Association of hypoglycemia problem-solving abilities with severe hypoglycemia in adults with type 1 diabetes: a Poisson regression analysis

Abstract

Background

Severe hypoglycemia (SH) poses a significant challenge in the management of type 1 diabetes (T1D); however, the factors that offer protection other than diabetes technologies are under-studied. The primary objective of this study was to examine the association between hypoglycemia problem-solving (HPS) abilities and severe hypoglycemic events in adults with T1D using Poisson regression analysis.

Methods

In this cross-sectional study, 287 adults with T1D (mean age: 50.3 ± 14.5 years, male: 36.2%, diabetes duration: 17.5 ± 11.2 years, mean HbA1c: 7.7 ± 0.9%) were included and categorized into two groups: non-SH (n = 262) and SH (n = 25). Data on diabetic complications, the hypoglycemia problem-solving scale (HPSS), and treatment details were collected. Impaired awareness of hypoglycemia (IAH) was evaluated using Gold’s method. Univariate and multivariable Poisson regression models were used for the analysis, and the findings were presented as incidence rate ratios (IRRs) at 95% confidence interval (CI).

Results

The incidence of SH was 16.7 (95% CI 7.5–26.0) per 100 person-years. In the univariate Poisson regression analysis, findings revealed associations between IAH, diabetic peripheral neuropathy (DPN), and HPSS1. On the other hand, the multivariate Poisson regression analysis, utilizing stepwise variable selection, identified DPN (IRR: 4.65, 95% CI 1.96–11.04; P < 0.001) and HPSS1 score (IRR: 0.51, 95% CI 0.34, 0.76; P = 0.001) as factors significantly associated with SH.

Conclusion

We identified HPS abilities, in addition to DPN, were associated with SH in adults with T1D.

Trial registration

University Hospital Medical Information Network (UMIN) Center: UMIN000039475), approval date: February 13, 2020.

Background

Severe hypoglycemia (SH) poses a significant problem in the management of type 1 diabetes (T1D). The mortality rate of SH in children and adolescents with T1D is 4–10% [1]. Repeated dropping of glucose levels below 3 mmol/L can retard the body’s natural warning signs, resulting in impaired awareness of hypoglycemia (IAH) and increase the risk of SH by 3- to 6-folds [2]. Notably, recurrent SH contributes to cognitive dysfunction characterized by severe memory impairment and reduced cognitive processing speed in adults with diabetes [3, 4].

Among 71 adults with T1D, recurrent SH was associated with two factors- a fear of hypoglycemia and diabetic peripheral neuropathy (DPN) [5]. Despite significant advancements in therapeutic options, such as continuous glucose monitoring (CGM) and continuous subcutaneous insulin infusion (CSII) [6], recurrent SH remains a concern for patients with T1D. However, our understanding on the protective and risk factors associated with SH in adults with T1D is limited [7].

Recently, researchers have used a validated questionnaire, the hypoglycemia problem-solving scale (HPSS) that comprises 24 items and 7 subscales: problem-solving perception, detection control, identifying problem attributes, setting problem-solving goals, seeking preventive strategies, evaluating strategies, and immediate management to measure patients’ problem-solving ability in terms of hypoglycemia [8]. Meanwhile, due to the rarity of SH events, even in patients with T1D, a careful evaluation is required when analyzing such occurrences. Poisson regression models have been employed in pediatric cases of T1D [9, 10]; however, data on Japanese adults with T1D are lacking.

Therefore, the study investigated the relationship between hypoglycemic problem-solving ability and SH events in adults with T1D using Poisson regression as a valuable tool for analyzing infrequent events.

