Association of HbA1c with VO2max in Individuals with Type 1 Diabetes: A Systematic Review and Meta-Analysis

The aim of this systematic review and meta-analysis was to evaluate the association between glycemic control (HbA1c) and functional capacity (VO2max) in individuals with type 1 diabetes (T1DM). A systematic literature search was conducted in EMBASE, PubMed, Cochrane Central Register of Controlled Trials, and ISI Web of Knowledge for publications from January 1950 until July 2020. Randomized and observational controlled trials with a minimum number of three participants were included if cardio-pulmonary exercise tests to determine VO2max and HbA1c measurement has been performed. Pooled mean values were estimated for VO2max and HbA1c and weighted Pearson correlation and meta-regression were performed to assess the association between these parameters. We included 187 studies with a total of 3278 individuals with T1DM. The pooled mean HbA1c value was 8.1% (95%CI; 7.9–8.3%), and relative VO2max was 38.5 mL/min/kg (37.3–39.6). The pooled mean VO2max was significantly lower (36.9 vs. 40.7, p = 0.001) in studies reporting a mean HbA1c > 7.5% compared to studies with a mean HbA1c ≤ 7.5%. Weighted Pearson correlation coefficient was r = −0.19 (p < 0.001) between VO2max and HbA1c. Meta-regression adjusted for age and sex showed a significant decrease of −0.94 mL/min/kg in VO2max per HbA1c increase of 1% (p = 0.024). In conclusion, we were able to determine a statistically significant correlation between HbA1c and VO2max in individuals with T1DM. However, as the correlation was only weak, the association of HbA1c and VO2max might not be of clinical relevance in individuals with T1DM.


Introduction
Regular physical activity (PA) represents a highly relevant non-pharmaceutical-glucose lowering activity in people with type 1 diabetes (T1DM) and is recommended to be conducted by means of moderate-to-vigorous intensity aerobic exercise for 150 min at least three days a week in most adults with T1DM [1]. Despite the challenge of PA potentially contributing to exercise-induced dysglycemia in T1DM, potential benefits have been demonstrated in regard to improving cardio-respiratory fitness, cardiovascular risk factors, and quality of life [2]. Besides the impact of PA on these aforementioned parameters, deteriorated glycemic control may act as an antagonist, which in T1DM remains a matter of debate [3] In people with type 2 diabetes (T2DM), research evidence suggests an inverse correlation between cardiorespiratory fitness assessed by maximum oxygen uptake (VO 2max ) and clinical markers of glycemic control (high HbA 1c , high fasting glucose) also after adjustment for age, body weight and markers of adiposity [4]. It has been speculated that VO 2max as a determinant for PA may be positively associated with β-cell compensation mechanisms such as improvements in glucolipotoxicity [5]. This may lead to lower levels of pro-inflammatory cytokines and increased secretion of various growth factors and hormones which may contribute to higher β-cell mass [5]. However, these suggested mechanisms in T2DM are hardly transferable to a T1DM population that often has completely lost their endogenous insulin production during the progress of the autoimmune disease.
We have recently shown that individuals with early-stage T1DM compared to healthy controls are characterized by a lower maximum oxygen uptake including decreased absolute and relative oxygen uptake (VO 2 ) as well as weaker oxygen reserves and oxygen pulse when determined during a cardiopulmonary exercise (CPX) test on a cycle ergometer [6]. Interestingly, these altered responses were not associated with HbA 1c but this might have been due to the tight glycemic control in the T1DM cohort of this study (mean HbA 1c 6.9% [6.2;7.7%]). Thus, the jury is still out whether or not glucose control assessed by HbA 1c is associated with maximum oxygen uptake.
In clinical practice, this question is highly relevant for physically active patients and in particular for competitive athletes with T1DM. Therefore, this systematic review investigates the association between HbA 1c levels and VO 2max assessed during CPX testing in individuals with T1DM.

Materials and Methods
This systematic review and comprehensive analysis were conducted according to the Preferred Reporting Items for Systematic Reviews (PRISMA) guidelines. The study was registered at the International Prospective Register of Systematic Reviews (Prospero) prior to initiation of the literature search (CRD42020141164).

