Prevalence and Influencing Factors of Mild Hypoglycaemia in People with Diabetes Type 1 and Type 2 in a Tertiary Care Centre: A Cross-Sectional Study

Severe hypoglycaemia (SH) is a potentially life-threatening condition, consequently its occurrence is an indicator for quality in diabetes care. In current therapeutic regimens SH is a rare event [1,2]. The Cochrane collaboration reviewed about long-acting insulin analogues compared to NPH insulin suggested mild hypoglycaemia (MH) as evaluation criterion for new drugs, as there was no significant difference for rates of SH shown in any of the included trials [3]. Reviews comparing sulphonylureas and glinides are exemplary for a lack of data in established therapeutic regimens, especially concerning MH [4,5]. Consequently the assessment of new drugs and diabetes therapies must consider markers occurring more frequently. In an effort to compare future studies with known therapeutic strategies, well-established regimens have to be reassessed. Another important disadvantage in using MH to assess quality of diabetes care is the inconsistency of definitions used throughout healthcare professionals, especially the threshold for hypoglycaemia differs between 2.8-3.9mmol/l [6-10].


Introduction
Severe hypoglycaemia (SH) is a potentially life-threatening condition, consequently its occurrence is an indicator for quality in diabetes care. In current therapeutic regimens SH is a rare event [1,2]. The Cochrane collaboration reviewed about long-acting insulin analogues compared to NPH insulin suggested mild hypoglycaemia (MH) as evaluation criterion for new drugs, as there was no significant difference for rates of SH shown in any of the included trials [3]. Reviews comparing sulphonylureas and glinides are exemplary for a lack of data in established therapeutic regimens, especially concerning MH [4,5]. Consequently the assessment of new drugs and diabetes therapies must consider markers occurring more frequently.
In an effort to compare future studies with known therapeutic strategies, well-established regimens have to be reassessed. Another important disadvantage in using MH to assess quality of diabetes care is the inconsistency of definitions used throughout healthcare professionals, especially the threshold for hypoglycaemia differs between 2.8-3.9mmol/l [6][7][8][9][10]. The aim of this study was to investigate the prevalence of MH and SH in people with type 1 diabetes mellitus (type 1 DM) and type 2 diabetes mellitus (type 2 DM) and to evaluate MH as a marker for quality of diabetes therapy. We propose a patientoriented definition of MH with occurring symptoms of MH as essential component and present data on the occurrence of MH in different diabetes types and therapies.

Patients and Methods
In 2009, all patients with type 1 or type 2 DM at a University outpatient Department for endocrinology and metabolic diseases in Germany were asked to answer a questionnaire about frequency and circumstances of hypoglycaemia in a crosssectional study. People were excluded if any of the following criteria were met: no medical diabetes treatment, new start of insulin therapy or other antidiabetic drug during the last 12 months as well as women with gestational diabetes.
MH was defined as a condition with symptoms consistent with hypoglycaemia and rapid attenuation after carbohydrate ingestion or a plasma glucose below 2.2mmol/l without any symptoms [11,12]. According to the guidelines of the German Diabetes Association, SH was defined as the necessity for intravenous injection of glucose or glucagon intramuscular [13]. Three to six months prior to the study all participants were instructed to document every hypoglycaemia in their diary. We asked for MH in the last week or in the last 12 months if participants had less frequencies of hypoglycaemia. Episodes of SH were accessed for the last 12 months and in lifetime. Blood glucose self-measurements below 2.2mmol/l without any symptoms were classified as impaired hypoglycaemia awareness.
Clinical and laboratory data such as gender, age, diabetes duration, antihyperglycaemic therapy, body mass index (BMI), glycosylated haemoglobin A 1c (HbA 1c ) and blood pressure were drawn from the digital patient record EMIL® [14] and collected on the day of the survey of the respective patient. HbA 1c was measured using high-performance liquid chromatography (TOSOH-Glykohämoglobin-Analyzer-HLC-723-GHbV;Tosoh, Tokyo, Japan). HbA 1c was adjusted according to the Diabetes Control and Complications Trial (DCCT): HbA 1c divided by the mean normal value 33mmol/mol (5.2%) and multiplied by the mean normal HbA 1c of healthy people of the DCCT, i.e. 32mmol/ mol (5.05%) [15]. Social status was determined by education, highest professional position achieved and household net income [16]. A score ranging from 3 to 21 points was established. Higher score value indicates a higher social status. Participation in structured patient education programmes is part of the standard treatment of all patients with diabetes in Germany [17].

