The relationship between glycated haemoglobin and blood glucose‐lowering treatment trajectories in type 2 diabetes: The Fremantle Diabetes Study Phase II

To examine the relationships between glycaemia and treatment complexity over 6 years in well‐characterized community‐based people with type 2 diabetes.


| INTRODUCTION
Several studies have used analyses of glycaemic trajectories to categorize people with type 2 diabetes and to explore their associated characteristics.A recent systematic review found that most of these studies reported a large subgroup of people with a relatively low and stable glycated haemoglobin (HbA1c) concentration over time, but that other subgroups with less optimal glycaemic control were more variable in number and proportional representation. 1 These latter subgroups could have sustained high HbA1c levels, improving control, deteriorating control, or hybrid temporal trends.
One of the variables considered to be an important determinant of HbA1c trajectory subgroups, especially those representing adverse glycaemic exposure, is diabetes duration. 1This reflects the progressive pancreatic beta-cell failure that characterizes type 2 diabetes 2 and leads to the need for intensification of blood glucose-lowering therapy (BGLT) including progression to insulin treatment.However, individuals with longer-duration type 2 diabetes may have poor outcomes with tight glycaemic control, 3,4 and consequent therapeutic rationalization 5 may be a factor in determining patterns of glycaemia.Of relevance to this point, some studies have examined the relationship between HbA1c trajectory subgroups and BGLT.A relatively consistent finding has been the association between insulin therapy and subgroups of participants with adverse trajectories, [6][7][8] but no study has assessed the temporal associations between glycaemic and therapeutic intensification in type 2 diabetes.
Based on the need to consider both diabetes duration and changes in BGLT in identifying and understanding glycaemic trajectory subgroups, the aim of the present study was to examine the relationship between changes in glycaemic control and treatment complexity over the course of 6 years in a well-characterized cohort of community-based people with type 2 diabetes.

| Study site, sample and approvals
The Fremantle Diabetes Study Phase II (FDS2) is a prospective, observational study of diabetes in a postcode-defined urban community of approximately 157 000 people surrounding the port of Fremantle in the state of Western Australia (WA).Sample characteristics, including classification of diabetes type and details of those identified but not recruited, have been described previously. 9Between 2008 and 2011, 4639 people with known diabetes were identified and 1732 were recruited, including 64 FDS Phase I participants (enrolment   1993-1996) who were no longer resident in the catchment area.Of these, 1551 were diagnosed with type 2 diabetes on clinical grounds.

The South Metropolitan Area Health Service Human Research Ethics
Committee approved FDS2 (reference 07/397) and written informed consent was obtained in each case.

| Clinical assessment
All participants underwent a detailed clinical assessment (history, physical examination and biochemical tests) at entry and then biennially, with multiple attempts to contact participants who did not attend for follow-up to minimize attrition. 9Questionnaires and interviews covered healthcare utilization, medical conditions, medication use, and socioeconomic, demographic and lifestyle data.A physical examination was performed by trained nurses according to a standard protocol.A body shape index (ABSI) was calculated as a more robust index of visceral obesity than body mass index (BMI). 10Fasting blood and urine samples were taken for biochemical tests performed in a single nationally accredited laboratory. 11ronic complications were identified using standard criteria. 12morbidities in FDS2 participants were ascertained from clinical assessments, complemented by access to WA hospital morbidity data through the WA Data Linkage System 13

| Statistical analysis
The computer packages IBM SPSS Statistics 28 (IBM Corporation, Armonk, New York) and StataSE 15 (StataCorp LP, College Station, Texas) were used for statistical analysis.Data from baseline (Year 0), Year 2, Year 4 and Year 6 assessments were used.Data are presented as proportions, mean ± SD, geometric mean (SD range), or, in the case of variables which did not conform to a normal or log e -normal distribution, median and interquartile range (IQR).For independent samples, two-sample comparisons were made using Fisher's exact test for proportions, Student's t-test for normally distributed variables, and the Mann-Whitney U-test for nonparametric variables.Comparisons among multiple groups for categorical variables were made using Fisher-Freeman-Halton exact or chi-squared tests, for normally or log e -normally distributed continuous variables by one-way analysis of variance (ANOVA), and for variables not conforming to normal or log e -normal distribution by Kruskal-Wallis test.Where the overall trend for these multiple comparisons was statistically significant, post hoc Bonferroni-corrected pairwise comparisons were performed.A two-tailed significance level of P < 0.05 was used throughout.

