Capturing the real‐world benefit of residual β‐cell function during clinically important time‐periods in established Type 1 diabetes

Abstract Aims Many individuals with type 1 diabetes retain residual β‐cell function, with increased endogenous insulin secretion associated with reduced hyperglycaemia, hypoglycaemia and glycaemic variability. However, it is unknown when these improvements occur during the day. Dysglycaemia is common in overnight and postprandial periods and associated with diabetes complications. Therefore, this study aimed to determine the influence of residual β‐cell function upon nocturnal and postprandial glycaemic control in established type 1 diabetes. Methods Under free‐living conditions, 66 participants wore a blinded continuous glucose monitor (CGM), kept a food diary, and completed a stimulated urine C‐peptide creatinine (UCPCR) test. Nocturnal, and postprandial CGM outcomes (participant means and discrete event analysis) were compared between UCPCR groups: undetectable (Cpepund), low (Cpeplow: 0.001–0.19 nmol/mmol) and high (Cpephigh: ≥0.2 nmol/mmol). Results Greater β‐cell function was associated with incremental improvements in glycaemia. Cpephigh spent significantly greater time in normoglycaemia than Cpepund overnight (76 ± 20% vs. 58 ± 20%, p = 0.005) and 0–300 mins postprandially (68 ± 22% vs. 51 ± 22%, p = 0.045), while also having reducing nocturnal variability (SD 1.12 ± 0.41 vs. 1.52 ± 0.43 mmol/L, p = 0.010). Analysis of individual events, controlling for diabetes duration, BMI, basal insulin, use of a continuous or flash glucose monitor and (for postprandial) meal type, carbohydrate and bolus insulin intake, replicated the group findings, additionally demonstrating Cpepund had increased hyperglycaemia versus Cpeplow overnight and increased postprandial hypoglycaemic events compared with Cpephigh. For all participants, breakfast had a significantly higher incremental area under the curve than lunch and dinner. Conclusions Residual β‐cell function is associated with improved nocturnal and postprandial glycaemic control. These data may be of clinical importance for identifying specific periods and individuals where further glycaemic management strategies would be beneficial.

One considerable advantage of using CGM devices over other measure of glycaemic control, such as HbA1c, is the ability to capture free-living daily glucose profiles. Indeed, CGMs have highlighted nocturnal and postprandial glycaemic events as common periods of dysglycaemia in type 1 diabetes. 9,10 These time periods are associated with diabetes complications and often cited as concerns by patients. [11][12][13][14][15][16] Additionally, individuals with type 1 diabetes tend to have large intraday variation in blood glucose, which may be influenced by circadian rhythms. Similar carbohydrate intake throughout a day can lead to substantially different postprandial events, 17 with early morning increases in circulating growth hormones 18 and reduced insulin sensitivity 19,20 likely contributing to large breakfast postprandial events. 17 The influence residual βcell function has on glycaemic control at these clinically important time periods has not previously been explored under free-living or laboratory settings. Therefore, this study examined the real world impact of residual βcell function upon nocturnal and postprandial glycaemic control, specifically, exploring incremental increases in stimulated C-peptide on individual events, time in range, and responses to different timed meals across the day.

METHODS
Participants with a clinical diagnosis of type 1 diabetes (primary osmotic symptoms, weight loss, hyperglycaemia, ketosis, insulin initiation at diagnosis), aged 18-65 years with a diabetes duration ≥1 year, HbA1c <86 mmol/mol (10.0%) and stable multiple daily injections or continuous subcutaneous insulin infusion regimen without changes over the preceding 6 months were recruited from the Newcastle Diabetes Centre. Participants provided written informed consent. This study was a secondary analysis of the participant recruitment phase of an observational exercise study exploring residual βcell function influence on post-exercise glucose. 21 Given that these findings were exploratory, it wasn't possible to pre-specify the anticipated outcomes at the time of study registration (ISRCTN50072340). The study was approved by the local National Health Service Research Ethics Committee, Newcastle, U.K. (code: 16/NE/0192).

K E Y W O R D S
continuous glucose monitoring, nocturnal, postprandial, residual βcell function What is already known?
• Many individuals with type 1 diabetes have residual βcell function secreting micro levels of insulin and C-peptide which is associated with improved glycaemic control. It is unknown when these improvements occur during the day.