Statistical Analysis
Patient characteristics and their frequency of hypoglycaemia are given by adequate statistical measures (mean, standard deviation, frequency). The frequency of MH was calculated in subgroups and given as events per patient per week. The prevalence of SH was calculated as the proportion of patients with at least one SH during the last year. The event rate of SH during the last year was calculated in defined subgroups.
Confidence intervals (CI) at 95% level were estimated to quantify the precision of prevalence estimates. Linear regression models were fitted for people with type 1 and type 2 DM to estimate the influence of age, diabetes duration, SH, HbA 1c , number of injections per day and insulin dose per day on MH. Associations were considered significant at p≤0.05. Statistical analysis was performed with SPSS for Windows 18.0 (SPSS Inc., Chicago, Il, USA) and R.2.11.0 [18].

Results
A total of 650 participants with type 1 DM (n =162) and type 2 DM (n =488) were interviewed. The characteristics of the enrolled cohort are shown in Table 1. Social status score was 12±4 on average in type 1 DM and 11±3 in type 2 DM. In type 1 DM, 77% (n =124) used an intensified insulin therapy, 3% (n =4) a conventional insulin therapy and 21% (n =34) were treated with an insulin pump therapy. In type 2 DM, 81% (n =393) were treated with and 19% (n =95) without insulin therapy. In people with insulin treated type 2 DM, 45% (n =176) had conventional, 35% (n =137) multiple, 18% (n =69) preprandial insulin therapy as well as 3% (n =11) basal insulin only. In addition to insulin therapy, 38 % were treated with oral agents, such as metformin.  In type 1 DM, frequency of MH decreased with age (ß=-0.24 per year, p=0.007). Duration of diabetes, SH during the last 12 months, HbA 1c , number of injections and insulin dose were not associated with MH. In type 2 DM with insulin therapy, one year longer diabetes duration increased the frequency of MH by 0.20 per year (p<0.001). There was also a significant increase of MH by SH (0.17 per each SH/year, p<0.001) and number of insulin injections (0.13 per each injection, p=0.023). In type 2 DM with oral hypoglycaemic agents, frequency of MH was not associated with age, diabetes duration and HbA 1c (Table 3). 11% of participants with type 1 DM as well as 1% of people with insulin treated type 2 DM measured blood glucose levels below 2.2mmol/l without any symptoms. These participants showed a lower threshold for first symptoms of hypoglycaemia (type 1 DM: 2.6±1.0 vs. 3.4±0.6mmol/l, p<0.001; type 2 DM: 3.0±1.0 vs. 3.7±0.7mmol/l, p=0.032) and monitored their blood glucose more often (type 1 DM: 39±8 vs. 34±8 per week, p=0.013; type 2 DM: 32±4 vs. 22±9 per week, p=0.024) than those with normal awareness of hypoglycaemia symptoms. The frequency of MH and SH is not associated with impaired hypoglycaemia awareness.