Materials and methods

This is a cross-sectional exploratory study and was approved by the National Hospital Organization (NHO) Central Research Ethics Committee (R2-0117002). From February 2020 to March 2022, adults with T1D from seven centers of NHO collaboration in Japan were enrolled. Individuals aged ≥ 20 years, with T1D for more than a year, and received care from one of the collaborating centers were included in the study. Participants were recruited on a continuous basis, and we did not employ a fixed number method at each center. Diabetic retinopathy, diabetic nephropathy, and DPN were diagnosed by certified diabetologists based on the treatment guidelines for diabetes. The diagnosis of DPN was based on meeting at least two of the following criteria: subjective symptoms in the lower limbs or feet, loss of ankle jerk reflex, and decreased vibration perception measured bilaterally at the medial malleoli using a C128 tuning fork [11]. The coefficient of variation of the R-R interval (CV-RR) was calculated, and a CV-RR < 3% was considered indicative of diabetic cardiac autonomic neuropathy [12]. The mean corrected QT (QTc) interval was determined using Bazett’s formula, and a QTc interval > 440 ms was classified as prolonged. Data on levels of HbA1c, glycated hemoglobin (GA), and liver enzymes and lipid profiles were obtained from the patients’ medical records. The GA/HbA1c ratio, which reflects glucose variability, was also calculated [13].

SH was defined as an event requiring assistance from another individual to actively administer carbohydrates or glucagon or to take corrective actions [14]. The Gold’s method was used to determine IAH [15]. The Gold Score is a single question (“Do you know when your hypos are commencing?”) rated on a 7-point Likert scale, from 1 = ‘always aware’ to 7 = ‘never aware.’ On the Gold method, a score of 4 or more implied IAH. Diabetes-related distress was assessed using the Problem Areas in Diabetes (PAID) questionnaire, with a cutoff score ≥ 40, indicating a high level of distress [16]. HPSS abilities were assessed using the HPSS [8, 17], which consists of 24 items and 7 subscales; HPSS1: problem-solving perception, HPSS2: detection control, HPSS3: identifying problem attributes, HPSS4: setting problem-solving goals, HPSS5: seeking preventive strategies, HPSS6: evaluating strategies, and HPSS7: immediate management. The HPSS exhibited exceptional internal consistency, as evidenced by a Cronbach’s alpha coefficient of 0.883 [18]. Respondents used a 5-point Likert-type scale, ranging from 0 to 4, to evaluate the relevance of each item with respect to their thoughts or behaviors during a hypoglycemic episode (https://researchmap.jp/yobouigaku). The scale ranges from “Not at all true of me” to “Extremely true of me,” with higher scores indicating an elevated problem-solving aptitude. A mean score of ≧ 3 denoted significant proficiency in addressing hypoglycemic problem-solving abilities.

Data on lifestyle behaviors (current smoking, regular exercise, dietary habits, drinking habits, and sleeping habits) were collected using a standardized questionnaire from the specific health check and guidance system [19]. Healthy lifestyle behaviors encompass a range of practices, such as incorporating fruit, fish, and milk into one’s diet, engaging in regular exercise, refraining from smoking, maintaining moderate alcohol consumption, and ensuring sufficient sleep duration [20].

An estimated sample size of 211 was determined for the Poisson regression, considering power = 0.95, alpha = 0.05, lambda = 0.05, and upper limit = 6. Qualitative variables were assessed using Fisher’s exact test, and quantitative variables were assessed using unpaired t-test and Mann–Whitney U test, depending on whether or not they follow a normal distribution, respectively. The accuracy of SH prediction was assessed by calculating the area under the receiver operating characteristic curve (AUC) to identify the sensitivity and specificity of specific cutoff values. The diagnostic accuracy based on AUC was categorized as follows: < 0.5 (indicating a non-useful test), 0.5–0.6 (classified as poor), 0.6–0.7 (considered adequate), 0.7–0.8 (deemed good), 0.8–0.9 (regarded as very good), and 0.9–1.0 (characterized as excellent). Poisson regression models were used, and the findings were presented as incidence rate ratios (IRRs) at 95% confidence intervals (CIs). Statistical significance was set at P < 0.05. When p < 0.10, and if p < 0.10, we included the effect size using Cohen Cramer’s rule. The guidelines are the same as for the equivalent phi value i.e., 0.10 represents a small effect, 0.30 represents a medium effect and 0.50 represents a large effect [21]. Patients with missing data were excluded from the analysis. Statistical analyses were performed using R version 4.1.2 and Stata software version 16.