What this study has found?
• Greater βcell function was associated with more time spent in normoglyceamia overnight and postprandially reduced hyperglycaemia and glycaemic variability overnight and reduced postprandial hypoglycaemic events. Even low C-peptide had glycaemic benefits compared with undetectable βcell function reducing hyperglycaemia overnight.
What are the implications of the study?
• Individuals with no βcell function may need more support to manage nocturnal and postprandial periods whereas those with C-peptide positivity should pursue more ambitious glycaemic targets.
monitors) and completed a food and insulin diary at home for 7-8 days. During the CGM collection period, participants completed a home 2 h-postprandial Urine C-peptide creatinine ratio (UCPCR) test, collecting urine in a Boricon container containing boric acid to stabilise C-peptide after their largest meal in a day, before posting to Exeter Clinical Laboratory. Samples were analysed for C-peptide using the routine automated E170 immuno-analyser from Roche Diagnostics, and creatinine was analysed on the Roche P800 modular analyser. UCPCR is highly correlated with post mixed meal tolerance test serum C-peptide, 22 the gold standard, and has inter-and intra-assay coefficients of variation of <4.5% and <3.3%, respectively. 23 Participants were grouped as follows: undetectable (Cpep und 0.000 nmol/mmol), low (Cpep low 0.001-0.19 nmol/mmol) or high (Cpep high ≥0.2 nmol/mmol). The lower limit of detection (0.001 nmol/mol) is the equivalent to serum C-peptide 3 pmol/L, while ≥0.2 nmol/mmol UCPCR is equivalent to >200 pmol/L serum C-peptide, a clinically defined level associated with reduced hypoglycaemia and microvascular complications. 3 Within the food and insulin diary, participants were asked to collect timings of all meals and snacks eaten during the CGM collection period, a description of the food eaten, estimated carbohydrate content and bolus insulin doses. Food and insulin data were recorded in real time on a paper diary and subsequently analysed after the data collection period.
CGM data were calibrated using capillary blood glucose values recorded in the participants' diaries, and downloaded into Microsoft ® Excel. Acceptance criteria for daily (midnight to midnight) CGM data were ≥4 calibrations a day, mean absolute relative difference <28% for a range of >5.6 mmol/L or <18% for a range <5.6 mmol/L, a correlations >0.79 between the calibrating blood glucose value and CGM and no missing data segments of >15 min. Only days that met the CGM data criteria, and the corresponding day's data from the food and insulin diary, were subsequently analysed. If the iPro2 failed to collect four valid days of data, the testing process was repeated. The primary outcome was percentage time spent in normoglycaemia (3.9-10 mmol/L). Secondary outcomes included measures of glycaemic variability ([GV], standard deviation [SD], coefficient of variation [CV]), mean, peak and delta glucose, [incremental] area under the curve ([i] AUC) and percentage time spent in/incidence of hypoglycaemia-1 (<3.9 mmol/L)/-2 (<3.0 mmol/L) and hyperglycaemia-1 (>10 mmol/L)/-2 (>13.9 mmol/L). 24 CGM data allowed group means and individual event outcomes to be analysed across 24hr (00:00-00:00 h), nocturnal (00:00-06:00 h) and postprandial (early: 0-120 mins, overall: 0-300 mins) periods. Prandial events started 15 mins prior to reported meal-time allowing for inaccuracy. Events were excluded if CGM values did not rise within 300 mins, or further food or insulin were taken within 120 mins (if taken within 120-300 mins then only 0-120 mins analysed). Values from all identified meals were used to calculate each participant's mean postprandial response, which was subsequently analysed between groups. Meal type was defined as breakfast: first carbohydrate-containing meal (06:00-10:00 h), lunch and dinner: largest carbohydrate-containing meals (11:00-15:00 h and 17:00-21:00 h) and 'other': remaining meals.
Analysis was performed using SPSS-27.0 (IBM CORP) and Rv4.04 using the lmer package. Normality and outliers were assessed, with skewed participants' mean data transformed. Participants' mean variables were compared between UCPCR groups using one-way ANOVA (Tukey post-hoc), or Kruskal-Wallis test, while a mixed-model ANOVA assessed glucose over time. Individual event analysis (continuous outcomes) were assessed by a mixed-effects linear regression, fitted with random effect for individuals and fixed effects for UCPCR category, adjusted for BMI and diabetes duration, basal insulin, use of a continuous or flash glucose monitor and (for postprandial) meal type, carbohydrate and bolus insulin intake. For binary outcomes, a mixed-effects generalised linear model was fitted, with random effects for individuals and fixed effects for UCPCR category, BMI and diabetes duration. Parameter effects and confidence intervals were extracted, with Wald test p-values. Data are presented as mean ± SD. A p-value <0.05 was considered statistically significant.