Discussion
Recently, the occurrence and frequency of MH, especially in people with type 2 DM, roused more attention due to upcoming GLP1 agonists and DDP4 inhibitors with a reduced risk of hypoglycaemia [19,20]. This potential advantage may lose its relevance in the light of the reported results of well-educated patients with type 2 DM under oral hypoglycaemic therapy with a low prevalence of MH with 0.08 events per week. In addition, the influence of MH on patients´ quality of life seems surprisingly low as Pramming et al. [21] have shown that worries about MH are not correlated with their frequency in people with type 1 DM. Another investigation of Kuniss et al. [22] showed that MH is not associated with increased diabetes-related distress or burden in people with type 1 or type 2 DM.
As the definition of hypoglycaemia significantly influences its prevalence, a consistent definition throughout studies, centres and healthcare professionals is a necessity for comparable data. A threshold between 2.7mmol/l (German Diabetes Association) and ≤3.9mmol/l (American Diabetes Association) is often used to define hypoglycaemia [6][7][8][9][10]. In our study only the half of people with type 2 DM and oral agents measure their blood glucose regularly, so this definition is inappropriate. Therefore, we defined hypoglycaemia on the basis of typical symptoms like sweating, loss of concentration or feeling shaky. Different criteria are necessary for different types of diabetes or different therapy regimes [23].
The rates of MH among our patients with type 1 DM correspond to the findings of one other study from Germany (1.6 per patient per week, both) [24] and are only slightly lower than the rates reported in an earlier European multi-centre study which used similar methods (1.8 per patient per week) [25]. Another prospective single-centre study in a Scottish cohort reported half as much incidents of MH (0.8 per patient per week) with a higher HbA 1c of 8.5% in patients with type 1 DM [26]. In people with insulin treated type 2 DM our results correspond to the data from the Scottish study (0.3 per patient per week, both) and are only slightly lower than the reported rates of MH in the German and European studies (0.4-0. 8 [27]. Considering the characteristics of the participants there is much difference to other studies which impairs comparability. Our participants had longer diabetes duration, were older, had lower HbA 1c and were assigned to structured education programmes. The criterion of third party assistance is the most probable reason for a tenfold lower prevalence of SH in our cohort in comparison to studies also based on recalled hypoglycaemia (type 1 DM: 0.1 vs. 0.7-1.0; type 2 DM with insulin: 0.03 vs. 0.1-0.5) [24,25]. Still, the ratio in the frequency of SH in type 2 DM versus type 1 DM of one to three is identical to the numbers reported in other studies [24][25][26].
We used a simple, straightforward item to assess hypoglycaemia unawareness, because there is no gold standard and even in the three standardised questionnaires there is a wide spectrum of prevalence (26-63%) [28]. The frequency of MH was not associated with impaired hypoglycaemia awareness in our patients. The value of HbA 1c to ensure appropriate treatment and prevent long term complications is largely agreed on <7.5% in patients with type 1 DM and 6.5%-7.5% in type 2 DM without SH [10,14,15,29,30]. There is ongoing debate whether focusing on HbA 1c alone may be sufficient to describe treatment quality. The position statement of the DDG working group for measurement of HbA 1c as well as Müller Wieland et al. [31] make mention of the considerable variability of HbA 1c measurements depending on the method in spite of the standardisation of the International Federation of Clinical Chemistry and Laboratory Medicine [31,32]. In our study HbA 1c was measured using highperformance liquid chromatography (official normal range 25-41mmol/mol (4.4-5.9%); mean 33mmol/mol (5.2%); TOSOH-Glykohämoglobin-Analyzer-HLC-723-GHbV; Tosoh, Tokyo, Japan). The normal range was found to have shifted in 2010 after retesting the normal range in 150 healthy volunteers without diabetes. This was confirmed by a mean HbA 1c of 38mmol/ mol (5.65±0.38%) in 1079 people without diabetes in 2009 [33]. This could have caused overtreatment in some patients in consequence of more MH. Additionally our study shows that the frequency of MH rises in people with insulin-treated type 2 DM with time since diagnosis and number of injections, regardless of HbA 1c levels or insulin dose, which might also be a hint for overtreatment.
Normal HbA 1c levels rise with age also in people without diabetes [34,35], whereas it gives no specific target values for elderly and also long time complications lose their relevance. So it seems to be more appropriate to reduce impairment in quality of life by avoiding hypoglycaemic events. Here, the frequency of MH as clinical and patient-oriented target value is the superior marker. The present study was conducted in a tertiary care centre in a University outpatient Department in a cross-sectional

005
setting. There are specific limitations to that design, such as recall bias. However, one study recording the frequency of MH in the same cohort prospectively as well as retrospectively indicates that recall bias in retrospective data collection of hypoglycaemia in the last week is not significant [24].

Conclusion
For the first time, estimates are provided for the frequency of MH in several common diabetic therapeutic regimens in people with type 1 DM and type 2 DM who have participated in a structured treatment and teaching programme.
Considering all the aspects above, we suggest to incorporate the systematic assessment of MH into clinical routine. Furthermore, consideration of the frequency of MH as an additional marker of diabetic treatment quality should find its way in diabetes treatment guidelines and disease management programmes. Frequency of MH may improve monitoring and evaluation of diabetes therapies when HbA 1c alone might not provide an adequate target value.