Results

This study enrolled 287 adults with T1D and IAH, with a mean age of 50.3 ± 14.5 years. The recruitment numbers of participants at each facility were 135, 93, 30, 12. 11, 3, and 3, in descending order of recruitment volume. Among them, 36.2% were males with diabetes for 17.6 ± 11.2 years on average. The mean HbA1c level was 7.7 ± 0.9%. The participants were categorized into two groups: the non-SH group (n = 262, 91.3%) and SH group (n = 25, 8.7%). The overall SH rate was 16.7/100 person-years (Fig. 1).

Fig. 1.

Distribution of severe hypoglycemia based on the frequency of severe hypoglycemic episodes

Compared with the non-SH group (Table 1), the SH group exhibited a higher prevalence of retinopathy, laser photocoagulation, DPN, IAH, and a CV-RR < 3%. The prevalence of CGM use (small effect size = 0.10), real-time CGM (rtCGM) use (small effect size = 0.11), mecobalamin treatment (small effect size = 0.13), and late-night dinners (small effect size = 0.12) were also higher for the SH group. The distribution of gender was comparable across the groups, showing no significant difference.

Table 1.

Clinical characteristics of study participants according to the SH category

Variables	SH categories		P value
	Non-SH	SH	(Effect size)
	(n = 262)	(n = 25)
Age, years	50.2 (14.4)	51.7 (15.9)	0.630
Male sex, %	36.3	36.0	> 0.999
Diabetes duration, years	17.2 (11.1)	21.0 (11.9)	0.111
BMI, kg/m2	23.3 (3.5)	24.2 (5.2)	0.235
Retinopathy and laser photocoagulation, %	9.5	24.0	0.039 *
Nephropathy, %	17.6	20.8	0.780
Diabetic peripheral neuropathy, %	11.9	47.6	< 0.001*
IAH, %	11.1	28.0	0.024*
HbA1c, %	7.7 (0.9)	7.7 (1.0)	0.673
GA/HbA1c	2.9 (0.3)	2.9 (0.4)	0.359
Alanine aminotransferase, IU/L	20.2 (7.3)	21.8 (8.4)	0.304
Aspartate aminotransferase, IU/L	17.3 (10.9)	18.6 (8.4)	0.554
Gamma-glutamyl transferase, IU/L	21.4 (18.5)	23.8 (24.6)	0.549
Low-density lipoprotein cholesterol, mg/dL	112.3 (27.3)	108.4 (27.2)	0.490
High-density lipoprotein cholesterol, mg/dL	78.8 (20.1)	80.8 (23.8)	0.639
Triglycerides, mg/dL	92.7 (54.9)	103.9 (91.8)	0.366
QTc > 440 ms, %	14.5	12.5	> 0.999
CV-RR < 3%, %	45.5	76.9	0.041*
CSII, %	37.4	28.0	0.393
SAP, %	22.9	8.0	0.124
CGM usage, %	57.6	40.0	0.096 (0.10)
isCGM, %	33.2	32.0	> 0.999
rtCGM, %	24.4	8.0	0.080 (0.11)
TDD / BW	0.6 (0.2)	0.7 (0.2)	0.458
Basal, %	32.7 (14.1)	33.3 (13.3)	0.824
Mecobalamin, %	3.1	12.0	0.060 (0.13)

IAH impaired awareness of hypoglycaemia, GA glycated albumina, QTc QT-correcteda, CV-RR coefficient of variation of the R-R interval, CSII continuous subcutaneous insulin infusion, SAP sensor-augmented pump, isCGM intermittently scanned continuous glucose monitoring, rtCGM real-time continuous glucose monitoring, TDD total daily dose, BW body weight