| Trajectory subgroups
Group-based multi-trajectory modelling was used to identify combined HbA1c/treatment trajectory subgroups. 14Multi-trajectory modelling identifies latent clusters of individuals following similar trajectories across multiple indicators of an outcome of interest.Censored normal models were used to estimate the combined trajectories over 6 years (four biennial assessments).To assist model selection, the Bayesian information criterion (BIC) was used to determine the optimum number of subgroups and their functional form. 14

| Characteristics of trajectory subgroups
The bivariable characteristics of the trajectory subgroups were determined and multinomial regression was used to identify independent associates of subgroup membership.Clinically relevant and biologically plausible variables were considered for entry into the multinomial models if bivariable P < 0.10, with backward conditional variable selection and variable entry and exit set at P < 0.05 and P ≥ 0.05, respectively.Loss to follow-up was quantified for each diabetes duration interval by trajectory.If the magnitude of dropout differed by trajectory subgroup, it was adjusted for in the final multinomial models.

| Trajectory subgroup selection and evaluation
For Group 2, the optimum number was four (Figure 2): "low" (zero order HbA1c/treatment trajectories; n = 84 [22.0%]); and "low medium"; "high medium" and "high" (all zero order/linear HbA1c/ treatment trajectories; n = 138, 96 and 64, respectively).For Group 3, the optimum number was four (Figure 3): "low", "low medium", "high medium" and "high' (all zero order HbA1c/treatment trajectories; n = 57, 146, 122 and 32, respectively).For each of the three duration groups, the observed and predicted mean HbA1c values were comparable for each trajectory subgroup at each timepoint, the estimated group probabilities and the proportion assigned to each subgroup according to the maximum posterior probability assignment rule were similar (Supporting Information Tables).The average posterior probabilities were ≥0.888, ≥0.976 and ≥0.925 for Groups 1, 2 and 3, respectively, and the odds of correct classification were ≥ 18.3, ≥62.2 and ≥54.7 (Supporting Information Tables).The rates of attrition during follow-up were not significantly different among the three duration groups (Supporting Information Tables).

| Characteristics of trajectory subgroups
In each low trajectory subgroup and regardless of diabetes duration, most participants maintained a flat HbA1c profile well below 53 mmol/mol (7.0%) on lifestyle measures alone or, in the case of Group 3, a single oral agent, throughout the 6 years of follow-up (Figures 1-3).However, representation in this subgroup fell progressively from over one third in Group 1 to one sixth in Group 3.
In the case of medium trajectory subgroups (one in Group 1 and two in Groups 2 and 3; Figures 1-3), these represented the majority of participants, increasing from 53.5% in Group 1 to 74.9% in Group 3. The HbA1c trajectory was a nonlinear (quadratic) increase in the medium subgroup in Group 1, but most participants maintained a level <53 mmol/mol (<7.0%) at Year 6 with a trend towards increasing numbers of oral therapies (Figure 1).The two Group 2 medium subgroups exhibited flat HbA1c profiles, with a mean HbA1c <53 mmol/mol (< 7.0%) in one and 58 mmol/mol (7.5%) in the other, but in the presence of increasingly complex oral therapy, especially in the latter subgroup (Figure 2).The medium subgroups in Group 3 had similar flat HbA1c trajectories, with a mean of approximately 58 mmol/mol 7(.5%) over the 6 years of follow-up (Figure 3), but one subgroup used multiple oral therapies and the other insulin.
In the high trajectory subgroups, which represented approximately one in eight participants, there was a cubic HbA1c trajectory in Group 1 with a nadir at Year 2 before a subsequent increase corresponding with a steady progression through multiple oral therapies to insulin (Figure 1).Despite this therapeutic progression, mean HbA1c levels were close to 80 mmol/mol (9.5%) at Year 6.In the case of Groups 2 and 3, the high subgroups showed flat HbA1c trajectories but at mean levels of just over 64 mmol/ mol (8.0%) and approaching 86 mmol/mol (10.0%), respectively (Figure 3).The vast majority of participants in these groups were insulin-treated throughout.

