the Journal of Applied Research
in Clinical and Experimental Therapeutics

Current Issue

Volume 1 - 2001

Volume 2 - 2002

Volume 3 - 2003

Volume 4 - 2004

Volume 5- 2005

Reprint Information

Back to The Journal of Applied Research

©2000-2005. All Rights Reserved. Therapeutic Solutions LLC

Click here for information on how to order reprints of this article.
the Journal of Applied Research
in Clinical and Experimental Therapeutics

Current Issue

Volume 6 - 2006

Volume 5- 2005

Volume 4 - 2004

Volume 3 - 2003

Volume 2 - 2002

Volume 1 - 2001

Reprint Information

Back to The Journal of Applied Research

©2000-2005. All Rights Reserved. Therapeutic Solutions LLC

Click here for information on how to order reprints of this article.

Quantifying Biobehavioral Determinants of Risk for Severe Hypoglycemia in Type 1 Diabetes*


Boris Kovatcheva

Daniel Coxa

Raina Robevab

Linda Gonder-Fredericka

Leon Farhic

William Clarked


aDepartment of Psychiatric Medicine

cDepartment of Medicine

dDepartment of Pediatrics

University of Virginia Health System

Charlottesville, VA 22908


bDepartment of Mathematical Sciences

Sweet Briar College

Sweet Briar, VA


*This research was supported by grants RO1 DK-51562 and RO1 DK-28288 from the National Institutes of Health and by LifeScan Inc, Milpitas, CA.



This study investigates, through a biobehavioral risk model, the perception, awareness, and self-treatment behavior accompanying hypoglycemia in type 1 diabetes, and their relationship with self-monitoring blood glucose (SMBG) profiles and episodes of severe hypoglycemia (SH).

Ninety-six adults with type 1 diabetes, aged 35 8 years, with a duration of diabetes of 16 10 years, HbA1c 8.6 1.8%, 43 of whom had a recent history of SH, used a hand-held computer over a month to complete 70 behavior-assessment trials followed by SMBG. For the next 6 months all subjects recorded occurrence of SH. Symptom awareness and self-treatment behavior indices accounted for 53% of the risk for hypoglycemia observed in SMBG profiles. Future SH episodes were equally dependent on two factors: symptom awareness and self-treatment/SMBG profile. Combined with subjects' history of SH, these factors explained 52% of future SH.

We conclude that a field assessment of awareness and self-treatment behaviors yields credible estimates of frequency of hypoglycemia and long-term risk for SH, thus providing a basis for behavioral intervention.


Key Words: [severe] hypoglycemia, awareness, self-regulation, self-monitoring of blood glucose



Extensive studies, including the 10-year Diabetes Control and Complications Trial (DCCT)1 and its European counterpart2 have demonstrated that the most effective long‑term control of type 1 diabetes mellitus (T1DM) is the strict maintenance of blood glucose (BG) levels within the normal range. The DCCT proved that chronic high BG levels cause many complications in multiple body systems over time, but, at the same time, intensive insulin therapy resulted in an increased risk of hypoglycemia.3 Without corrective action, hypoglycemia can rapidly progress to severe hypoglycemia (SH), a condition identified as low BG resulting in stupor, seizure, or unconsciousness that precludes self‑ treatment.3 If the patient does not receive treatment during a SH episode, death can occur. It is estimated that approximately 4% of deaths in all T1DM patients are attributed to SH.4 Recently, other negative consequences of SH have been documented: a magnetic resonance imaging (MRI) study reported cases in which occurrences of SH were associated with anatomical changes in the brain5; other studies report permanent cognitive dysfunction associated with SH.6-8 Consequently, hypoglycemia has been identified as the major barrier to improved glycemic control,9,10 and the progression of complications is most rapid for patients in whom glycemic control is worst. In short, patients with T1DM face the life‑long clinical optimization problem of maintaining strict glycemic control without increasing their risk for hypoglycemia. Thus, it is imperative that the biologic and behavioral factors contributing to the occurrence of hypoglycemia be well understood.

