1. IntroductionAccording to the National Institute for Health and Care Excellence (NICE) and the American Diabetes Association, metformin is the first-line initial drug monotherapy of choice for patients with type 2 diabetes (T2D) [1American Diabetes Association 7. Approaches to Glycemic Treatment., ]. There are a wide range of glucose-lowering therapies as possible second-line agents to add to metformin if glycaemic targets are not reached. When dual therapy does not achieve optimal glycaemic targets, tripe therapy or insulin therapy is recommended, the latter is considered the most effective treatment to improve glycaemic control [[3]Khunti K. Damci T. Meneghini L. Pan C.Y. Yale J.-F. Study of Once Daily Levemir (SOLVETM): insights into the timing of insulin initiation in people with poorly controlled type 2 diabetes in routine clinical practice.]. Insulin requiring occurs around 5–10 years from diagnosis [[3]Khunti K. Damci T. Meneghini L. Pan C.Y. Yale J.-F. Study of Once Daily Levemir (SOLVETM): insights into the timing of insulin initiation in people with poorly controlled type 2 diabetes in routine clinical practice.], characterised physiologically as increased peripheral insulin resistance and/or when beta cell function has fallen to 15–20 % of normal function [[4]Insulin secretagogues: old and new.]. However, the time from becoming potentially insulin-requiring to initiation of insulin treatment do not often coincide. The reasons for this are poorly understood and likely to be multifactorial including clinical inertia and patient factors. One patient factor includes depression (referred here as diagnosis of depressive disorder and depressive symptoms) which is twice as common than in the general population [[5]Ali S. Stone M.A. Peters J.L. Davies M.J. Khunti K. The prevalence of co-morbid depression in adults with Type 2 diabetes: a systematic review and meta-analysis.] and is associated with reduced diabetes self-care [[6]Nanayakkara N. Pease A. Ranasinha S. Wischer N. Andrikopoulos S. Speight J. de Courten B. Zoungas S. Depression and diabetes distress in adults with type 2 diabetes: results from the Australian National Diabetes Audit (ANDA) 2016.], diabetes complications [[7]Pouwer F. Nefs G. Nouwen A. Adverse effects of depression on glycemic control and health outcomes in people with diabetes.] and premature mortality [[8]Ismail K. Winkley K. Stahl D. Chalder T. Edmonds M. A cohort study of people with diabetes and their first foot ulcer.].In cross sectional studies, depressive symptoms are also associated with negative insulin beliefs [[9]Depression and dysfunctional beliefs. Predictors of negative appraisal of insulin treatment.] and longer duration in acquiring skills to inject [[10]Dzida G. Karnieli E. Svendsen A.L. Sølje K.S. Hermanns N. Depressive symptoms prior to and following insulin initiation in patients with type 2 diabetes mellitus: prevalence, risk factors and effect on physician resource utilisation.]. Two cohort studies observed that depressive symptoms were not associated with longer time to insulin initiation [11

I. M.M, N. G, T. G.S, E. B, M. K, G. M, P.F.A.O.-I.M.M. O, http://orcid.org/0000-0001-9954-171X, M.M. Iversen, G. Nefs, G.S. Tell, B. Espehaug, K. Midthjell, M. Graue, F. Pouwer, Anxiety, depression and timing of insulin treatment among people with type 2 diabetes: Nine-year follow-up of the Nord-Trondelag Health Study, Norway, J. Psychosom. Res. 79, 2015, pp. 309–315. 〈http://dx.doi.org/10.1016/j.jpsychores.2015.07.004〉.

, 12Nefs G. Pop V.J.M. Denollet J. Pouwer F. The longitudinal association between depressive symptoms and initiation of insulin therapy in people with type 2 diabetes in primary care.]. Nevertheless, both studies recruited predominantly white people, which may represent a bias in terms of representativeness. Also, neither study recruited newly diagnosed people with T2Dnor measured insulin requiring status as a variable.

There are no studies that examine both the association between depressive symptoms and time to becoming potentially insulin-requiring and to insulin initiation. Detecting depression could identify people with T2D at high risk for requiring insulin and who may need additional support during initiation of insulin therapy hence the clinical relevance of ascertaining the association.

We hypothesized that presence of depressive symptoms at the time of T2D diagnosis was associated with shorter duration to a) potentially insulin-requiring status and b) insulin initiation compared to those who did not have depressive symptoms, after adjusting for known potential confounders such as age, gender, ethnicity, body mass index, HbA1c, macrovascular and microvascular complications.

