Does Carbohydrate Counting from Diabetes Onset Make a Difference to Glycemic Outcomes? A Swedish Registry Study

Abstract

Introduction

Carbohydrate counting (CC) for calculating insulin for meals is recommended in type 1 diabetes to optimize the postprandial glycemic response. Few studies have studied the effect of CC on glycemic outcomes when introduced from diabetes diagnosis. Our aim was to investigate if CC effects metabolic outcomes over time between individuals taught CC from diabetes onset and those who were not. A secondary aim was to examine if use of CC was associated with glycemic outcomes at 8 years diabetes duration.

Methods

This was a cohort study analyzing pediatric data from the Swedish Diabetes Registry (NDR), comparing two groups with diabetes onset 2011–2013, where one was taught CC from diagnosis and the other group used fixed doses. HbA1c data were analyzed from first year of diabetes up to 8 years. A cross-sectional analysis at 8 years diabetes duration, compared groups with reported use of CC to the NDR to examine the association with glycemic outcomes.

Results

There was no difference in HbA1c over time between the group taught CC from diagnosis (n = 253) and those that were not (n = 221) after adjusting for gender, age, and insulin regimen (difference fixed CC: 1.0; 95% confidence interval: −0.8 to 2.9; p = 0.27). The group using CC after 8 years duration had significantly lower HbA1c than those not using CC (56.2 mmol/L vs. 64.8 mmol/L, p = <0.001).

Conclusion

The results suggest that CC can be effective in achieving target glycemic outcomes over time and support its introduction at any time in the life course to improve metabolic outcomes.

Introduction

Type 1 diabetes (T1D) is a chronic disease characterized by loss of endogenous insulin secretion, resulting in burdensome insulin injections and attention to diet as well as constant monitoring of glucose levels [1]. Carbohydrate counting (CC) is an important part of nutritional management in T1D as carbohydrates need to be covered by the correct amount of insulin to prevent dysglycemia [1–3]. It is based on the knowledge of the carbohydrate content of each meal consumed, divided by an individually set insulin to carbohydrate ratio (ICR), and it is the recommended method for insulin dosing for meals to optimize postprandial glycemic response [1, 4]. CC, as part of diabetes management, has been shown across nine studies in a recent systematic review to improve glycemic outcomes by 5.7 mmol/mol (0.52%) and is often cited as the preferred method to dose insulin for meals [2].

However, ascribing positive outcomes to one specific part of diabetes self-management such as dietary management is challenging as diabetes care is complex involving numerous daily tasks and psychosocial factors with access to diabetes technology and health care affecting glycemic outcomes [5, 6]. Despite improvements in diabetes technology, there have been only small improvements in the proportion of children meeting glycemic targets, suggesting factors such as dietary behaviours and interventions play a key role [6–8]. A study comparing registry data with a lifespan perspective, from Sweden, Germany/Austria, and USA, found the only improvement in HbA1c over all age groups was seen in the Swedish data with uniform access to healthcare, diabetes technology and where CC is recommended from diabetes diagnosis across all pediatric diabetes clinics [7, 9]. Data from the Norwegian diabetes registry show improved glycemic outcomes over the last decade, attesting the improvements to diabetes technology, use of CC and systematic quality improvements in their pediatric departments [10].

Achieving glycemic targets is important as it is a predictor of present and future health, and it has been shown that the first years of glycemic control impacts HbA1c and diabetes related complications later in life [11, 12]. It is therefore important to understand if methods for insulin dosing, such as CC is better taught at onset of diabetes, and if this impacts glycemic outcomes. There is, to our knowledge, only one study examining glycemic outcomes over time when CC was introduced from diabetes onset [13]. Results from that study showed that CC, as part of their intensive management, increased the proportion of patients reaching target glycemic outcomes <58 mmol/mol.

Objectives

This study aimed to investigate the effect of CC on glycemic outcomes and specifically examine if learning it at onset impacts HbA1c. The hypothesis was that introduction of CC at diabetes onset would improve HbA1c at onset and over time, when compared to those who were taught fixed doses. A secondary aim was to examine if use of CC was associated with improved glycemic outcomes at 8 years of diabetes duration.