In addition to intensive insulin therapy, research has identified many other factors contributing to increased likelihood of hypoglycemia. For example, patients who do not hormonally counterregulate during insulin-induced hypoglycemia were reported to be 250 times more likely to have future SH than those who do hormonally counterregulate,10 and patients who report loss of hypoglycemic symptoms are reported to be three times more likely to have SH in the immediate future.11 All of these risk factors (ie, deficits in counterregulation, hypoglycemic unawareness, and intensive insulin therapy) appear to be associated with hypoglycemia-associated autonomic failure.12 Specifically, episodes of hypoglycemia (<3.9 mmol/L) can cause hormonal counterregulation to be delayed or weakened in response to a low BG event occurring in the next 24 to 48 hours. In this case, only extremely low BG would trigger counterregulation and, when it occurs, epinephrine secretion may be inadequate to keep BG from falling further. Consequently, autonomic symptoms may also be delayed or dampened, making it much more likely that hypoglycemia will be undetected, thus increasing the risk of SH. In general, studies that investigated the occurrence of hypoglycemia-related symptoms found that such warning signs occur and are recognized by patients in less than 50% of all hypoglycemic episodes and are associated with low BG levels (3.9 mmol/L and below).13-16 This means that about half of all hypoglycemic episodes are asymptomatic (ie, unrecognized). In many cases, even if such hypoglycemic episodes are recognized, the patient's BG level may be too low to allow for self-treatment due to neuroglycopenia's interference with cognitive/motor functioning. In addition to symptom awareness/unawareness, the accuracy and the promptness of self-treatment of low BG contributes to the risk of progression of hypoglycemia into SH and would be generally related to the magnitude of BG fluctuations in the low BG range.

In order to describe formally the behavioral precursors to (severe) hypoglycemia, we have developed and validated a biopsychobehavioral model postulating that SH is a consequence of a complex interplay between physiologic, psychologic, and behavioral factors.17-21 The model is based on the concept of stochastic transitions22 that describe sequentially the links between symptom awareness, self-treatment judgments, and behaviors. These assessments are included in the biopsychobehavioral model that contains seven steps, each having a theoretical binary outcome of yes or no. These sequential steps are linked by paths reflecting the increase or decrease of patients' idiosyncratic risk of SH through the steps of the model. We continue this research by presenting a detailed analysis of the biobehavioral sequence: history of SH, symptom awareness, self-treatment behaviors, self-monitoring BG profiles, and long-term risk for future SH. We will demonstrate that long-term risk of SH is linked to autonomic failure, while neuroglycopenia and self-treatment behavior predict short-term BG excursions into hypoglycemia. In order to do so, we will introduce appropriate tools that quantify perception and awareness of autonomic and neuroglycopenic symptoms, self-treatment behavior, and self-monitoring BG profiles using data collected in patients' natural environment.




Ninety-six individuals, 58 women and 38 men, who had T1DM for at least 2 years and were taking insulin since the time of diagnosis were recruited through advertisements in newsletters and diabetic clinics and through direct referrals. All subjects were routinely using self-monitoring devices to measure their BG. Their average age at the time of recruitment was 35 years (standard deviation [SD] = 8), the average duration of diabetes was 16 years (SD = 10) and the average daily insulin dose was 0.58 U/kg (SD = 0.19). Since the goal of this research was to study risk factors for SH, subjects who had problems with recurrent SH were preferentially recruited. History of SH was recorded as the number of SH episodes in the previous year. The preferential recruitment resulted in 43 participants who reported having at least two SH episodes in the previous year (SH group) and 53 who reported none (No SH group). The SH group included 45% of all subjects, which is greater than the estimated 4% to 22% of T1DM patients who have problems with SH.3 Consequently, the incidence of SH in this study was high compared with reports from population-based studies.



Subjects were invited to an orientation meeting in groups of 4 to 10. They were informed about the study, their questions were answered, written informed consent was obtained, and blood samples were collected for determination of HbA1c. Each subject then participated in four consecutive data collections: (1) screening questionnaire, (2) hand-held computer field behavioral assessment, (3) parallel self-monitoring of blood glucose (SMBG) using LifeScan OneTouch Profile meters, and (4) subsequent 6 months of diaries of moderate/SH.