Mean (standard deviation) or P value (effect size). *P < 0.05. An effect size of 0.30 indicates a small effect, while 0.50 suggests a medium effect, and 0.50 represents a large effect

No significant difference in the total score on the PAID scale or severe distress rate was observed between the two groups (Table 2). The optimal cutoff value for problem-solving perception in predicting SH was 3.8, with a sensitivity of 40.2% and specificity of 92.0% (AUC: 0.693) (Table 3). However, the accuracy of other items in the HPSS was poor with AUC < 0.6.

Table 2.

Lifestyle, diabetes distress, and hypoglycemic problem-solving ability of study participants according to the SH category

Variables	SH categories		P value
	Non-SH	SH	（Effect size）
	(n = 262)	(n = 25)
Late-night dinner eating, %	25.0	44.0	0.056 (0.12)
Exercise habits, %	31.2	36.0	0.655
Healthy lifestyle score, points	4.3 (1.4)	4.3 (1.3)	0.989
PAID, points	30.0 (19.2)	34.9 (20.7)	0.225
Severe distress (≥ 40 points), %	32.2	48.0	0.123
HPSS, points	55.9 (16.3)	54.6 (14.4)	0.701
HPSS1: Problem-solving perception	3.4 (0.7)	2.9 (1.0)	< 0.001*
HPSS2: Detection control	2.5 (1.3)	2.7 (0.9)	0.370
HPSS3: Identifying problem attributes	2.1 (1.2)	2.4 (1.0)	0.322
HPSS4: Setting problem-solving goals	1.8 (1.1)	2.0 (0.8)	0.294
HPSS5: Seeking preventive strategies	1.5 (1.0)	1.6 (0.9)	0.826
HPSS6: Evaluating strategies	2.5 (1.0)	2.3 (0.8)	0.372
HPSS7: Immediate management	2.9 (1.0)	3.2 (0.9)	0.168

PAID Problem Areas in Diabetes, HPSS Hypoglycemia problem-solving scale

^*P < 0.05. An effect size of 0.30 indicates a small effect, while 0.50 suggests a medium effect, and 0.50 represents a large effect

Table 3.

Mean score, area under the curve, cutoff points, sensitivity, and specificity for each item on the hypoglycemic problem-solving scale

Item	≧ 3 a, %	AUC	95% CI	Cut-off point	Sensitivity	Specificity
HPSS1: Problem-solving perception	78.5	0.693	0.594–0.791	3.8	0.402	0.920
HPSS2: Detection control	46.5	0.537	0.433–0.640	1.5	0.196	0.958
HPSS3: Identifying problem attributes	29.2	0.552	0.444–0.661	1.0	0.173	1.000
HPSS4: Setting problem-solving goals	20.3	0.576	0.480–0.671	1.3	0.294	0.917
HPSS5: Seeking preventive strategies	9.8	0.518	0.407–0.628	1.0	0.286	0.833
HPSS6: Evaluating strategies	40.9	0.573	0.459–0.686	2.8	0.366	0.792
HPSS7: Immediate management	59.1	0.581	0.460–0.701	4.0	0.706	0.458

AUC area under the curve, CI confidence interval, HPSS Hypoglycemia problem-solving scale

^aA mean score of ≧ 3 was defined as a high hypoglycemic problem-solving ability

In the univariate Poisson regression analysis, findings revealed associations between IAH, diabetic peripheral neuropathy (DPN), and HPSS1. On the other hand, the multivariate Poisson regression analysis, utilizing stepwise variable selection, identified DPN (IRR: 4.65, 95% CI 1.96–11.04; P < 0.001) and HPSS1 score (IRR: 0.51, 95% CI 0.34, 0.76; P = 0.001) as factors significantly associated with SH (Table 4).

Table 4.