Methods

Study Design

This was a longitudinal cohort study based on the Swedish National Diabetes Registry (NDR), to investigate metabolic outcomes over 8 years of diabetes duration between two groups, those who were taught CC at onset and those who were taught fixed doses.

As many of the participants in the fixed group were taught CC in the years following diabetes onset, a cross-sectional examination of participants using CC after 8 years of diabetes duration was performed, irrespective of the method taught at onset. Metabolic outcomes, such as HbA1c, mean blood glucose (BG), time in range (TIR), glucose variability, and BMI were investigated. The study was approved by the Swedish Ethical Review Authority (2022/01/11 Dnr 2021-04903).

Study Population

The Astrid Lindgren Children’s Hospital, part of the Karolinska University Hospital in Stockholm, Sweden, is a pediatric hospital with two multidisciplinary diabetes units providing care to 1,300 children and adolescents 0–18 years with T1D. Annually 100–150 children are diagnosed with T1D across the two sites. On January 1, 2012, advanced CC in gram increments and with individual ICRs was implemented as the standard nutritional intervention for all newly diagnosed children at both units, as compared to fixed doses that had been used prior for insulin dosing for meals. Training in CC was initiated directly when patients started meal bolusing on the ward, instructed, and supported by the ward nurses as well as the diabetes team. Continued education proceeded the following 1–3 weeks with the multiprofessional diabetes team, including at least four visits with the dietician and individual adaption. The clinical protocol is to offer four visits per year, interchangeably to endocrinologist and diabetes nurse where assessment of glucose outcomes, insulin pump settings, as well as CC skills was assessed. Referrals to the dietitian was prompted if needed and encouraged yearly but not mandatory. With a team approach, implementation of CC for all patients was carried out with training individually or in group when starting insulin pump therapy (IPT). When CGMs became available in 2016, this was included as clinical practice and commenced in the first weeks after diabetes diagnosis and offered to all existing patients.

Data Collection and Measurements

To answer the primary aim, all children and adolescents diagnosed with T1D at Astrid Lindgren Children’s Hospital between 2010 and 2013 were eligible for inclusion. The inclusion date was defined as date of T1D diagnosis. Participants with less than 4 years of follow-up data were excluded from the analysis, as well as one participant with a condition impacting HbA1c, excluding 6 participants. This resulted in the inclusion of 474 participants.

Based on the date of diagnosis, we defined two study groups: fixed dose group (n = 221), children diagnosed 2010–2011, who were managed with fixed insulin doses from onset; and CC group (n = 253), children diagnosed 2012–2013, who received education on CC from onset. Each participant was followed based on diabetes duration, where 0 represents the first year after diagnosis, 1 represents the second year, and so on.

To answer the second aim, reported use of CC to the NDR (yes/no) at 8 years diabetes duration was used to create two separate groups, irrespective of method taught at onset of diabetes; those who used CC and those who did not use CC. A total of 341 (72%) participants had a reported variable for use of CC after 8 years of diabetes duration, n = 250 who used CC and n = 91 who did not use CC.

Standardized data are reported by the diabetes teams to the NDR and NDR kids, following each visit to an endocrinologist or nurse/diabetes educator. The frequency of visits varies across clinics and between patients depending on individualized need and, consequently, the reporting of variables to the NDR. Pediatric clinics typically conduct 3–4 visits per year, whereas adult clinics average 1–2 visits annually.

Variables reported include glycemic outcomes (HbA1c), growth (height, weight), type of treatment (MDI or IPT), and insulin requirements (U/kg, TDD). HbA1c is measured in clinic using a point-of-care analyzer, from capillary blood test, DCA 2000 (Siemens Healthcare AB). In the years between 2010 and 2020, there was a continuous update in variables reported to the NDR as new variables came of interest, i.e., TIR (3.9–10 mmol/L), time in tight range (TITR 3.9–7.8 mmol/L), and time below range (TBR <4 mmol/L) as well as if patients use CC (yes/no).