The screening questionnaire included demographic data and history of subjects' diabetes, including detailed description of the SH episodes in the previous year (history of SH).

Hand-held computer (HHC) assessment: Subjects were instructed to use the Psion P-250 (Psion Corp, England) HHC whenever they believed that their BG was low or high, based on internal or external cues, and before their routine SMBG. For each subject, seventy trials were stored over a 3- to 4-week period. At each trial, the HHC collected data on two sets of potential sources of information concerning hypoglycemia status: current symptoms and preceding self-treatment behaviors.

Symptoms: Subjects rated on a scale from 0 to 6 four common autonomic symptoms (sweating, pounding heart, trembling, jittery/tense/stressed feelings) and four common neuroglycopenic symptoms (difficulty concentrating/slow thinking, light-headedness/dizziness, visual disturbance, uncoordinated movements).

Self-treatment behaviors: At each trial, the HHC asked the subjects to recall their previous BG and to rate the subsequent amount of food eaten, insulin taken, and exercise on a three-point scale: more, less, or usual. At the end of each trial, the HHC prompted the subjects to measure BG and enter the reading. Since it requires at least 45 seconds for someone to lance a finger, collect a blood sample, and generate a BG reading, a trial was considered invalid and excluded from the data analysis if less than 45 seconds elapsed between the prompt "measure BG" and the actual BG entry. Six percent of all trials were excluded for this reason, leaving 4111 records for analysis. We have previously demonstrated that the HHC assessment yields very reliable and useful findings.23-25

SMBG profile: Parallel to the HHC assessment, subjects were instructed to measure their BG four times a day using LifeScan OneTouch Profile meters (lifescan Inc, Milpitas, CA). This yielded 135 53 SMBG readings per subject over a month. These data were electronically downloaded and the low BG index of each subject was computed. The low BG index is a statistic based on our previously published transformation of the BG measurements scale (26) and has been repeatedly validated and proven to be the best predictor of SH from SMBG data (27, 28, 29, 30).

Monthly diaries of SH: For the following 6 months, the subjects recorded in diaries any occurrence of SH together with the date and time of the episode. The diaries were mailed in monthly and documented a total of 215 SH episodes (2.4 5.3 per subject).



Biobehavioral model of hypoglycemia (Figure 1): The risk for hypoglycemia in T1DM is a continuous process that is contingent on patients' perception and awareness of hypoglycemic symptoms, quality of judgment of their condition, and accuracy and swiftness of execution of corrective actions. The short-term outcome of this control is reflected by SMBG profiles, while the long-term outcome is reflected by the frequency of events such as SH. To formally describe this process, we introduce a biobehavioral model of hypoglycemia (Figure 1) that consists of sequential steps and appropriate indices developed to quantify each step:

Step 0 is history of SH, assessed by screening questionnaires.

Step 1 quantifies symptom awareness by introducing Autonomic and Neuroglycopenic Symptom Awareness Indices. These indices are summary scores, based on the ratings of autonomic/ neuroglycopenic symptoms during hypoglycemia recorded in 70 HHC trials. Each index is computed as a logistic function of autonomic or neuroglycopenic symptom ratings, optimized with respect to subjects' history of SH. In other words, each index is given by the formula:


Where Z(sx) =a1Sx1+a2Sx2+a3Sx3+a4Sx4+a5 is a linear combination of four autonomic or four neuroglycopenic symptoms, respectively, with coefficients optimized to differentiate SH versus No SH subjects (ie, on the basis of Step 0). Symptom Awareness Indices range between 0 and 100.

