Introduction

Living with type 1 diabetes (T1D) can be challenging due to the daily self-management required, including administering insulin and monitoring blood glucose levels at least 4–6 times a day. A substantial proportion of people with T1D develop diabetes distress (DD), defined as the emotional distress associated with the worries, burdens and concerns that occur when living with and managing diabetes over time.1 It is estimated that up to 42% of adults with T1D experience DD1; however, prevalence differs across groups of adults with diabetes and tends to be higher for women and younger adults, those with less social support, as well as for individuals with relatively high body mass index and prevalent diabetes complications.2–7 In addition, high levels of DD are associated with suboptimal diabetes self-management and higher glycaemic levels,2 4 7 and DD should preferably be reduced or prevented.

People with T1D are mainly treated with either multiple daily insulin injections (MDI) or insulin pump therapy (IPT). Compared with MDI, several studies have found that IPT can reduce haemoglobin A1C (HbA1c) levels and decrease variability in HbA1c.8–13 IPT is also associated with improved treatment satisfaction,9 12 14–16 lifestyle and treatment flexibility,9 as well as improved quality of life15 and reduced fear of hypoglycaemia.11 17 With regard to DD, several observational studies have found that IPT is associated with decreased DD levels, compared with MDI10 11 18; however, some studies report no differences.19 Interestingly, only two studies explored DD as their primary outcome and reported conflicting results. Whereas no significant differences in DD levels were observed between IPT and MDI users in the study by Wardian et al,19 a more recent study by Dowling and Maidment18 found that those using IPT reported lower levels of DD, specifically management and physician-related distress, compared with those using MDI.

Monitoring glucose levels is essential for treating T1D, independent of the insulin delivery method used. Compared with a traditional blood glucose meter (BGM), the use of either intermittently scanned continuous glucose monitoring (isCGM) or real-time continuous glucose monitoring (rtCGM), collectively CGM, has been found advantageous on several parameters. These include reduced HbA1c,20–26 improved treatment satisfaction,21 improved quality of life and well-being21 27 28 and less fear of hypoglycaemia21 27–29 compared with BGM, regardless of insulin delivery method. Moreover, the use of CGM has been associated with lower levels of DD in observational studies20 24 28 and in a randomised controlled trial (RCT).30

It is clear from the literature that the use of IPT and/or CGM has the potential to improve both glycaemic outcomes and reduce the psychosocial challenges of diabetes, including DD. In this study, we aimed to analyse associations between the insulin delivery method (IPT vs MDI), glucose monitoring method (CGM vs BGM) and DD in individuals with T1D, both independently and for distinct combinations of devices.

Research design and methodsSetting

In Denmark, most individuals with T1D receive care in specialist diabetes outpatient clinics funded by the public healthcare system.31 Thus, diabetes care is generally free at the point of delivery, except for medications, which are, nonetheless, heavily subsidised. In addition, fully subsidised IPT is available by prescription and indicated according to certain criteria as specified by the Danish Health Authority, including not being able to achieve HbA1c levels below 53 mmol/mol (7.0%) and experiencing recurring hypoglycaemic episodes with MDI.32 Likewise, fully subsidised CGM is available by prescription to adults with T1D with HbA1c levels above 70 mmol/mol (8.5%), hypoglycaemia unawareness, problems measuring capillary finger prick glucose during work and/or other individual indications. At the time of this study, types of IPT available by prescription included conventional pumps such as Ypsopump, patch pumps such as Omnipod and the first generation of hybrid closed systems, for example, the MiniMed 670G. Types of CGM included Dexcom 4, 5 and 6 and Eversense. At the time of the survey, the first-generation Flash Libre device was the only isCGM system available through the Danish healthcare system.

Study design and participants

Data used in this study came from two questionnaire-based online surveys conducted in Denmark in 2020, namely the Steno Tech Survey33 (see online supplemental material 1 for the original version and online supplemental material 2 for an English translation) and the T1-DDS validation survey,34 as well as data from national Danish registries. Individuals aged 18 years with T1D who were using IPT and attending Steno Diabetes Centre Copenhagen or Nordsjællands Hospital Hillerød were invited to participate in the Steno Tech Survey (n=1591); individuals using MDI at Steno Diabetes Centre Copenhagen and individuals treated at Holbæk Hospital or Sjællands Universitetshospital, regardless of treatment modality, were invited to participate in the T1-DDS validation survey (n=4205). Invitations to both surveys were sent out between June and October 2020 via a secure email system. Non-responders and partial responders received two reminders. Informed consent from all participants was obtained digitally before they responded to either survey.