| RESULTS
Data from 66 participants (Cpep und [n = 34], Cpep low [n = 13], Cpep high [n = 19]) were collected. Age, HbA1c, BMI and use of MDI versus CSII were not statistically different between groups, while Cpep high were older at diagnosis with shorter diabetes duration. The majority of CGM/flash glucose monitoring users were in the Cpep und group, but there were no statistically significant differences between groups (Table 1).
For event analysis, 337 nights were analysed, and 843 postprandial events identified. Following removal of ineligible events (further food or insulin intake), 599 and 252 were analysed for early and overall postprandial periods. Figure 1 displays participant mean glycaemic outcomes across groups.
Mixed-effects analysis of individual events replicated overall group findings (Supplementary Table S1). Additionally, Cpep und had significantly higher nocturnal CV than Cpep high and greater %time hyperglycemia-2 versus Cpep low . Increased BMI was associated with higher mean, pre-bed, peak, AUC and nadir glucose, increasing %time in hyperglycemia-1 and decreasing %time in normoglycaemia. Use of a CGM or a flash glucose monitor did not statistically change time spent in normoglycaemia, F I G U R E 3 Group glycaemic outcomes during the early postprandial periods (0-120 mins). (A) displays group mean percentage time spent in glycaemic ranges. Orange represents time spent >13.9 mmol/L, yellow represents time spent 10-13.9 mmol/L, green represents time spent 3.9-10 mmol/L, pink represents time spent 3.0-3.9 mmol/L, and red represents time spent <3.0 mmol/L. (B) displays group mean (±SD) glucose time course post-prandially. (C) displays group box (representing median and interquartile range) and whiskers (representing 10-90th percentile values), and outlier data as individual participant data points for CGM metrics. Values from all identified meals were used to calculate each participant's mean postprandial response, which was subsequently analysed between the groups Cpep und (red lines) n = 32; Cpep low n = 13 (orange lines); Cpep high n = 19 (green lines). *Significantly different to Cpep und hyperglycaemia-1 or 2, or hypoglycaemia-1 or 2, despite CGM use reducing % time spent <3.9 mmol/L by a mean 6.3% (95% confidence intervals 15.5 to −2.9%, p = 0.205) compared with not using a glucose monitor.
Mixed-effects analysis (Supplementary Table S1) also found reduced time spent in hyperglycaemia-2 for Cpep high compared with Cpep und in the early postprandial period, while Cpep low had reduced post-prandial CV compared with Cpep und . Over 0-300 min, Cpep high were significantly less likely to have a hypoglycaemic-1 event than Cpep und . Increased BMI and meal insulin bolus were associated with higher postprandial glucose. Compared with breakfast, dinner had a significantly higher pre-meal glucose, while lunch, dinner and other meals had a significantly reduced iAUC. Use of flash glucose monitoring significantly reduced peak glucose, SD and CV in the early postprandial period and iAUC in both the early and overall postprandial period compared with not using a glucose monitor. CGM use did not significantly change any postprandial variable measured.