Step 2 quantifies self-treatment behaviors pertinent to low BG. We introduce a Self-treatment [Behavior] Index based on patients' HHC reports of changes in insulin, food, and exercise following a specific BG level. Essentially the index is a conditional estimate of patients' self-treatment reaction (change in insulin, food, exercise) to a previous BG level, computed according to a previously published table.18 For example, if a patient's BG is low and subsequently s/he increases her/his insulin dose, or eats less, or exercises more, a "penalty" is imposed. This "penalty" is 1, 2, or 3, depending on the range of the previous BG: 6.2 to 3.9 mmol/L, 3.9 to 6 mmol/L, or below 3 mmol/L, respectively. If the patient takes an appropriate action, the "penalty" is zero.

The self-treatment index is the average "penalty score" across 70 HHC trials, calibrated between 0 and 100, with higher scores indicating inappropriate treatment responses.

Step 3 includes the low BG index, computed from SMBG data, using a previously published formula.28,30 This Index could theoretically range between 0 and 100.

Step 4 is directly accessible by prospective diaries recording the occurrence of moderate/SH together with the date and time of each episode.

The analysis of awareness and self-treatment indices, Low BG Index, history of SH and prospective records of SH yielded the following results:

Retrospectively, the two subject groups, No SH and SH, were differentiated by (1) Autonomic Symptom Awareness Index, t = 3.5, P < .001; (2) Neuroglycopenic Symptom Awareness Index, t = 2.3, P < .05; (3) Self-Treatment Behavior Index, t = -3.6, P < .001, and the Low BG Index, t = -4.2, P < .001. Table 1 presents the means, standard deviations, t and P values for these comparisons.

Prospectively, the Autonomic Symptom Awareness Index correlated with future SH episodes, r = -0.53, P < .001, while the Self-Treatment Behavior Index correlated with subjects' Low BG Index, r = 0.73, P < .001. None of the Symptom Awareness Indices correlated with the Self-Treatment Behavior Index.

A linear regression predicting subjects' low BG index from awareness and self-treatment indices had r2 = 53% and was statistically significant (F = 51, P < .0001). The partial correlations of the Neuroglycopenic Symptom Awareness Index and Self-Treatment Behavior Index were -0.23 and 0.72, respectively; both variables entered significantly into the model.

A linear regression model using Autonomic and Neuroglycopenic Symptom Awareness and Low BG Indices, in combination with history of SH, to predict future SH episodes in the following 6 months had r2=52% and was significant (F = 24, P < .0001). Table 2 presents the partial correlations of the variables in the model together with their significance.

In the set of four variables (Autonomic and Neuroglycopenic Symptom Awareness, Self-Treatment Behavior and Low BG Indices), factor analysis identified two orthogonal factors: Factor 1, associated with Self-Treatment and Low BG Indices (factor loads of 0.93 and 0.91, respectively); and Factor 2, associated with the Autonomic and Neuroglycopenic Symptom Awareness Indices (factor loads of 0.84 and 0.85, respectively). Both factors differentiated SH from No SH subjects, P levels < .01). Most importantly, future SH was almost equally dependent on these two factors (loads of 0.55 and -0.49, respectively). Thus, we can conclude that the risk of SH can be (statistically) presented as a linear combination of the independent Factor 1 and Factor 2 (Figure 2).



We have previously developed and reported a biopsychobehavioral model of SH that presents in detail the possible paths through which a low BG episode could progress into SH.21 We now generalize this model with a formal description of behavioral patterns leading to hypoglycemia and test the model using HHC technology for implementation in the field of our behavioral assessments. The general idea of the formal biobehavioral model of hypoglycemia is that the daily control of T1DM is a continuous process dependent on patients' perception and awareness of hypoglycemic symptoms, quality of judgment of their condition, and accuracy and swiftness of execution of corrective actions. This process has a memory (the history of SH episodes) and receives continuous feedback from SMBG. In the long-term, the quality of hypoglycemic control is reflected by frequency of SH. In the short term, patients can rely on SMBG profiles, such as the Low BG Index. In order to fully understand this process, we need two sets of estimates: one for each of its steps and another for the relationships between its sequential steps. Step 1 and Step 2 of the model (Figure 1) are complex sets of psychobehavioral characteristics that have their own internal structure, which we previously presented. 21 ,22 The sequential relationship between awareness, self-treatment, SMBG profile and future SH is stochastic (eg, a specific combination of symptoms may or may not trigger self-treatment action; an appropriate treatment behavior is a precursor to a safe SMBG profile only with some probability). In addition, knowledge of SBMG readings, or profile, as well as experience of SH results, with some probability, in feedback about patients' awareness or behavior . This feedback may be negative (eg, recurrent hypoglycemia may reduce symptom awareness, leading to a "vicious cycle,"12 or positive (eg, knowledge about the SMBG profile could lead to improvement in self-treatment behavior). Thus, the feed-forward and feedback arrows in Figure 1 mark stochastic transitions22 and are subject to probability modeling and statistical evaluation.