Procedures

To facilitate comparative analysis of DD levels, we combined data from the two surveys, retaining only the information collected in both, namely the insulin delivery method (IPT or MDI), glucose monitoring method (isCGM/rtCGM or BGM) and DD scores. The 28-item Type 1 (T1-DDS)2 34 was used for measuring DD. The T1-DDS provides a total DD score, as well as seven subscale scores for, respectively, powerlessness, management distress, hypoglycaemia distress, negative social perceptions, eating distress, physician distress and friend and family distress. Each item consists of a statement that the respondent rates on a six-point Likert scale, where 1 corresponds to ‘not a problem’ and 6 to ‘a very serious problem’. The total DD score is calculated by averaging scores across all items; scores for each subscale are calculated by averaging items belonging to the subscale in question. DD scores are categorised as little or no distress at a score of 1.0–1.40, mild distress at 1.5–1.9, moderate distress at 2.0–2.9 and high distress at ≥3.0.2

Covariates

Using the unique Danish personal identifier, we linked the combined survey data to sociodemographic and clinical data from national Danish registries. Sociodemographic data included variables on age, sex, civil and cohabitation status, number of children living at home, educational attainment, employment history, annual disposable income and country of origin. Clinical variables included diabetes duration, the number of diabetes-related hospital visits (during 2020), microvascular and macrovascular diabetes complications and prevalent psychiatric disorders (depression or anxiety). We also calculated a 5-year Charlson Comorbidity Index score, reflecting the relative risk of 1-year all-cause mortality, to account for other somatic comorbidities.35 In the case of both sociodemographic and clinical variables, inclusion as covariates was based on an assessment of the potential impact that they have on both access to and management of diabetes technology and on the manifestation of diabetes distress.

Statistical analysis

We assessed differences in sociodemographic and clinical characteristics between IPT and MDI users with χ2 tests for categorical variables and non-parametric Mann-Whitney U tests for continuous variables, except for the Charlson Comorbidity Index for which we used a two-sided t-test.

To investigate associations of total T1-DDS and subscale scores between IPT and MDI users and between CGM and BGM users, we used generalised linear regression models with heteroskedastic robust SEs. We adjusted each model for all previously mentioned sociodemographic and clinical variables, except for diabetes duration due to multicollinearity with age, to control for potential confounding. We adjusted the IPT versus MDI model also for the method of glucose monitoring (CGM vs BGM), and similarly, we adjusted the CGM versus BGM model for the insulin delivery method (IPT vs MDI) to obtain estimates independent of the possible combination of devices.

To investigate the potential effect modification of distinct combinations of devices, that is, IPT with or without CGM vs MDI with or without CGM, on total T1-DDS and subscale scores, we conducted generalised linear regressions, including an interaction between the insulin delivery method and glucose monitoring method variables. This resulted in four distinct groups: MDI with BGM, MDI with CGM, IPT with BGM and IPT with CGM. Moreover, to investigate the potential effect modification of different types of CGM (isCGM vs rtCGM) on total T1-DDS and subscale scores, we conducted similar analyses but on the following groups: MDI with isCGM, MDI with rtCGM, IPT with isCGM and IPT with rtCGM. Thus, in this analysis, individuals using BGM were not included. The significance level was set to 0.05 in all analyses. Analyses were conducted in SPSS V.27 and Stata V. 17.

Patient and public involvement

The preliminary version of the questionnaire was pilot-tested with seven HCPs experienced in insulin pump therapy but with no involvement in the project and six people with T1D, purposefully sampled according to age, profession, complication status and insulin pump type. Further cognitive interviewing with people with T1D led to minor revisions of content and language-related shortcomings, and the comprehensiveness of response categories for non-standardised items was confirmed.

