleeds, is common in individuals with type 1 diabetes (30). As a limitation, we did not study these silent brain manifestations as part of this study. Therefore, some of our nonstroke control subjects likely had asymptomatic brain manifestations. To conclude, previous studies on Hp1/2 genotype and stroke were contradictory and inconclusive. Our study suggests that Hp CNV is not associated with the risk of stroke in type 1 diabetes or with stroke subtypes or survival after stroke.

Acknowledgments. We are indebted to the late Dr. Carol Forsblom (1964–2022), the international coordinator of the FinnDiane Study Group, for his considerable contribution. The authors thank all FinnDiane study participants and skilled nurses and clinicians at the FinnDiane study sites (Supplementary Table 9). We also thank Maija Parkkonen (Folkhälsan Research Center) for the great work in optimizing the Hp genotyping.

Funding. This study was supported by the Folkhälsan Research Foundation, Wilhelm and Else Stockmann Foundation, Liv och Hälsa Society, Helsinki University Hospital Research Funds (TYH2018207), Novo Nordisk Foundation (NNF OC0013659), Sigrid Jusélius Foundation, the Academy of Finland (299200 and 316664), and the Aarne Koskelo Foundation.

Duality of Interest. M.I.E. is a shareholder of BCB Medical Oy. P.-H.G. has received investigator-initiated research grants from Eli Lilly and Roche, is an advisory board member for AbbVie, Astellas, AstraZeneca, Bayer, Boehringer Ingelheim, Cebix, Eli Lilly, Janssen, Medscape, Merck Sharp & Dohme, Mundipharma, Nestlé, Novartis, Novo Nordisk, and Sanofi, and has received lecture fees from Astellas, AstraZeneca, Boehringer Ingelheim, Eli Lilly, ELO Water, Genzyme, Medscape, Merck Sharp & Dohme, Mundipharma, Novartis, Novo Nordisk, PeerVoice, Sanofi, and SCIARC. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. A.S. conducted the laboratory investigation. A.S. analyzed the data. A.S. and E.H.D. performed the imputation. A.S., S.H.-H., C.F., M.I.E., V.H., J.P., and L.M.T. curated the data. A.S. and L.M.T. wrote the original draft of the manuscript. C.F., P.-H.G., N.S., and L.M.T. conceptualized the project. All authors critically revised and edited the manuscript. P.-H.G. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.