Variables of interest for predicting severe hypoglycemia using Poisson regression analysis

Variables	Univariate Poisson regression		Multivariate Poisson regression
	Incident rate ratio (95% CI)	P value	Incident rate ratio (95% CI)	P value
IAH	2.71 (1.13, 6.49)	0.025*	–	–
DPN	5.24 (2.23, 12.35)	< 0.001*	4.65 (1.96, 11.04)	< 0.001*
HPSS 1: Problem-solving perception	0.52 (0.36, 0.76)	0.001*	0.51 (0.34, 0.76)	0.001*
rtCGM	0.29 (0.07, 1.23)	0.094	–	–
CSII	0.67 (0.28, 1.61)	0.376	–	–

IAH impaired awareness of hypoglycaemia, DPN diabetic peripheral neuropathy, HPSS hypoglycemia problem-solving scale, rtCGM real-time continuous glucose monitoring, CSII continuous subcutaneous insulin infusion

*P < 0.05

Discussion

Using Poisson regression analysis, this study made a significant contribution by being the first to identify protective factors, such as HPS perception, as well as risk factors, including DPN and IAH, to predict SH in Japanese adults with T1D.

DPN and SH

Numerous cross-sectional and observational studies have consistently demonstrated an association between SH and DPN or cardiac autonomic neuropathy [22, 23]. Conversely, Olsen et al. reported no significant association between IAH and autonomic dysfunction or DPN in adults with T1D [24]. In our study, DPN and an abnormal CV-RR were associated with an increased risk of SH. An abnormal CV-RR is a reliable indicator of autonomic and peripheral neuropathy [25]. Furthermore, animal experiments have reported that repeated episodes of hypoglycemia can induce the development of DPN [26, 27]. Hence, close monitoring of SH is crucial for adults with T1D who exhibit an abnormal CV-RR. To validate these findings, extensive large-scale long-term investigations must be conducted in future.

Diabetes technologies and SH

In the present study, rtCGM use was associated with a decreased risk of SH, but not CGM use. Moreover, the CSII treatment was not associated with a decreased risk of SH. A previous study by Wohland et al. showed that treatment with a CSII reduced the risk of SH [28]. Sensor-augmented insulin pump therapy with predictive low-glucose suspension may be effective in reducing the risk of recurrent SH. Davis et al. reported that CGM use did not prevent recurrent SH [29]. Switching from CGM-to-rtCGM use reduces the risk of SH [30]. Hásková et al. [31] reported that rtCGM was superior to CGM in reducing the risk of hypoglycemia and improving T1D duration and normal hypoglycemia awareness [31]. The use of rtCGM reduced the number of hypoglycemic events in individuals with T1D treated for MDI, IAH, or SH [32]. Further examinations, including hypoglycemic alerts and approaches using trend arrows [33], are required to confirm these findings. The Dexcom G6 became eligible for insurance coverage with the C150 SMBG device code in December 2022. Participants in this study were recruited from seven NHO collaboration centers in Japan between February 2020 and March 2022. The number of rtCGM users was relatively low in this study and one possible reason could be that it was before the period of insurance coverage. Also, there were also few users of sensor-augmented pump (SAP). This could be attributed to both economic conditions and potential related factors. There might be other reasons, such as differences in education levels or social stigma, but we did not investigate those aspects.