Data were acquired from NDR Kids and NDR in April 2022. Because date of diabetes onset ranged from January 2010 and December 2013, diabetes duration varied in the extracted dataset. To achieve uniform follow up periods for both groups, we have analyzed data between the years 2010–2020 for the fixed group and 2012–2022 for the CC group, which equals 8 years of diabetes duration for each participant (2 years difference in diabetes onset).

The primary outcome was HbA1c, calculated as means for each participant measurements for each year with diabetes, which means that each participant contributed a maximum of one measurement per year. This approach reduced variability due to multiple measurements per year, providing a standardized and comparable measure of HbA1c for each participant. Categorical variables, such as insulin regimen and use of CC, were calculated as medians.

Secondary outcomes were mean BG, TIR, TITR, variability (SD), and BMI. BMI was translated into categories (normal weight or overweight) using BMI SDS values for participants <18.0 years of age and BMI for participants ≥18 years of age, to enable comparisons between adults and children [14]. We categorized age groups (0–5, 6–12, 13–17, 18–23, and ≥24 years) based on classifications from the annual reports of the NDR. All baseline variables were calculated as means for the first year, with the exclusion of all measurements for the first 3 months after diagnosis to exclude high HbA1c measurements related to diagnosis and allow for weight regain.

Statistical Analyses

Primary Aim

We report baseline characteristics as means with standard deviations (SDs) for continuous variables and percentages for categorical variables. We compared groups using independent t-test or Mann-Whitney U test for continuous variables, using the Shapiro-Wilk test to assess normality. We used chi-square test to compare proportions.

To compare how CC affects HbA1c levels over time between participants who learned CC from diagnosis (CC group) and those who did not, we used linear mixed-effects modeling with an autoregressive AR (1) correlation structure. Outcome variable was HbA1c. Participants were treated as random effects to account for individual differences. Using mixed models allows for the analysis of repeated measurements over time while accounting for individual variability, leading to more accurate and robust estimates of the comparisons of HbA1c [15].

Our statistical strategy involved three models: first, we estimated the crude mean difference in HbA1c between the fixed and CC group, using group and diabetes duration as factors (model 1); second, we adjusted for gender, age category (defined above), and insulin regimen by adding these factors (model 2). Finally, we tested whether the difference in HbA1c between the groups changed over time, by adding an interaction between group and duration (model 3).

Secondary Aim

A cross-sectional analysis was undertaken to examine if there was a difference in mean HbA1c 8 years after diabetes diagnosis, between those who had a reported variable to NDR of current use of CC (yes or no). We used a general linear model (UNIANOVA), with HbA1c as outcome variable. First, we estimated the crude difference in mean HbA1c and 95% confidence intervals (CIs) between the two groups. Then, we adjusted the model adding covariates that could affect HbA1c and differed between the groups: gender, insulin regimen, BMI categories, and age.

Two-sided p values <0.05 were considered statistically significant. Analyses were conducted using IBM SPSS STATISTICS 29.0 for MAC (2024).

Results

Baseline comparisons between the fixed and CC groups are presented in Table 1. There were small differences that did not reach statistical significance.

Table 1.

Comparison of characteristics at the first year of diabetes between fixed and CC group

End first year with diabetes	Fixed (n = 221)	CC (n = 253)	p value for differences between fixed and CC
# Visits to clinic	498	597	0.426
Age at onset	8.8 (4.7)	9.5 (4.5)	0.128
Age category, n (%)
0–6 years	70 (31.7)	67 (26.5)	0.321
7–13 years	108 (48.9)	125 (49.4)
14–18 years	43 (19.5)	61 (24.1)
Gender, n (%)
Female	105 (47.5)	115 (45.5)	0.654
Male	116 (52.5)	138 (54.5)
HbA1c
mmol/mol (IFCC)	51.4 (10.7)	49.9 (11.0)	0.151
% (DCCT)	6.9 (3.1)	6.7 (3.2)
BMI SDS, n = 369 (<5 years excluded), n (%)
Under weight	3 (1.8)	4 (2.0)	0.803
Normal weight	121 (71.6)	149 (74.5)
Overweight	38 (22.5)	37 (18.5)
Obesity	7 (4.1)	10 (5.0)

The number of visits per patient ranged from 0 to 12 per year, with a total of 7–85 visits per patient over the 8-year follow-up. The total number of visits over 8 years were 4,832 for the fixed group and 5,491 for the CC group. Mean number of visits per participant and group over the 8 years was 22.0 for the fixed and 21.6 for the CC group.