| CONCLUSIONS
Through analysis of clinically meaningful time periods during free-living, we have demonstrated for the first time that increasing βcell function in type 1 diabetes is associated with improved glycaemia overnight and postprandially. This builds upon previous work demonstrating that individuals with residual βcell function spend less overall freeliving time in hyperglycaemia or hypoglycaemia compared with those with undetectable βcell function. [5][6][7] Specifically, we show that Cpep high spent the most time in normoglycaemia overnight and had the lowest nocturnal variability. We also demonstrated Cpep und spent less time in normoglycaemia in postprandial periods (0-300 min). In a mixedeffect model, accounting for covariates, higher C-peptide was still associated with improved glycaemic outcomes. Additionally, a very low level of C-peptide was also significantly associated with meaningful reductions in overnight hyperglycaemia, whereas a high C-peptide was associated with less postprandial hypoglycaemic events, in comparison with individuals with no βcell function.
Increased βcell function likely improves nocturnal glycaemia due to endogenous responsivity to changing blood glucose. Rickels et al. 6 found that those with superior βcell function have greater C-peptide response to hyperglycaemic clamps, and greater glucagon response to hyperinsulinemic-hypoglycaemic clamps. Continued secretion of insulin into the portal vein, even the small amount seen in type 1 diabetes, likely attenuates highs through reduction of hepatic glucose production, with increased glucagon response lessening hypoglycaemia. The mechanisms explaining the enhanced glucagon response observed in those with high C-peptide is unclear, with theories including the suppression of functional βcells activating neighbouring αcells within intact islets 25 or the presence of C-peptide itself enhancing the counter regularity responses to hypoglycaemia. 26 This may lower GV and protect against nocturnal dysglycaemia, potentially giving individuals confidence to adhere to more intensive insulin regimens, lowering nocturnal mean glucose.
Despite Cpep high spending greater time in normoglycaemia postprandially (0-300 min), improvements in postprandial glucose with increasing βcell function were less marked. When analysed by mixed-model, Cpep low group had limited improvements in glycaemia compared with Cpep und , with Cpep high only having reduced likelihood of hypoglycaemic events. Paradoxical postprandial glucagon rises, caused partly by dysfunctional βcells unable to suppress αcells, 6,25 and unaffected by C-peptide level, 6 may partly explain the limited group differences.
Glycaemic improvements appear to be incremental with increasing βcell function. In our mixed-effects regression, Cpep low spent significantly less time in hyperglycemia-2 than Cpep und overnight. Combined with research demonstrating that residual βcell function reduces time spent in hypoglycaemia, 5 it appears that minimal Cpeptide secretion may offer glycaemic benefits. However, unlike Gibb et al., 5 no significant differences in %time in hypoglycaemia existed, despite Cpep high appearing to have clinically significant improvements nocturnally. This is likely due to the extremely skewed nature of time spent in hypoglycaemia data, making statistical analysis difficult, and our study being underpowered to detect these small, but important differences, if indeed genuine.
Irrespective of UCPCR, an increased BMI and larger meal bolus insulin were associated with higher postprandial and nocturnal glucose, potentially due to relative insulin resistance, 27 further highlighting the importance of maintaining a healthy weight in type 1 diabetes. 28 Despite increased mean glucose and %time spent in hyperglycaemia, a higher BMI was not protective against hypoglycaemia, with no association with %time spent in or incidence of hypoglycaemia level 1 or 2.
The association between postprandial glucose and bolus insulin is likely to be influenced by the timing of the insulin in comparison with the food intake, which was not recorded in this study and is, therefore, a limitation. Previous studies have demonstrated that bolus administration 20 min before a meal leads to reduced postprandial excursions in comparison with immediately before a meal. 29 Newer ultra-fast bolus insulins (Fiasp) also reduce postprandial excursions compared with more commonly used rapid acting bolus insulin (Humalog and Novorapid). 30 Fiasp was only used by 3 participants in the current study (1 with Cpep und and 2 with Cpep low ) and was, thus, not considered as a parameter in our mixedeffects models.
As standard care was maintained throughout the study, some participants were using real time CGM or flash glucose monitoring. Surprisingly, use of a CGM or flash monitor was not associated with improved time spent in normoglycaemia during nocturnal or postprandial periods, unlike previous studies that have found improvements over 24 h and nocturnally. 31,32 Use of a flash monitor, but not a CGM, was associated with improved post-prandial glucose, reducing peak glucose and glycaemic variability measures. It is possible that the proactive nature of the flash glucose monitoring system helps individuals' inform meal time bolus insulin dosage decisions, 33 with cumulative usage teaching individuals their "normal" postprandial responses improving future events. At the time of data collection, CGMs were prescribed using NICE Guidelines (NG17) 34 and only offered to individuals with episodes of severe hypoglycaemia or complete hypoglycaemia unawareness. It was therefore used as a safety net to prevent severe hypoglycaemic events in individuals' at risk rather than for proactive management of postprandial glucose control. Despite only being prescribed to individuals who are more likely to spend time in hypoglycaemia, mean time spent <3.9 mmol/L overnight was reduced by CGM, although this did not reach statistical significance.
The improvements in glycaemic control in overnight and postprandial period are similar to our previous findings demonstrating that individuals with a higher residual βcell function display a substantially greater amount of time spent in normoglycaemia in the hours following a bout of moderate-intensity exercise. 21 It is likely that residual βcell function offers some partial protection against dysglycaemia at all time periods throughout a day. This further demonstrates the importance of interventions aiming to preserve βcell function in the recently diagnosed and why they should target maintaining a high C-peptide (>0.2 mmol/mol) secretion, 6 albeit preserving a smaller amount of function likely confers clinical benefits compared with absolute loss. 5 In conclusion, we demonstrate association of residual βcell function with improved free-living glycaemic control in type 1 diabetes overnight and postprandially. The amount of support needed to manage these time periods may be divergent between those with detectable and undetectable levels of C-peptide. In situations with limited resources, the increased difficulties those with no C-peptide face overnight and after meals could be a way of helping allocate diabetes technology and support. In addition, the assistance of residual βcell function in managing blood glucose may allow tighter more ambitious glucose targets for C-peptide positive type 1 diabetes individuals.