3. ResultsThe flowchart of the study is presented in Fig. 1. Of the original SOUL-D cohort (n = 1735), 1663 people provided informed consent for medical record data to be screened for 20 years. From October 2017-December 2020 (representing the window for the 8-year follow-up), 1588 medical records were screened, 381 were inactive (n = 105 of these were traced and medical records obtained), and 75 records were not obtainable. In total 1166 (out of 1663, 70.1 %) records were available for analysis.

Fig. 1Flow of participants in SOUL-D follow-up study.

3.1 Original SOUL-D versus 8-year follow-up SOUL-D characteristicsThe baseline characteristics of the original cohort compared to those who had no follow-up data is shown in Table 1. Participants with no follow-up data were younger and had poorer glycaemic control at baseline. Other variables showed no significant difference between groups.

Table 1Baseline characteristics of original SOUL-D cohort versus 8-year follow-up sample.

Data are n (%) or mean (SD). a There are missing data for some variables in the original SOUL-D cohort, in the 8-year follow-up sample and in the records with no follow- up data, resulting in different percentages, the total number of individuals for each variable is therefore given. b Records with no follow- up data vs. 8- year follow- up data. Continuous variables were measured with T tests and qualitative with χ2 tests.

3.2 Eight- year follow-up SOUL-D characteristics

For the 8-year follow-up SOUL-D study (n = 1166), the median follow-up time was 92 months (IQR = 80–103). By follow-up, 84 people had died. Reasons for death include unspecified cardiovascular disease (n = 4), myocardial infarction (n = 8), sepsis (n = 1), cancer (n = 22), liver failure (n = 3), respiratory failure/pneumonia (n = 5), and information unknown from medical records (n = 39). The mean HbA1c at follow-up was 57.74 (18.42) mmol/mol with 61.1 % (n = 713) of the sample with optimal HbA1c (<58 mmol/mol) compared to 32.8 % (n = 382) with suboptimal HbA1c (≥ 58 mmol/mol).

3.3 Insulin requiring status

On average, the time take to become insulin requiring was 14.67 (15.53) months after the 2nd OAD was prescribed. Thirty-two observations were excluded from analysis because insulin-requiring occurring on or before the original SOUL-D study data collection (baseline), resulting in 1132 observations for analysis. Fewer people had been insulin requiring (n = 380) than those who had not been insulin-requiring (n = 752) during the follow-up period. The overall median time to insulin-requiring status was 84 months (IQR 63–100).

A Kaplan-Meier survival estimate curve found a steady increase in insulin-requiring events up to 120 months after which the number of events then remained static (Fig. 2).

Fig. 2Kaplan Meier survival estimate of insulin requiring events.

3.4 Survival distribution for insulin-requiring status by depressive symptoms: non-parametric comparisonOn average, those who had depressive symptoms at baseline required insulin earlier (n = 163, mean [SD] 73.64 [32.16]) than those who did not have depressive symptoms (n = 958, 79.05 [29.07]) by 5.41 months, (Fig. 3). A log-rank test observed a statistically significant difference in survival distribution between the two groups, χ2(1) = 7.17, p = 0.007.

Fig. 3Survival distribution of insulin-requiring status according to presence of depressive symptoms at time of diagnosis.

3.5 Cox regression: insulin-requiring status

A univariable Cox regression with depressive symptoms (continuous) as predictor found that depressive symptoms are significantly related to time to insulin requiring status (HR = 1.43 [1.10–1.85], p = 0.01).

In the second, step potential confounders (age, gender, ethnicity, body mass index, HbA1c, macrovascular and microvascular complications) were added to the model. The effect of depressive symptoms was reduced and not significant anymore (HR = 1.16 [0.86–1.57], p = 0.34). Two independent variables made a unique statistically significant contribution to this model, age and HbA1c (Table 2). People with younger age (HR = 0.97 [0.96–0.98], p 

Table 2Cox regression analysis for insulin requiring status and insulin initiation.

3.6 Insulin initiation

Forty observations were excluded from the analysis because insulin initiation had already occurred at or before baseline, providing 1127 observations for analyses. Fewer people had been prescribed insulin (n = 109) than those who had not been prescribed insulin during the follow-up period (n = 1016). Most starting on intermediate-acting insulin (NPH; n = 38). The median time to insulin initiation was 93 months (IQR = 79–105).

A Kaplan-Meier survival estimate curve (Fig. 4) observed that after 120 months there was a more rapid increase in insulin initiation events, and this the number of events remains static.

Fig. 4Kaplan Meier survival estimate of insulin initiation.

3.7 Survival distribution for insulin initiation by depressive symptoms: non-parametric comparisonOn average, those who had depressive symptoms at baseline initiated insulin earlier (n = 163, 82.53 [30.19]) than those who did not have depressive symptoms (n = 953, 89.72 [22.02]) by about 7 months (Fig. 5). A log-rank test found a statistically significant difference in survival distribution between the two groups, χ2(1) = 25.64, p 

Fig. 5Survival distribution of insulin initiation according to presence of depressive symptoms at the time of diagnosis.