Comparison of HbA1C between Fixed and CC over Time

Over 8 years, we analyzed 474 participants and 4,042 mean HbA1c measurements using a generalized linear mixed model. There was no difference in HbA1c between the fixed dose group (adjusted mean 59.0 mmol/mol) and the CC group (adjusted mean 57.9 mmol/mol) after adjusting for gender, age, and insulin regimen (difference fixed CC: 1.0; 95% CI: −0.8 to 2.9; p = 0.27), Table 2, model 2. HbA1c increased in both groups over the study period. When analyzing interaction effect between groups and time, there was a difference in HbA1c between groups year 5 and 7 (Table 2, model 3; Fig. 1a). To add context to this difference, Figure 1a shows HbA1c levels by calendar year, as treatment standards changed with the introduction of CGMs. This contributed to improvements in HbA1c levels in both groups but at different time points.

Table 2.

Generalized linear model, crude comparison of HbA1c and the different groups (model 1), second model adjusted for age categories, diabetes duration, insulin regimen (model 2), and a third model adding interaction effect to analyze changes over time (model 3)

Factor	Category	Valid observations, N	Crude model (1)			p value	Adjusted model (2–3)			p for comparison between fixed and CC	Interaction
			estimated mean fixed	estimated mean CC	mean difference (95% CI for difference)		adjusted mean estimates fixed	adjusted mean estimates CC	mean difference (95% CI for difference)		p for interaction group * duration
Treatment	​	​	​	​	​	0.23	​	​	​	0.27	​
​	Fixed	1,884	58.6	57.5	1.1 (−0.7 to 3.0)	​	59.0	57.9	1.0 (−0.8 to 2.9)	​	​
​	CC	2158	​	​	​	​	​	​	​	​	​
​	Total	4,042	​	​	​	​	​	​	​	​	​
Duration, years	​	​	​		​	<0.01	​	​	​	​	<0.001
​	0	460	50.5		​	​	50.9	49.8	1.1 (−1.2 to 3.5)	0.35	​
​	1	469	55.4		​	​	56.6	54.6	1.9 (−0.4 to 4.3)	0.10	​
​	2	458	58.0		​	​	58.8	57.9	0.9 (−1.4 to 3.2)	0.46	​
​	3	448	59.6		​	​	60.0	60.0	−0.1 (−2.4 to 2.3)	0.97	​
​	4	450	60.2		​	​	60.7	60.8	0.0 (−2.4 to 2.3)	0.98	​
​	5	452	60.1		​	​	61.4	59.9	1.4 (−0.9 to 3.7)	0.24	​
​	6	439	60.1		​	​	62.5	58.7	3.8 (1.5 to 6.2)	0.00	​
​	7	438	59.4		​	​	60.9	58.7	2.2 (−0.1 to 4.6)	0.06	​
​	8	429	59.5		​	​	59.7	59.9	−0.2 (−2.6 to 2.1)	0.85	​

The table shows no differences between CC and fixed, but HbA1c changes differently over duration in the fixed and the CC group.

Fig. 1.

Associations between time and mean HbA1c. a Estimated mean HbA1c levels according to diabetes duration and calendar year across treatment groups. b Estimated mean HbA1c between groups, stratified by insulin regimen.

IPT was associated with lower HbA1c in both groups (Fig. 1b). Use of insulin pumps increased over the study period in both groups, reaching 64% in the fixed group and 62% in the CC group after 8 years diabetes duration.