In this article, we give quantitative representations for each step of this model and evaluate the feed-forward relationships of the model. Further investigation and a repeated HHC assessment at the end of the study are needed to evaluate feedback relationships. We conclude that the Autonomic Awareness Index is related to long-term risk for SH, while the Neuroglycopenic Index mostly reflects patients' frequency of exposure to low BG (as estimated by the Low BG Index). In addition, the Low BG Index depends primarily on subjects' self-treatment behaviors. Thus, interventions to improve such behaviors would likely reduce subjects' recurrence of hypoglycemia. The latter is even more important in light of the finding that self-treatment behavior (and the Low BG Index) are, to a large extent, independent of symptom awareness.

Finally, our analyses revealed the internal structure of the biobehavioral determinants of hypoglycemia. The occurrence of future SH episodes was largely (and approximately equally) dependent on two clearly identified linearly independent factors: Symptom Awareness and Self-Treatment Behavior + the resulting SMBG profile (Figure 2). In combination with history of SH, these factors predicted 52% of the variance in future SH episodes. Interestingly, the history of SH that is largely considered to be the single best predictor of future SH3 contributed about 16% to that prediction. The remainder of the variance was accounted for by current characteristics of patients' behavior. The latter gives an optimistic perspective for a behavioral treatment intended to reduce SH in T1DM.



1.         The Diabetes Control and Complications Trial Research Group: The effect of intensive treatment of diabetes on the development and progression of long-term complications of insulin-dependent diabetes mellitus. NEJM 329:978-986, 1993.

2.         Reichard P, Phil M: Mortality and treatment side-effects during long-term intensified conventional insulin treatment in the Stockholm Diabetes Intervention study. Diabetes 43:313-317, 1994.

3.         The Diabetes Control and Complications Trial Research Group: Hypoglycemia in the Diabetes Control and Complications Trial. Diabetes 46:271-286, 1997.

4.         Lorenz R, Siebert C, Cleary P, et al for the DCCT Research Group: Epidemiology of severe hypoglycemia in the DCCT. Diabetes 37:3A, 1988.

5.         Perros P, Deary IJ, Sellar RJ, et al: Brain abnormalities demonstrated by magnetic resonance imaging in adult IDDM patients with and without a history of recurrent severe hypoglycemia. Diabetes Care 20:1013-1018, 1997.

6.         Deary IJ, Crawford JR, Hepburn DA, et al: Severe hypoglycemia and intelligence in adult patients with insulin-treated diabetes. Diabetes 42:341-344, 1993.

7.         Gold AE, Deary IJ, Frier BM: Recurrent severe hypoglycaemia and cognitive function in type I diabetes. Diab Med 10:503-508, 1993.

8.         Lincoln NB, Faleiro RM, Kelly C, et al: Effect of long-term glycemic control on cognitive function. Diabetes Care 19:656-658, 1996.

9.         Cryer PE: Hypoglycemia begets hypoglycemia. Diabetes 42:1691-1693, 1993.

10.     Cryer PE, Fisher JN, Shamoon H: Hypoglycemia. Diabetes Care 17:734-755, 1994.

11.     Gold AE, MacLeod KM, Frier BM: Frequency of severe hypoglycemia in patients with type I diabetes with impaired awareness of hypoglycemia. Diabetes Care 17:697-703, 1994.