ResultsSample characteristics

In total, 2201 individuals responded to one of the two surveys, resulting in a response rate of 38%. Due to missing responses on key variables, 133 individuals were excluded from further analysis. Thus, data for 2068 individuals were analysed. The characteristics of the study participants are shown in table 1. Of these, 799 (38.6%) were IPT users and 1269 (61.4%) were MDI users. Significantly more IPT users, compared with MDI users, used rtCGM (48.7 vs 14.8%) for glucose monitoring, while fewer used BGM (24.4 vs 49.7%) or isCGM (26.9 vs 35.5%). In addition, IPT users were significantly younger (median (IQR): 50 (37–60) vs 56 (45–66)), more likely to be female (60.9 vs 44.2%), have one or more children living at home (38.5 vs 29.2%), cohabitate with someone else (77.2 vs 76.1%) or only with children (4.5 vs 2.8%), have a medium cycle (28.9 vs 23.1%) or long cycle (23.9 vs 17.3%) higher education and be employed (67.1 vs 62.7%) or under education (10.3 vs 3.2%). IPT users also had lower HbA1c (56 (50–61) vs 58 (52–65) mmol/mol, 7.3 (6.7–7.7) vs 7.5 (6.9–8.1) %), scored lower on the Charlson Comorbidity Index (mean±SD: 0.84±0.87 vs 0.94±0.95) and were less likely to have microvascular (62.6 vs 68.2%) and macrovascular (36.8 vs 45.2%) complications, compared with MDI users. However, IPT users had more diabetes-related hospital contacts within a 1-year period (7 (4–10) vs 5 (4–8)) and were more likely to have been diagnosed with a psychiatric illness (9.8 vs 6.1%) than MDI users.

Table 1

Sample characteristics

The distribution of DD levels and total T1-DDS and subscale scores by insulin delivery and glucose monitoring methods are shown in table 2. The highest proportions of individuals with moderate or high levels of DD were among individuals using a combination of MDI and isCGM (48.6%) or IPT and isCGM (50.7%), while the lowest proportions were among individuals using MDI and BGM (31.7%) or IPT and BGM (39.5%). Individuals using a combination of MDI and BGM also had the lowest total T1-DDS score (1.79±0.65), while the highest total T1-DDS score was found among those using a combination of MDI and isCGM (2.07±0.79). Among all six groups, the highest T1-DDS scores were found on the powerlessness distress subscale.

Table 2

Distribution of diabetes distress levels by insulin delivery and glucose monitoring methods

Impact of various insulin delivery and glucose monitoring methods

The results from the generalised linear regressions are shown in table 3. Focusing first on IPT vs MDI and controlling for the glucose monitoring method, we observed a significantly lower total T1-DDS core in IPT users (−0.09, 95% CI−0.16 to −0.03). Similarly, we found that the powerlessness subscale score (−0.10, 95% CI −0.21 to −0.01), the management distress score (−0.17, 95% CI −0.25 to −0.09), the hypoglycaemia distress score (−0.14, 95% CI −0.24 to −0.05), the negative social perception score (−0.10, 95% CI −0.18 to −0.01), the eating distress score (−0.12, 95% CI −0.22 to −0.02) and the physician distress score (−0.16, 95% CI −0.24 to −0.07) were lower among IPT users compared with MDI users. On the contrary, compared with BGM and controlling for insulin delivery method, use of CGM was associated with a significantly higher total T1-DDS score (0.11, 95% CI 0.05 to 0.18), powerlessness score (0.23, 95% CI 0.14 to 0.33), hypoglycaemia distress score (0.22 95% CI 0.14 to 0.31), negative social perception score (0.12, 95% CI 0.04 to 0.19), eating distress score (0.20, 95% CI 0.11 to 0.29) and friend and family distress score (0.09, 95% CI 0.03 to 0.16).

Table 3

Associations between IPT and diabetes distress scores and CGM and diabetes distress scores

Pairwise comparisons of device combinations

Pairwise differences of T1-DDS scores from generalised linear models for distinct combinations of devices revealed that individuals using MDI with CGM had a higher total T1-DDS score compared with those using MDI with BGM (0.15, 95% CI 0.08 t0.22) as shown in table 4. With regard to IPT users, regardless of using either BGM (−0.17, 95% CI −0.28 to −0.06) or CGM (−0.13, 95% CI −0.21 to −0.05), they had a significantly lower T1-DDS total score when compared with users of MDI with CGM. We observed the same tendency across most of the T1-DDS subscales (table 4).

Table 4

Pairwise differences between IPT versus MDI with or without CGM

Table 5

Pairwise differences between IPT versus MDI with either isCGM or rtCGM

Similar to table 4, table 5 also presents pairwise differences of T1-DDS scores from generalised linear models but with CGM separated into isCGM and rtCGM. Despite being based on a smaller sample size, that is, without BGM users, the magnitude of main effect sizes was larger for IPT versus MDI; however, there were no differences in main effect sizes between rtCGM and isCGM (apart from friend and family distress). We observed no differences in total T1-DDS or in any subscale score, neither in the MDI with rtCGM versus MDI with isCGM comparisons (apart from friend and family distress), nor in the IPT with rtCGM versus IPT versus isCGM. Comparisons across insulin delivery methods revealed, however, that, for instance, users of IPT and isCGM had lower total T1-DDS and subscale scores, compared with users of MDI and isCGM, which was also the case for IPT with rtCGM versus MDI with isCGM.