Comparison of HbA1c and Use of CC at Eight Years of Diabetes Duration

A total of 341 (72%) participants had a reported variable for use of CC (yes/no) after 8 years of diabetes diagnosis. Table 3 shows that the group using CC was younger and used IPT to a greater extent.

Table 3.

Characteristics of participants per group, yes CC vs. no CC after 8 years diabetes duration

Carbohydrate counting, n = 341		Yes CC, n = 250	No CC, n = 91	p value
Age, years		14.9 (3.8)	17.8 (3.6)	<0.001*
Gender (female), n (%)		122 (48.6%)	49 (53.3%)	0.466
HbA1c, mmol/mol
No, n = 91		56.2 mean	64.8 mean	<0.001*
Yes, n = 250		55 median	59 median
Glucose metrics	Mean blood glucose	9.2 mean	9.8 mean	0.074
		9.2 median	9.4 median
No, n = 53	TIR (3.9–10 mmol/L)	56.9% (19.7)	49.7% (20.4)	0.023*
Yes, n = 181	TITR (3.9–7.9 mmol/L)	40.7% (14.2)	35.4% (14.7)	0.020*
	Variability (SD)	3.6 (0.8)	4.0 (1.1)	0.004*
BMI
No, n = 72	Normal weight	185 (77.7%)	48 (66.7%)	0.063
Yes, n = 238	Overweight	53 (22.3%)	24 (33.3%)
Insulin regimen
No, n = 77	MDI	n = 35 (14.0%)	n = 64 (70.3%)	<0.001*
Yes, n = 233	Pump	n = 215 (86.0%)	n = 27 (29.7%)
Taught CC from onset of T1D		n = 156 (62.2%)	n = 52 (56.6%)	0.383

Participants who did not use CC at 8 years had higher HbA1c levels compared to those who reported use of CC (64.8 mmol/mol vs. 56.2 mmol/mol, crude difference: 8.6 mmol/L [95% CI: 5.2–12.0]) (p=<0.001) irrespective of insulin regimen (Fig. 2). The univariate general linear model showed that, after adjusting for gender, insulin regimen, and BMI categories with age as a covariate, the difference in HbA1c remained significant, with a mean difference of 6.0 mmol/L (95% CI: 1.9–10.2) (p = 0.004).

Fig. 2.

Cross-sectional comparison of HbA1c, irrespective of insulin regimen, at 8 years diabetes duration, between group who uses CC (n = 250) and group that does not use CC (n = 91) as reported to the NDR. Independent samples t-test, p = <0.001.

Analysis of participants with missing data, where there was no reported variable for use of CC, showed that the missing group was significantly older (mean age 15.7 years vs. 22.4 years). There were more males with missing values (p = 0.023) and they were less likely to use IPT (p=<0.001). There was no difference in proportion normal weight or overweight between groups, nor HbA1c between those with missing values and those with values for using CC.

Out of the participants with a reported variable for CC, 75% of those who had been taught CC used it 8 years after diabetes diagnosis and 70% of those who had been taught fixed doses at diagnosis had learnt and still used CC over the 8-year period. When comparing mean HbA1c within each group based on method taught at diabetes onset, there was no difference (p = 0.537, p = 0.215).

All glucose metrics showed improved outcomes when using CC except mean BG (Table 3). TIR as well as TITR was higher in those using CC and the glucose variability was lower when using CC, as compared to not using CC when adjusted for insulin delivery method.

Discussion

To our knowledge, this is the first study to examine if CC introduced at diabetes diagnosis to enable flexible dosing for meals, affects glycemic outcomes over time. In this study based on Swedish registry data, we found no difference in HbA1c outcomes over 8 years in the group taught CC at diabetes onset versus the group taught fixed dosing. The cross-sectional comparison of using CC or not 8 years after diabetes onset showed that the group using CC had significantly lower HbA1c and improved glycemic metrics including TIR, TITR, and SD irrespective of mode of insulin delivery. An interaction effect, that HbA1c levels differed between groups over time, was found and is explained by the introduction of CGM around the year 2016, which caused an improvement in HbA1c for both groups in the following years.