| Associates of trajectory subgroup membership
There were significant differences in baseline characteristics between the trajectory subgroups in Groups 1, 2 and 3 over a large range of variables (Supporting Information Tables).The results of multinomial regression analysis with the respective low group as reference are shown in Table 1.For Group 1, non-Anglo-Celt ethnic background, and higher BMI and serum triglycerides were significant associates of high subgroup membership.Of these variables, only serum triglyceride level was in the multinomial model for the medium subgroup, but a higher ABSI was also an independent associate.For Group 2, a younger age at diagnosis and lower total serum cholesterol level was associated with both medium subgroups and high subgroup membership, as was retinopathy, albeit not reaching statistical significance in the high-medium subgroup.Longer diabetes duration, higher BMI and higher serum triglyceride level were each associated with high-medium and high subgroups.The Group 2 high subgroup was associated with outpatient clinic/specialist attendance, self-monitoring of blood glucose (SMBG) and a higher urinary albumin: creatinine ratio.For Group 3, participants in all trajectory subgroups were younger at diagnosis than those in the reference low subgroup.Low-medium subgroup membership was associated with a higher plasma C-peptide concentration, highmedium subgroup membership with increased systolic blood pressure and retinopathy, and high subgroup membership with an increased BMI and heart rate.For combinations of two of the three Group 3 subgroups, low-medium and high subgroup membership were associated with non-Anglo-Celt ethnic background, low-medium and high-medium with SMBG, and high-medium and high with outpatient clinic/specialist attendance, although there were trends towards statistical significance in the remaining subgroup for most of these variables.
F I G U R E 1 Mean (solid line) and 95% confidence intervals (dashed lines) for glycated haemoglobin (HbA1c) and treatment trajectories over 6 years for three subgroups of participants in Group 1 (diabetes duration ≤1.0 year; n = 160)."Diet" refers to lifestyle measures (dietary modification and exercise), and "single" and "multiple" to oral blood glucoselowering therapies.
The novel use of combined HbA1c and treatment trajectory modelling in people with type 2 diabetes from a multi-ethnic urban Australian community in the present study provides insight into how these two variables are linked in clinically relevant subgroups of participants.In addition, the demographic, clinical and laboratory features of subgroup membership have implications for management.The key findings are that, for those who are recently diagnosed, non-Anglo Celt ethnicity, a high BMI and hypertriglyceridaemia are strongly associated with rapid progression of glycaemia and the need for insulin therapy.In the intermediate and long diabetes-duration groups, a younger age at diagnosis and higher BMI were independent associates of the most adverse HbA1c trajectories.The persistence of suboptimal glycaemic control in these latter groups despite the use of insulin therapy in the majority of participants raises questions regarding adherence and the adequacy of clinical management despite the fact these people are significantly more likely to be accessing outpatient clinic/ specialist care than those in other trajectory subgroups.
The results of our analysis are consistent with previous use of HbA1c trajectory modelling that has characterized clusters of people with type 2 diabetes.In a recent systematic review, a consistent finding in published studies has been the existence of a group with stable and acceptable glycaemic control (average HbA1c slightly above 53 mmol/mol [7.0%]) that is typically the largest of those identified. 1 Although the three low subgroups in the present study represented approximately one third or less of each group and were not the largest, inclusion of the medium (in Group 1) or low-medium (in Groups 2 and 3) subgroups, which had stable HbA1c trajectories just above a mean of 53 mmol/mol (7.0%), would include the majority (>57%) of participants in each of our three duration-defined groups.In previous studies, groups with a stable HbA1c of between 54 and 64 mmol/mol (7.1% and 8.0%) comprised 13% to 38% of participants. 1 The equivalent subgroups in our analysis would be the high-medium subgroups F I G U R E 2 Mean (solid line) and 95% confidence intervals (dashed lines) for glycated haemoglobin (HbA1c) and treatment trajectories over 6 years for four subgroups of participants in Group 2 (diabetes duration >1.0 to 10.0 years; n = 382)."Diet" refers to lifestyle measures (dietary modification and exercise), and "single" and "multiple" to oral blood glucose-lowering therapies.in Groups 2 and 3 which, consistent with previous studies, included 33.7% and 25.8%, respectively, of participants.Other studies have found groups with stable high or progressively deteriorating HbA1c levels representing up to 15% of participants, similar to the 9.2% to 16.4% of participants who were in the three high subgroups of Groups 1 to 3.