3.8 Cox regression: insulin initiation

A univariable Cox regression with depressive symptoms (continuous) as predictor showed that depressive symptoms are not significantly related to time to insulin initiation (HR=1.00, 95 % CI=0.99–1.00, p = 0.44).

In the second, step potential confounders (age, gender, ethnicity, body mass index, HbA1c, macrovascular and microvascular complications) were added to the model. The effect of depressive symptoms was reduced and not significant anymore (HR = 1.00 [0.99–1.00], p = 0.49). Three independent variables made a unique statistically significant contribution to this model, age, HbA1c and macrovascular complications (Table 2). People of younger age (HR = 0.95 [0.93–0.97], p 4. Discussion

This was a prospective cohort study of an 8-year follow-up of an incident cohort of people with T2D.

The median time to insulin-requiring status was around 7 years from baseline data collection. People with depressive symptoms at baseline had a significantly shorter time to insulin-requiring status than people without depressive symptoms. However, baseline depressive symptoms were not associated with time to insulin-requiring status after adjusting for age, gender, ethnicity, body mass index, HbA1c, macrovascular and microvascular complications. In this adjusted analysis, factors associated with time to insulin-requiring status younger age and higher HbA1c.

The median time to starting insulin therapy in this cohort was nearly 8 years from baseline data collection. People with depressive symptoms at baseline had a significantly shorter time to insulin initiation than people without depressive symptoms. However, baseline depressive symptoms were not associated with time to insulin initiation after adjusting for age, gender, ethnicity, body mass index, HbA1c, macrovascular and microvascular complications. In an adjusted analysis, factors associated with shorter time to insulin initiation were younger age, higher HbA1c, and presence of macrovascular complications.

This study is in keeping with two previous cohort studies that found no association between depressive symptoms and time to insulin initiation [11

I. M.M, N. G, T. G.S, E. B, M. K, G. M, P.F.A.O.-I.M.M. O, http://orcid.org/0000-0001-9954-171X, M.M. Iversen, G. Nefs, G.S. Tell, B. Espehaug, K. Midthjell, M. Graue, F. Pouwer, Anxiety, depression and timing of insulin treatment among people with type 2 diabetes: Nine-year follow-up of the Nord-Trondelag Health Study, Norway, J. Psychosom. Res. 79, 2015, pp. 309–315. 〈http://dx.doi.org/10.1016/j.jpsychores.2015.07.004〉.

, 12Nefs G. Pop V.J.M. Denollet J. Pouwer F. The longitudinal association between depressive symptoms and initiation of insulin therapy in people with type 2 diabetes in primary care.] Our research added to these previous studies by examining this association with a multi-ethnic cohort and recruiting participants within 6 months of T2D diagnosis. In addition, we included an additional outcome variable, insulin-requiring status.Older people with T2D are often given higher glycemic targets due to the increased risk of hypoglycemia which can be associated with cognitive impairment, falls, and hospital admissions [[23]Sinclair A.J. Abdelhafiz A.H. Forbes A. Munshi M. Evidence-based diabetes care for older people with Type 2 diabetes: a critical review.] In addition, some older people may have a functional limitation which means they are unable to self-administer insulin [[24]Yakaryılmaz F.D. Öztürk Z.A. Treatment of type 2 diabetes mellitus in the elderly.] These factors may contribute to older people with T2D being less likely to be prescribed insulin therapy, which is consistent with our findings of younger age being associated with shorter time to insulin initiation in this study. In addition, younger age was associated with insulin-requiring status. This is in line with previous observations that people diagnosed in older age have better glycemic levels than those diagnosed at a younger age [[25]Management of diabetes mellitus in older people with comorbidities.] For example, in the long-term follow-up of the TODAY study set the US, more than 50 % of participants diagnosed with T2D were adolescents and followed up for 10 years had hyperglycaemia and developed microvascular disease [[26]Long-Term Complications in Youth-Onset Type 2 Diabetes.].NICE guidelines recommend starting insulin when despite dual OAD therapy, HbA1c is above 58 mmol/mol [], to prevent long-term diabetes complications. The findings of this study appear to suggest that NICE guidance is followed as those with higher HbA1c at T2D diagnosis initiated insulin sooner.

The main difficulty encountered during the study is that not all participants from the original SOUL-D cohort were followed-up. This study relied in clinical records and as such real reasons why patients who could not been traced and appeared as “inactive” are not known.