Conclusions

In this study, we investigated the independent and joint associations of insulin delivery methods (MDI and IPT) and glucose monitoring methods (BGM and CGM) with DD in adults with T1D. While the use of IPT was associated with a significantly lower total T1-DDS score, including most subscale scores, CGM was associated with a higher total T1-DDS score and most subscale scores. Moreover, when investigating DD levels between different combinations of insulin delivery and glucose monitoring methods, we found that IPT was associated with a lower total T1-DDS score, as well as lower scores on most subscales, regardless of being combined with BGM or CGM, compared with MDI with CGM. The use of MDI was associated with higher total T1-DDS and subscale scores when combined with CGM compared with BGM. When restricting our analysis only to individuals using isCGM or rtCGM, we found no significant associations between the type of CGM (isCGM vs rtCGM) and DD levels among either IPT users or MDI users. However, the use of IPT combined with isCGM or rtCGM was associated with lower levels of DD, compared with MDI with isCGM.

Several observational studies have found IPT to be associated with lower levels of DD among adults with T1D, regardless of whether DD was measured with the T1-DDS18 or with the Problem Areas in Diabetes Scale.10 11 Thus, our findings are consistent with the existing literature, except for the study by Wardian et al who found no difference in DD between individuals using IPT and MDI.19 One reason for this difference could be that participants in Wardian et al’s study generally reported low levels of DD compared with participants in this and other studies. The level of ‘total’ DD in this study is similar but slightly lower than that of the study by Dowling and Maidment, who also used the T1-DDS (2.57 for IPT users and 2.83 for MDI users).18 The use of IPT eases the daily management of diabetes, thereby providing several benefits that may explain the lower levels of DD among individuals using IPT compared with MDI. For instance, adults with T1D report feeling more in control of their diabetes and being able to ‘live more freely’ after changing from MDI to IPT.36 IPT provides the ability to enjoy a more flexible lifestyle, avoid the discomfort of injecting insulin several times a day and allow for a more discrete administration of insulin. Furthermore, IPT users report being able to be more spontaneous with regard to physical activity and diet, for example, eating outside of the home.14

As with IPT, several studies, including cross-sectional, cohort and RCTs, have also found CGM to be associated with better psychosocial outcomes, such as lower DD levels, among adults with T1D.20 24 28 30 Our results regarding BGM versus CGM to some degree conflict with these findings; however, this difference may due to the fact that some studies exclusively included individuals using MDI for insulin delivery,24 30 did not adjust for insulin delivery method28 or exclusively included adolescents20 who may be more comfortable with engaging with new technology.37 Despite these differences, the overall levels of DD found in other studies using the T1-DDS are similar to ours: participants in Al Hayek et al’s study had a mean total T1-DDS score of 2.5, 3 months after initiating CGM, versus 3.8 at baseline24; Polonsky et al found a mean total score of 2.2 among CGM users versus 2.5 among controls,28 and Polonsky et al found a mean total score of 1.8 among CGM users versus 1.7 in controls.30

Although the use of CGM has been found to produce several benefits, such as feeling more confident in managing glucose levels, experiencing less fear of hypoglycaemia during sleep, providing a better understanding of the glycaemic effect of food and easing social participation at home and work,38 it may also be related to several burdens for some people. For instance, the use of CGM is related to barriers, such as spending a lot of time interpreting data from the CGM (information overload), the stress of being confronted with poor results and blood glucose fluctuations,36 38 perceived inaccuracy of the CGM, having to wear a device on the body and being disturbed by hypoglycaemia or hyperglycaemia alarms, creating so-called ‘alarm fatigue’.38 Alarm fatigue, in particular, has been reported as a major barrier related to CGM use.39 In our study, however, alarms were only available with rtCGM systems, as the only available isCGM system (FreeStyle Libre) in Denmark at the time of data collection did not come with alarms. Thus, since we found no significant difference in DD levels between individuals using rtCGM and isCGM, alarm fatigue is unlikely to explain the higher levels of DD found among CGM users, compared with BGM users.