Data Sources and Study Selection
The following electronic libraries were searched to identify relevant publications: EM-BASE, PubMed, Cochrane Central Register of Controlled Trials, and ISI Web of Knowledge. Studies from January 1950 until July 2020 were included in the analysis. Search items were included as shown in Supplemental Tables S1-S4. Randomized and observational studies with a minimum number of three participants were included if HbA 1c measurements and CPX tests to determine VO 2max had been performed. Only published studies were considered. Moreover, no in silico or animal studies were included. Moreover, systematic reviews and meta-analysis were excluded. Additionally, duplicates of articles were discarded. Study titles and abstracts were reviewed including relevant studies fulfilling the inclusion criteria. Then, the full text of these studies was digitally saved and read by two independent authors (MLE, F.J.R.D.). Another independent author (O.M) monitored the identified studies and solved potential disagreements. The SD values were converted to standard error (SE = SD/ √ n) [7]. Interquartile range (IQR) or confidence intervals were converted using the formula provided in the supplementary file. If studies included more than one appropriate data set, these data were extracted and analyzed separately.

Criteria for Inclusion in the Review
The following criteria had to be met for the study abstract to be considered eligible for manuscript data extraction: (a) reporting of an association between HbA 1c at baseline and VO 2max , (b) at least three participating children aged 6-12, adolescents aged 13 to 17 or adults older than 18 years, (c) participants with type 1 diabetes, (d) observational, cross-sectional or randomized controlled study design, (e) conducted a CPX test. Studies without clear specification of diabetes type were not included.

Data Synthesis and Analysis
A narrative descriptive analysis was performed to summarize the characteristics of studies, such as population, age and type of CPX testing. VO2max was defined as the maximum oxygen consumption given in each study measured via a CPX test on a spirometric device independent of manufacturing company. HbA1c values were recorded as given in the anthropometry section of the included cohorts within each manuscript. Studies were excluded if VO2max values were not measured according to the guidelines of the American College of Sports Medicine [10,11].

Meta-Analysis
The meta-analysis was performed using the random effects model and Hedges' g method as a number of studies only had a small sample size. The effect size (delta BG) was summarized and presented as the pooled mean with corresponding 95% confidence interval (CI). The negative pooled mean indicated a higher decrease in BG following physical exercise. The heterogeneity in the effect size was assessed by estimating I2 statistics and Cochran's Q test for homogeneity. The difference in effect size with respect to cycling versus running studies and other study-level categorical covariates was assessed by performing sub-group analysis of the effect size for each covariate and group differences in the effect size were assessed via Cochran's Q test for homogeneity. The difference in effect size with respect to the study-level continuous covariates was assessed by meta-regression analysis. Furthermore, simple and multiple meta-regression were performed to assess the crude and adjusted association of each study-level covariate with the effect size within the strata of cycling and treadmill studies. The results of the meta-regression were

Data Synthesis and Analysis
A narrative descriptive analysis was performed to summarize the characteristics of studies, such as population, age and type of CPX testing. VO 2max was defined as the maximum oxygen consumption given in each study measured via a CPX test on a spirometric device independent of manufacturing company. HbA 1c values were recorded as given in the anthropometry section of the included cohorts within each manuscript. Studies were excluded if VO 2max values were not measured according to the guidelines of the American College of Sports Medicine [10,11].

Meta-Analysis
The meta-analysis was performed using the random effects model and Hedges' g method as a number of studies only had a small sample size. The effect size (delta BG) was summarized and presented as the pooled mean with corresponding 95% confidence interval (CI). The negative pooled mean indicated a higher decrease in BG following physical exercise. The heterogeneity in the effect size was assessed by estimating I2 statistics and Cochran's Q test for homogeneity. The difference in effect size with respect to cycling versus running studies and other study-level categorical covariates was assessed by performing sub-group analysis of the effect size for each covariate and group differences in the effect size were assessed via Cochran's Q test for homogeneity. The difference in effect size with respect to the study-level continuous covariates was assessed by meta-regression analysis. Furthermore, simple and multiple meta-regression were performed to assess the crude and adjusted association of each study-level covariate with the effect size within the strata of cycling and treadmill studies. The results of the meta-regression were reported as a coefficient with corresponding 95% CI and p values. Publication bias was assessed in terms of meta-bias using Egger's test and visualized via a funnel plot. The results of the meta-analysis are presented in the supplementary material.