Patients that could not be followed were significantly older and had increased HbA1c levels than those who continued in the study. Significant differences in mean baseline values do not necessarily suggest there is a difference in the association between variables, as adjusting for age and HbA1c in the Cox regression accounts for this imbalance [28Detecting and correcting attrition bias in longitudinal family research., 29Gustavson K. von Soest T. Karevold E. Røysamb E. Attrition and generalizability in longitudinal studies: findings from a 15-year population-based study and a Monte Carlo simulation study.].The point at which a person was insulin-requiring might not be the exact point as it was based on the available HbA1c values from the participants’ medical record. For example, for some, it could have been 6 months since the last HbA1c test and therefore they could have been insulin-requiring at any point over that period. In addition, we assume the person with T2D self-administered the 2nd OAD as prescribed. This means the true time to insulin-requiring status could not be determined. The definition of insulin-requiring status for this analysis was based on NICE guidelines (2 OADs plus HbA1c >58 mmol/mol), however, NICE guidelines also recommend setting individualised targets based on other factors such as comorbidities and personal preferences. This might account for why a proportion of people were coded as ‘insulin-requiring’ (according to this definition) but did not initiate insulin. Furthermore, this definition of insulin-requiring may not be directly transferable to other international guidance, such as for the joint American Diabetes Association (ADA)/European Association for the Study of Diabetes (EASD) [[30]Davies M.J. D’Alessio D.A. Fradkin J. Kernan W.N. Mathieu C. Mingrone G. Rossing P. Tsapas A. Wexler D.J. Buse J.B. Management of hyperglycemia in Type 2 Diabetes, 2018. A consensus Report by the American Diabetes Association (ADA) and the European Association for the Study of Diabetes (EASD).]. However, despite this, there is parity across guidance when HbA1c remains out of target within 6 months of dual or triple combinations [[31]Hanefeld M. Fleischmann H. Siegmund T. Seufert J. Rationale for timely insulin therapy in Type 2 diabetes within the framework of individualised treatment: 2020 update.].

The date of first insulin prescription was the best prospective measure of insulin initiation, eliminating potential recall bias in retrospective patient accounts. However, this might not be the true point of insulin initiation. Baseline data collection was up to 6 months from T2D diagnosis, therefore, participants who were prescribed insulin before baseline data collection were excluded from the analysis, reducing insulin initiation events in our analysis.

This analysis did not examine adherence or persistence to insulin therapy. Previous research on OAD adherence suggests that more than 50 % of people do not take their medication as recommended [[32]Donnan P.T. MacDonald T.M. Morris A.D. Adherence to prescribed oral hypoglycaemic medication in a population of patients with Type 2 diabetes: a retrospective cohort study.]. Therefore, when glycaemic targets are not being met (because of not taking prescribed treatment) insulin treatment may be being prescribed prematurely [[33]Rozenfeld Y. Hunt J.S. Plauschinat C. Wong K.S. Oral antidiabetic medication adherence and glycemic control in managed care.]. Future research with T2D cohorts could compare prescription databases with medical records to help identify adherence issues with insulin therapy.There is a possibility that age and macrovascular complications were colinear with depressive symptoms as opposed to being completely independent predictor variables. In this cohort, baseline depressive symptoms were associated with baseline and 2-year follow-up macrovascular complications [[34]Ismail K. Moulton C.D. Winkley K. Pickup J.C. Thomas S.M. Sherwood R.A. Stahl D. Amiel S.A. The association of depressive symptoms and diabetes distress with glycaemic control and diabetes complications over 2 years in newly diagnosed type 2 diabetes: a prospective cohort study.] which indicates the need to initiate insulin sooner. Different types of study designs could be used to test this in the future such as mediational analyses.We did not adjust for depressive symptoms at the point of insulin-requiring status or insulin initiation as the latter were routinely collected data. However, depressive symptoms tend to persist over time in people with T2D [[35]de Groot M. Kushnick M. Doyle T. Merrill J. McGlynn M. Shubrook J. Schwartz F. Depression among adults with diabetes: prevalence, impact, and treatment options.].

The strength of the study is its design. A prospective cohort design allows for an estimation of time to an event minimising the risk of recall bias of the independent variable. We had sufficient power to detect statistical differences. This is supported by assessments of the confidence intervals which are not higher than 1.57 in the upper bound for insulin requiring status. Furthermore, our cohort was derived from an ethnically diverse target population, while previous studies have been in predominantly white populations.

Younger age and higher HbA1c predicted shorter time to insulin-requiring status and insulin initiation. Future research to test whether modifying depressive symptoms in young people and those with higher HbA1c at time of diagnosis may delay insulin requiring. Further research should examine whether depressive symptoms reduce after insulin initiation.