Other studies that have examined CC are mainly intervention studies with shorter follow-ups for 3–30 months [2, 16, 17]. The findings from these studies are in favor of CC as an insulin dosing strategy for meals; however, the early studies on CC showed conflicting results on its impact on HbA1c. In a more recent review, the lack of statistically significant results was assigned to heterogeneity of the studies, explained by CC technique (basic or advanced) and trial comparator (usual care or alternative dietary advice) [2]. To our knowledge, there is one cohort study, from a clinical setting where CC was introduced, but not from diagnosis [18], and one study on young children examining CC being introduced at diabetes diagnosis, with cross-sectional follow-ups in three 6-year intervals [13]. Both studies found that using CC improved HbA1c. The findings from our study add to the previous knowledge on the use of CC as we studied its use from diabetes diagnosis and over 8 years. Though we found no change in HbA1c over time between groups based on method learnt at diabetes onset, there were clinically significant lower HbA1c levels in the group who used CC after 8 years, with the majority of the cohort using CC.

The study settings arose from implementation of a new clinical treatment at the largest pediatric diabetes clinic in Sweden, changing the initial meal dosing strategy from fixed doses to CC at diabetes diagnosis. The change also impacted all existing patients, who were encouraged to learn CC and was required by anyone starting IPT. The high use of CC, also in the group that initially was taught fixed doses, likely impacted outcomes.

An important finding was the high use of CC after 8 years of diabetes duration, irrespective of method taught at onset of diabetes. CC has been recognized to allow flexible eating, as well as improve glycemic outcomes, but the method has also been questioned to add diabetes burden due to the time and effort required to count the carbohydrate content of each meal which could impact adherence [2, 17, 19, 20]. Some studies, that have examined adherence to CC, have reported reduced adherence rates over time [21]. Centenaro and colleagues reported an adherence rate of 69.2% at the end of their cohort study [18]. Donzeau and colleagues [21] found a decrease in CC use from 85% to 65% between months 3 and 12 of their intervention with no impact on HbA1c. Though burden in relation to CC or quality of life was not examined in this study, our finding that 75% used CC implies that CC is a feasible method of calculating mealtime doses that participants continue to use [2, 21]. Furthermore, other studies that have examined CC in association to quality of life or treatment satisfaction have found a positive relationship, which along with our findings demonstrate that advantages outweigh hindrances [2, 21, 22].

IPT was associated with improved glycemic outcomes over time in both groups. Modern technologies, such as both IPT and CGM, have simplified diabetes management. Insulin pumps offer flexible meal dosing strategies as well as insulin dosing without injection, which is an advantage as number of bolus doses are associated to improved glycemic outcomes [23, 24]. However, early comparisons of HbA1c outcomes, before the introduction of CC in Sweden, between those using IPT therapy and MDI, showed no differences in HbA1c [25, 26]. Similarly, international comparisons of HbA1c outcomes in registry data also imply that increased access to diabetes technology alone is not a predictor for improved glycemic outcomes [8, 9]. The insulin pump offers simplified bolus calculation, as ICRs are programmed into the pump and studies examining CC and IPT, using the bolus calculator vs. only the insulin pump, report greater improvements in HbA1c and postprandial variability when the bolus calculator was used [27–29].

When Norway presented improved outcomes over a 10-year period from their quality registry, the improvements were attributed to both diabetes technology such as IPT and CGM but also CC and participating in quality projects [10]. Furthermore, the diabetes team at Astrid Lindgren Children’s Hospital participated in a national quality improvement collaborative with 13 other diabetes teams across Sweden, starting in 2012 [30]. The work involved improved local guidelines, the introduction of CC with all patients as well as lowering the national HbA1c target to 52 mmol/mol (<7%). Having a consistent team approach in implementing and establishing CC as the standard meal dosing strategy for all new-onset diabetes, but also for all other patients as well, was likely to impact results, across the clinic and not only those with diabetes onset from 2012 [31].