Results
A total of 122 studies were extracted from the initially screened 4416 titles. The steps of the article selection process are described as a flow diagram in Figure 2. Studies published between 1989 and 2020 included data from 3278 participants with T1DM. Due to the crossover designs of the studies the results were split for analysis and led to a total number of 187 study results included. We included 57 case-control studies, 41 randomized controlled trials, 2 secondary outcome analyses from randomized controlled trials, 7 non-randomized comparative studies and 15 cohort studies. Groups were separated by adolescents, adolescents/adults and adults, since some study cohorts did not include both age-groups without a clear separation within the results of the study. Detailed information is shown in Figure 2.  All study-specific outcomes included in this systematic review and meta-analysis are shown in Table 1. Types of exercise identified as "other" were conducted outside on a track with a spirometric wireless device or if the type of exercise during CPX was not specified.
There was a weak, but significant correlation between VO 2max and HbA 1c of r = −0.19 (p < 0.001). The meta-regression revealed a slope between VO 2max and HbA 1c of −1.46 [−2.35-−0.58], p < 0.001), i.e., for each increase in HbA 1c of 1% the relative VO 2max was lower by 1.46 mL/kg/min. This association is shown in Figure 3. There was a weak, but significant correlation between VO2max and HbA1c of r = −0.19 (p < 0.001). The meta-regression revealed a slope between VO2max and HbA1c of −1.46 [−2.35-−0.58], p < 0.001), i.e., for each increase in HbA1c of 1% the relative VO2max was lower by 1.46 mL/kg/min. This association is shown in Figure 3.

Multivariate Meta-Regression
Multivariate meta-regressions for VO 2max in association with HbA 1c , sex, age group and type of exercise were conducted. The results are shown in Table 2. Univariate Metaregression can be found in the supplementary material.