The high subgroup in Group 1 had the most complex glycaemic prolife of all subgroups identified, namely, a biphasic HbA1c trajectory with an initial reduction from a high mean of approximately 75 mmol/ mol (9.0%) to a nadir at Year 2 and then a progressive increase thereafter, despite therapeutic intensification.In a trajectory analysis of newly diagnosed Germans with type 2 diabetes, 16 there was a similar pattern in a subgroup of 12% of patients, although the post-nadir rise in HbA1c was substantially smaller than in the present study.We have previously reported, in analysis of data from the first phase of the FDS initiated 15 years before the FDS2, that people of Southern European ethnicity with type 2 diabetes (from migrant families and the major non-Anglo Celt ethnic subgroup in both FDS phases) exhibited early non-antibody-mediated pancreatic beta-cell failure and early insulin requirement, 17 which would be consistent with the present findings.The fact that the present high subgroup was no more likely than other recently diagnosed FDS2 participants to access outpatient clinic/specialist management despite their poor glycaemic control could indicate that there are cultural, linguistic and/or health literacy barriers to healthcare access 18 with subsequent inadequate management of glycaemic progression, but reluctance to progress, or adhere, to insulin therapy may contribute. 19The relatively high serum triglyceride levels in this subgroup probably reflects their baseline high HbA1c levels 20 but may also relate to their greater BMI. 21r Groups 2 and 3, younger age at diagnosis, BMI and resting heart rate were strongly associated with the most adverse HbA1c trajectories.We were not able to include estimates of beta-cell function and insulin resistance in our multinomial models because of incomplete data (Supporting Information Tables).However, it is possible F I G U R E 3 Mean (solid line) and 95% confidence intervals (dashed lines) for glycated haemoglobin (HbA1c) and treatment trajectories over 6 years for four subgroups of participants in Group 3 (diabetes duration >10.0 years; n = 357)."Diet" refers to lifestyle measures (dietary modification and exercise), and "single" and "multiple" to oral blood glucose-lowering therapies.
T A B L E 1 Multinomial regression models of independent associates of trajectory group membership (reference: low in each case) for Group 1 (n Note: Data presented are odds ratios with 95% confidence intervals.Variables with bivariable P < 0.10 (Supporting Information Tables) were considered for entry.Backward stepwise selection was performed with P < 0.050 for entry and P ≥ 0.050 for removal.Bold values are those with 95% confidence intervals that did not span unity.
a An increase of 1 in log e (serum triglycerides) or in log e (uACR equates) to an increase of 2.72 in serum triglycerides or uACR, respectively.
that, due to the faster loss of beta-cell function in younger individuals with type 2 diabetes, 22 age at diabetes onset may be a surrogate for this variable.Similarly, in the absence of adequate insulin resistance data, BMI may be a surrogate, these observations showing that individuals with the most severe underlying pathophysiology are understandably at risk of both poor glycaemic control and therapeutic progression.The association between resting heart rate and high subgroup in Groups 2 and 3 could reflect the contribution of sympathetic activation to glycaemia, 23 but autonomic dysregulation and dehydration associated with poor glycaemic control may also be involved. 24ere were other expected associations between clinical and laboratory variables and subgroup membership in the two groups with longer diabetes duration in the present study.There were several medium or high subgroups in which the prevalence of retinopathy, use of SMBG and outpatient clinic/specialist access were increased relative to the equivalent low-trajectory subgroups.Nevertheless, continuing poor control in the high subgroups, especially in Group 3 despite the almost universal use of insulin, is of concern.In people in this latter subgroup, there is an argument for less intensive management to avoid hypoglycaemia and an increased risk of death in those with, or at risk of, cardiovascular disease. 3,4An HbA1c target of <64 mmol/mol (<8.0%) has been suggested in this situation. 5This is well below the stable mean HbA1c of close to 86 mmol/mol (10.0%) in this subgroup, a level at which hyperglycaemic symptoms may have a negative effect on quality of life, 25 and the risk of new or worsening microangiopathy remains. 26Although persistent poor glycaemic control despite insulin therapy is likely to reflect factors such as variable patient adherence and healthcare system shortcomings, 27 our data were collected from a cohort studied before the widespread availability of the glucagon-like peptide 1 receptor agonists (GLP-1RAs) and sodium-glucose cotransporter-2 (SGLT2) inhibitors in Australia.These agents may be particularly beneficial for people in the high, largely insulin-treated, subgroups in the present study. 28e association between a low total serum cholesterol and the medium and high subgroups versus the low subgroup in Group 2 probably represents the fact that, compared to the low subgroup, the use of statin therapy was greater (Supporting Information Tables).Statin use was relatively uniform across the subgroups in Group 3 and therefore there was no such difference.The over-representation of non-Anglo Celt participants in the medium and high Group 3 subgroups, especially those with the highest HbA1c trajectory, may, as in Group 1, be a reflection of their progressive pancreatic beta-cell failure but perhaps also of cultural issues relating to health literacy and healthcare access.
The design of the present study, and many of those in which only glycaemic trajectory groups have been identified, 1 was different from those in which distinct clusters of people with subtypes of adult-onset diabetes have been proposed. 29,30In these latter studies, apart from not using HbA1c trajectories in the categorization process, only newly 29 or recently diagnosed 30 participants were included, the subgroups were based on multiple clinical/laboratory variables and endpoints that included biochemical and genetic tests that may not be available in routine clinical care, and follow-up was limited to ≤8 years.