We found that HbA1c increased over time, which can be a result of loss of remission but also that many participants reached adolescence and young adulthood where missed boluses are common and glycemic outcomes are known to deteriorate [32]. Even though the cohort is mainly teenagers and young adults, estimated mean HbA1c levels reached 60 mmol/mol, which was only slightly above the ISPAD guideline recommendations at the time [33]. Also, the adult clinics where adolescents were referred had not participated in quality improvements, had not lowered their HbA1c targets, and did not use CC as default insulin method for meals, thus meaning the young people had less support. Continued education from a multidisciplinary diabetes team is crucial to support both CC and dietary counseling and to sustain glycemic outcomes [5, 31, 34].

Limitations

This study is based on registry data, manually reported by the diabetes team following each visit. The allowance of reporting some, but not all variables, rendered incomplete reported data for some visits. This was seen in the reporting of height which impacted the possibilities to have complete BMI data for all participants over time. New variables, such as those for glucose metrics, were added to the NDR during the study years, as was the variable for CC. Furthermore, the variables were added to NDR kids and NDR at different time points. This compromised some of the variables of interest in this study, which could not be analyzed due to no reporting or low reporting rate. Furthermore, the use of registry data fails to address the complexity of diabetes management as many factors affect glucose outcomes that could not be measured or adjusted for in this study, e.g., diet quality, meal structure, and level of physical activity. Nor is there information on how and when insulin pump treatment was started and training in CC for those in the fixed group which could impact outcomes.

The variable CC is a subjective assessment by the diabetes team that may not be consistent with daily practice, nor does it relate, to what extent, someone uses CC. This fact, together with the lack of information on insulin dosing behaviors, such as missed boluses or timing of dose in relation to meals, is a limitation in this study [3, 24, 35]. As implied in the systematic review by Builes-Montaño, heterogeneity in CC technique and level of CC may impact HbA1c outcomes [2]. Furthermore, accuracy of counting carbohydrates is associated with improved outcomes, and this was not measured in this study [1, 36].

This study was not a randomized control trial, which would have been the study of choice to more robustly answer the primary aim. More studies are needed to further examine the long-term effect of CC, including meal bolus behaviors, if precise carbohydrate estimation, as well as ICR calculations, are needed for every meal to achieve in target glycemic outcomes.

Conclusions

The results suggest that CC is a well-accepted meal dosing strategy that can be introduced at any time after diabetes diagnosis. After 8 years of diabetes duration, a majority used CC, and those using CC had better glycemic outcomes.

Statement of Ethics

This study was performed in accordance with the Declaration of Helsinki. This human study was approved by Swedish Ethical Review Authority – approval: 2021-04903. Parent, guardian, or next of kin consent was not required for the minors because the data used are from a national quality registry. Each legal guardian and adult participant has given consent for data to be reported to the Swedish National Diabetes Registry.

Conflict of Interest Statement

C.E.S. was a member of the journal’s editorial board the time of submission. The remaining authors have no conflicts to declare.

Funding Sources

This research was funded by The Swedish Diabetes Foundation, Grant No. DIA2022-711, and Svenska Diabetesstiftelsen, Grant No. 2025-003. The funding sources had no role in the study design, execution, or manuscript preparation.

Author Contributions

E.J. and A.-L.O. conceptualized the study. E.J., H.J.P., and A.-L.O. performed statistical analyses. E.J. wrote the first draft of the manuscript. A.-L.O., C.E.S., H.J.P., U.K., and A.-L.B. were responsible for writing, review, and editing of the manuscript. E.J. coordinated the data collection of the study. A.-L.O., C.E.S., U.K., and A.-L.B. supervised the project. All authors interpreted the data, edited, and approved the final version of the manuscript.

Funding Statement

This research was funded by The Swedish Diabetes Foundation, Grant No. DIA2022-711, and Svenska Diabetesstiftelsen, Grant No. 2025-003. The funding sources had no role in the study design, execution, or manuscript preparation.

Data Availability Statement

The data presented in this study are available upon request from the corresponding author. Registry data used to support the findings of this study are not available due to ethical reasons.