Taken together, our findings indicate that the use of different technologies in the management of T1D can be both supportive and induce distress. Naranjo et al examined psychological factors associated with technology use among adults with T1D and found that all participants experienced moderate levels of DD, regardless of technology use, which they speculate might be due to the fact that technology does not remove the daily tasks related to managing a chronic disease.40 Thus, practical implications of our findings include that decisions about insulin delivery and glucose monitoring methods should be based on an assessment of and with the individual with T1D to identify which combination of technology will generate the best clinical and psychosocial outcomes for that specific person. In addition, assessments of DD around the initiation of IPT and CGM may be relevant. While the need for regular assessment of DD is increasingly recognised in both national and international treatment guidelines,7 the change from one insulin delivery or glucose monitoring method to another may require additional assessment of DD. According to Fisher et al, screening for DD should vary by clinical need and be performed more frequently around high-risk events, such as the onset of new complications or during diabetes education.41 Initiation of new technology may constitute such a high-risk event, which requires additional monitoring of the individual’s psychosocial well-being so that relevant support for managing DD can be offered if necessary.

Strengths related to this study include the large study population, making it possible to compare DD levels between people using different combinations of insulin delivery and glucose monitoring methods, as well as to delimit analysis to those using CGM and investigate the effect of CGM type (isCGM vs rtCGM). Also, the ability to distinguish between isCGM and rtCGM to explore their potentially different associations with DD, including the effect of alarms, is a strength and no previous studies have done that. Furthermore, the availability of data on a large number of sociodemographic as well as clinical characteristics of our study population allowed us to adjust for several variables known to be associated with DD, for example, age, sex, HbA1c and diabetes complications. Lastly, the use of a well-established and validated scale, the T1-DDS, for measuring the study outcome is also a strength.

The study also has limitations. First, the cross-sectional nature of the survey precludes a causal inference about the relationship between insulin delivery, glucose monitoring methods and DD levels. The reasons for the prescription and use of IPT and, especially, CGM in Denmark are not only based on clinical outcomes but also on the clinical assessment of a person’s ability and motivation for investing time in using these technologies in their daily lives. We cannot exclude that the observed associations are due to confounding indications. It is plausible to believe that IPT is prescribed to individuals with better psychosocial well-being, while CGM in some situations is prescribed to those with worse psychosocial well-being, but as device prescription, in addition to guidelines, is based on individual decisions, there is no definitive support for this supposition. Second, the average response rate to the Steno Tech and T1-DDS validation surveys of 38.2% was relatively low, increasing the risk of selection bias. For instance, most non-responders in the Steno Tech survey were both younger and had higher HbA1c33 than responders, and those characteristics are both positively correlated with DD. Thus, the study population may not be representative of the general population of adults with T1D in Denmark, and the magnitude of the observed associations may be underestimated due to the selection of people with better clinical outcomes. Third, as information about DD came from two surveys, potential differences in the overall framing of questions in these surveys may have influenced how participants responded. Also, as both surveys were conducted in 2020, the presence of the COVID-19 pandemic may have affected participants’ psychosocial well-being, including DD levels, negatively. However, the prevalence of moderate to high DD in this study was similar to the prevalences reported before the pandemic.1 The large number of statistical comparisons included in our results entails an increase in the risk that statistical significance is the outcome of chance, but we have sought to assess the results in more general terms rather than focussing on statistically significant outcomes seen in isolation. Lastly, we did not consider the use of automatic insulin delivery (AID) systems, where an insulin pump automatically delivers insulin based on data collected by a CGM. Such systems have been shown to improve glycaemic outcomes and reduce DD levels42 and will likely play an important role in diabetes management in the future. As it was, however, only a few participants in this study used AID.

In conclusion, this study showed that compared with MDI, the use of IPT for insulin delivery was associated with lower levels of DD among adults with T1D, regardless of glucose monitoring modality. However, the use of CGM for glucose monitoring was associated with higher levels of DD, regardless of the insulin delivery method. The type of CGM used (isCGM vs rtCGM) did not affect associations between insulin delivery method and DD. Inasmuch as the results may be interpreted as reflecting bias derived from the more or less formal criteria for the prescription of devices, the data we present do not provide support for the view that either IPT or CGM, alone or in combination, have a transformative impact on psychosocial outcomes. Due to the cross-sectional design of the study, however, clearcut conclusions about the impact of devices on the psychosocial burden of living with diabetes remain elusive. In the future, therefore, the effect of diabetes technology on psychosocial outcomes should be investigated using different study designs, for example, longitudinal designs or RCTs.