Treatment progression was used to differentiate clusters in one of these studies, 29 but in the form of time to initiation rather than through trajectories as in the present study.In addition, and in contrast to the two cluster-based analyses, 29,30 we excluded participants with LADA and MODY, given that these diagnoses are associated with treatment progression that typically differs from that in true type 2 diabetes.Our data represent a characterization of glycaemic and treatment profiles and their associates across the clinical course of type 2 diabetes rather than a detailed multivariable profiling around the time of diagnosis.
This study has some limitations.Although they were from a representative community-based sample, the FDS2 participants included in the present analyses were those with near-complete relevant data collected over 6 years.As acknowledged, they were younger, more likely to be male, had shorter duration diabetes with better glycaemic control, and were less likely to be treated with insulin than those who were excluded, but these differences were largely minor.In addition, the subgroups identified as a result of trajectory analyses were consistent with previously published studies. 1 The strengths of the present study include the long-term detailed patient-level data, which allowed exploration of variables associated with distinct HbA1c trajectories.
There are various clinical implications of our findings.
BIC values balance model fit with model complexity, and the closer the negative BIC value is to zero the better the fit.Other selection criteria included: (i) adequate numbers of subjects in each subgroup; (ii) distinct trajectories (non-overlapping confidence intervals); (iii) acceptably narrow confidence intervals; (iv) average posterior probabilities of subgroup membership >0.70; (v) odds of correct classification based on posterior probabilities of subgroup membership >5; and (vi) close correspondence between each subgroup's estimated probability and the proportion of participants classified to that subgroup according to the maximum posterior probability assignment rule.
Trajectories were identified separately for those with baseline diabetes duration (i) ≤1.0 year (Group 1, recently diagnosed: median [IQR] 0.6 [0.3-0.8]years; n = 160), (ii) >1.0 to 10.0 years (Group 2, intermediate duration: 5.0 [2.7-7.6]years; n = 382) and (iii) >10.0 years (Group 3, long duration: 16.0 [13.6-20.0]years; n = 357).More complex categorization by duration did not reveal additional clearly identifiable trajectory patterns, including splitting Group 2 into >1.0 to 5.0 years and >5.0 to 10.0 years (Supporting Information Figures).At baseline, members of these three duration groups were significantly different in age (61.5 ± 9.8, 64.0 ± 10.6 and 68.0 ± 9.8 years, respectively, for Groups 1, 2 and 3; ANOVA P < 0.001) but not sex (P = 0.157).The best models for the observed data (those with the lowest BICs), the matrices of the observed and predicted values for mean HbA1c and of observed values for diabetes treatment by trajectory subgroup, the estimated probability and the proportion of study participants classified to each group according to the maximum posterior probability assignment rule for each of the combined HbA1c/diabetes treatment trajectory subgroups, and the average posterior probability and odds of correct classification for each of the combined HbA1c/diabetes treatment trajectory subgroups are shown in the Supporting Information Tables.
First, people with short-duration type 2 diabetes who are from non-Anglo Celt ethnic groups, and have a high BMI and high serum triglyceride levels are likely to progress relatively rapidly to insulin requirement.They should be counselled, using culturally and linguistically appropriate methods,31 on the probable early need for insulin at or soon after diagnosis, and merit close glycaemic monitoring that guides therapeutic intensification.The GLP-1RAs may be particularly useful early therapy in this group because of the potential for weight reduction as well as improved glycaemic control.Second, younger people who are overweight or obese with established type 2 diabetes are more likely to develop poor glycaemic control over time and may not respond adequately to the introduction of insulin therapy.These individuals may also benefit from relatively close monitoring and proactive management including, where appropriate, measures to improve adherence to insulin.32The early introduction of the newer BGLTs, such as GLP-1RAs or SGLT2 inhibitors, before or in combination with insulin should be considered.Indeed, the influence of these recently introduced therapies on glycaemic and treatment trajectories in type 2 diabetes should generate significant interest in future years.for performing laboratory tests.The authors also wish to thank the staff from the Department of Health WA's Data Linkage Services, the Hospital Morbidity Data Collection, and the Western Australian Registry of Births, Deaths and Marriages.Open access publishing facilitated by The University of Western Australia, as part of the Wiley -The University of Western Australia agreement via the Council of Australian University Librarians.
as approved by the WA Department of Health Human Research Ethics Committee.All public/private hospitalizations in WA are recorded in the Hospital Morbidity Data Collection that was established in 1970.The latest linkage has data up to the end of 2022.