Discussion
Our systematic review and meta-analysis showed that HbA 1c and VO 2max are inversely, albeit weakly associated. Furthermore, if groups were separated for HbA 1c , significant differences in VO 2max between low (<7.5%) and high (≥7.5%) values were found. Additional subgroup analyses showed that VO 2max was highest in male (in comparison to female) participants whereas age only had a small, insignificant effect on VO 2max and on the impact of HbA 1c on VO 2max .
Physical exercise has previously shown to not necessarily decrease HbA 1c for a variety of reasons [134,135]. Individuals with T1DM are advised to reduce insulin doses in preparation for specific exercise sessions, which may elevate BG and eventually (at least with regular physical exercise) HbA 1c [136]. In addition, patients are advised to supplement CHO with falling glucose values during exercise [136]. Even though the amounts of administered CHO are small, they may weaken the positive exercise induced effects on body mass and glycemic control.
Moreover, previous exercise studies have shown that slightly elevated BG levels or CHO supplementation during physical exercise may increase the performance of individuals without T1DM [136]. In combination with elevated pre-exercise BG values (due to insulin reductions) and higher post-exercise BG levels (in response to supplemented CHO during exercise), HbA 1c might be negatively impacted in those people with T1DM who are physically active, and fear of hypoglycemia might be another contributor.
However, a large multi-center study demonstrated that physical activity has a positive impact on glycemic control but also on diabetes-related comorbidities [137]. This is supported by the study from King et al. that found a positive association between glycemic control and the amount of physical activity in children with T1DM [138]. These studies are reflective of the results from our systematic review and meta-analysis that show an inverse relationship between HbA 1c and VO 2max , highlighting that a 1% increase in HbA 1c decreases VO 2max by 1.46 mL/kg/min. This may appear negligible at first, but Laukkanen et al. have shown that a decrease of 1 mL/kg/min of VO 2max is associated with a 9% increase in all-cause mortality in a population-based study of 579 men without diabetes [139]. VO 2max has previously been suggested to predict longevity. While still a matter of research, an elevated VO 2max allows individuals to be more active which may increase well-being and longevity [140]. It should be noted that HbA 1c worsens between the ages of 8-18, while from the age of 16 it steadily improves due to a higher awareness of diabetes management. In addition, puberty and hormonal changes at that time contribute to a more complicated glycemic management that may ease over time [141]. Adolescents in our study were already in the 'steadily improving' age range hence the impact of puberty and hormonal changes are not as pronounced or influential on our study results. Furthermore, adolescents with T1DM do not necessarily have a smaller VO 2max compared to adults since studies in healthy individuals [142] and individuals with T1DM [143] align with our overall findings regarding VO 2max .
Even though the association was weak in our results, HbA 1c has shown to reduce skeletal muscle mitochondrial ATP production, which consequently reduces performance, deteriorating VO 2max [144]. In individuals with an increased HbA 1c , capillary density around skeletal musculature has also been shown to be decreased [145]. This may in addition lead to compromised oxygen supply systems influencing an individual's functional capacity.
These studies and our systematic review and meta-analysis are not without limitations. Although still considered the gold standard of glycemic control, HbA 1c may not be the ideal parameter to detect the effects of physical exercise on glucose control in individuals with T1DM as it is an average of low and high BG values over a rather long period of time. Low BG values will reduce HbA 1c although they are not exactly an indicator of good glycemic control [146]. On the other hand, exercise-induced hyperglycemic BG values which are often accepted during physical activity will increase HbA 1c which might cast doubt on the benefit of exercise in people with T1DM. Another contributor to the effects seen in our analysis could be diabetes duration. This may have an impact on HbA 1c but also on VO 2max since it may decrease over time with insufficient training. However, it might not necessarily have an impact since diabetes duration could be influential once individuals with T1DM have gained more knowledge about their own diabetes which could thus improve their overall glycemic control and engagement in physical activities increasing VO 2max . Unfortunately, this could not be included in our systematic review and meta-analysis, since the majority of the studies did not show the diabetes duration in detail, omitting this important detail which should be considered in future studies.
Future studies should focus on more differentiated parameters than HbA 1c such as time in range which represents the standard parameter to assess quality of glycemic control in users of continuous glucose monitoring systems. Hypoglycemic and hyperglycemic glucose values do not compensate for each other in time in range. Therefore, time in range might be a more appropriate parameter to evaluate quality of glycemic control in particular for interventions potentially triggering hypoglycemia, including physical activity. It must be mentioned, that several confounding parameters may have an impact on HbA 1c without influencing VO 2max . This could be an adaption in dietary behavior, e.g., meal timing or overall carbohydrate intake, or this could be a change in therapy from pen to pump therapy or an upgrade to a novel insulin analogue that may lower HbA 1c immediately. Future studies on this important research topic will hopefully investigate additional parameters potentially affecting HbA 1c and its association with VO 2max , such as body mass, comorbidities and socioeconomic status-parameters that could not be included in this meta-analysis since as they were either not measured or reported insufficiently, leaving important information about the association of HbA 1c and VO 2max unexplored.
In our systematic review and meta-analysis, we chose the gold standard parameters for glycemic control and maximum oxygen uptake, HbA 1c and VO 2max , since they are the most popular and most often applied parameters in diabetes research, but VO 2max has its limitations, too. It is unable to fully represent the entire spectrum of physical performance since cardiological (heart rate max), metabolic (lactate max) and physiological parameters such as power and speed are not reflected. Since we solely included studies with individuals that conducted exercise tests until volitional exhaustion for the determination of VO 2max , results from our study cannot be transferred to all types of physical exercise.
Nevertheless, future studies are needed to investigate additional parameters that may impact long-term glycemic control and its association with VO 2max in individuals with T1DM.

Conclusions
This meta-analysis demonstrates an inverse association between physical performance and HbA 1c showing an increase in VO 2max with decreasing HbA 1c . Further studies relating to time in range will be needed to confirm a positive impact of glucose control on physical performance.  Table S1: Pubmed search, July 2020; Table S2: Embase search, July 2020; Table S3: Web of science search, July 2020; Table S4: Cochrane search, July 2020; Table S5: Statistical analysis of HbA1c of all included groups; Figure S1: Funnel Plot title.