Limitations of the study

The strength of this study is that it included a validated self-administered questionnaire and Poisson regression analysis for rare events. HPSS is the established standard in this area, but as of now, only five studies [8, 17, 34–36] have been conducted on HPSS, leading to a lack of a systematic review. Wu et al. introduced HPPS and identified factors (younger age, higher education, type 1 diabetes, and lower negative mood) that influence the ability to avoid hypoglycemia [35]. However, this study has some limitations. This study used a cross-sectional design to make causal inferences. DPN status was estimated based on the presence or absence of this condition; therefore, the severity of DPN was not evaluated. Diabetic cardiovascular autonomic neuropathy (DCAN) is an under-diagnosed cardiovascular complication in patients with diabetes. In an animal model, DCAN was evaluated based on histologic patterns and cardiac nerve density. The QTc interval is affected by other factors such as obesity, arteriosclerotic macroangiopathy, and autonomic nerve function. Nerve conduction studies and sympathetic skin responses are reliable methods of detecting DCAN. Further examinations, including the definite DCAN method, are required to compare DCAN and IAH status. In addition, predictive low-glucose suspend and sensor-augmented pumps (PLGS-SAP) have been shown to assist patients in avoiding hypoglycemia [37]. However, we did not assess the efficacy of PLGS-SAP in our study. Furthermore, it is worth noting that after the completion of our study, the use of CSII with a hybrid closed-loop (HCL) system was initiated in Japan. Therefore, future studies should include investigations involving the HCL system to provide a comprehensive understanding of its effectiveness. The potential influence of social stigma on problem-solving perception cannot be ruled out. While the reasons for other sub-scale of HPSS not being associated with SH remain unclear, further examination with an increased sample size may be necessary.

Implications for practice

Problem solving, which means a self-directed cognitive-behavioral process used by individuals to cope with challenging situations, plays a crucial role in diabetes management as a behavioral strategy. It encompasses mental processes involving the identification, analysis, and problem solving. Among the seven factors examined, problem-solving perceptions accounted for a significant proportion of the variance. This specific sub-scale comprises four reverse items: feelings discouraged by failures to prevent hypoglycemia, experiencing emotional distress or anger due to difficulties in hypoglycemia prevention, worrying about prevention methods without taking action, and low self-esteem. Wu et al. found that implementing an intervention centered around a HPSS program effectively enhanced both HbA1c levels and the ability to manage hypoglycemia in a group of 71 individuals experiencing diabetes-related hypoglycemia [17]. In future, there is a need to develop programs adapted to Japanese culture and healthcare system and conduct research on SH prevention within this program.

Notably, DPN is associated with cognitive impairment in adults with T1D, potentially influencing the perception of hypoglycemia. Implementing interventions based on acceptance and commitment therapy for diabetes-related distress [38] may prove benefits in enhancing patients’ problem-solving perceptions. Furthermore, our study revealed that late-night dinner consumption, which is unhealthy and disrupts circadian rhythms, was associated with an increased risk of recurrent SH. Late-night dinner is linked to smaller breakfast portions and skipping breakfast, and both are independent risk factors for diabetes and cardiovascular diseases [39, 40]. Omitting late-night dinner could potentially lead to hypoglycemia in individuals with well-controlled type 2 diabetes who are receiving basal insulin treatment [41]. However, the relationship between having a late-night dinner and SH in individuals with T1D remains unclear. To confirm these issues, further examination, including the assessment of breakfast and bedtime snack quantities, is required to prevent nocturnal hypoglycemia. Considering these findings, it is advisable to structure educational programs that emphasize enhancing the problem-solving abilities and adopting healthy lifestyles in accordance with the recommendations [42] for adults with T1D and SH.

Conclusions

In clinical practice, it is essential for healthcare professionals specializing in diabetes care to take a proactive approach to inquire about individuals’ experiences during hypoglycemic episodes and enhance their problem-solving abilities to prevent SH.

Funding

The PR-IAH study with funding from the National Hospital Organization Clinical research (NHO) (Grant number: H31-NHO (Endocrinology and Nephrology)-01).

Declarations

Conflict of interest

The authors declare no conflicts of interest associated with this manuscript.

Ethical approval

This study conformed to the standards of the Declaration of Helsinki.

Approval of the research protocol

This study was approved by the National Hospital Organization Central Review Board (NHOCRB/ R2-0117002, R3-0614027, R4-0819005、R5-0519005).

Research involve human and animal

N/A

Informed consent

Informed consent or substitute for it was obtained from all patients for being included in the study.