12.     Cryer PE: Iatrogenic hypoglycemia as a cause of hypoglycemia induced autonomic failure in IDDM: A vicious cycle. Diabetes 41:255-260, 1992.

13.     Clarke WL, Cox DJ, Gonder-Frederick LA, et al: Reduced awareness of hypoglycemia in IDDM adults: A prospective study of hypoglycemia frequency and associated symptoms. Diabetes Care 18:517-522, 1995.

14.     Clarke WL, Gonder-Frederick LA, Richards FE, Cryer PE: Multifactorial origin of hypoglycemic symptom unawareness in IDDM: Association with defective glucose counterregulation and better glycemic control. Diabetes 40:680-685, 1991.

15.     Cox DJ, Gonder-Frederick LA, Antoun B, et al: Perceived symptoms in the recognition of hypoglycemia. Diabetes Care 16:519-527, 1993.

16.     Cox DJ, Gonder-Frederick LA, Cryer P, Clarke W: Sex differences in BG thresholds for counterregulatory hormone release and low blood glucose symptom perception. Diabetes Care 19:269-270, 1996.

17.     Clarke WL, Gonder-Frederick LA, Cox DJ, et al: The relationship between nonroutine use of insulin, food and exercise predicts the occurrence of hypoglycemia in IDDM adults with varying degrees of hypoglycemic awareness and metabolic control. Diabetes Educator 23:55-58, 1997.

18.     Clarke WL, Cox DJ, Gonder-Frederick LA, et al: The bio-psycho-behavioral model of severe hypoglycemia II: Self-management behaviors. Diabetes Care 22:580-584, 1999.

19.     Cox DJ, Kovatchev BP, Gonder-Frederick LA, et al: Bio-behavioral factors contributing to occurrence of severe hypoglycemia. Diabetes 46(Suppl 1):A269, 1997.

20.     Cox DJ, Gonder-Frederick LA, Kovatchev BP, et al: Bio-psycho-behavioral model of severe hypoglycemia. II. Understanding causes of severe hypoglycemia. Diabetes Care 22:2018-2025, 1999.

21.     Gonder-Frederick L, Cox D, Kovatchev B, et al: A biopsychobehavioral model of risk of severe hypoglycemia. Diabetes Care 20:661-669, 1997.

22.     Kovatchev BP, Cox DJ, Gonder-Frederick LA, et al: Stochastic model of self-regulation decision making exemplified by decisions concerning hypoglycemia. Health Psychol 17:277-284, 1998.

23.     Cox DJ, Gonder-Frederick LA, Schroeder DB, et al: Disruptive effects of acute hypoglycemia on speed of cognitive and motor performance. Diabetes Care 16:1391-1393, 1993.

24.     Driesen NR, Cox DJ, Gonder-Frederick LA, Clarke WL: Reaction time impairment in insulin-dependent diabetes: Task complexity, blood glucose levels and individual differences. Neuropsychology 9:1-9, 1995.

25.     Gonder-Frederick LA, Cox DJ, Driesen NR, et al: Individual differences in neurobehavioral disruption during mild and moderate hypoglycemia in adults with IDDM. Diabetes 43:1407-1412, 1994.

26.     Kovatchev BP, Cox DJ, Gonder-Frederick LA, WL Clarke: Symmetrization of the blood glucose measurement scale and its applications. Diabetes Care 20:1655-1658, 1997.

27.     Cox DJ, Kovatchev B, Julian D, et al: Frequency of severe hypoglycemia in IDDM can be predicted from self-monitoring blood glucose data. JCEM 79:1659-1662, 1994.

28.     Kovatchev BP, Cox DJ, Gonder-Frederick LA, et al: Assessment of risk for severe hypoglycemia among adults with IDDM: Validation of the Low Blood Glucose Index. Diabetes Care 21:1870-1875, 1998.

29.     Kovatchev BP, Straume M, Farhy L, Cox DJ: Estimating the speed of blood glucose transitions and its relationship with severe hypoglycemia. Diabetes 48(Suppl): A363, 1999.

30.     Kovatchev BP, Cox DJ, Farhy LS, et al: Episodes of severe hypoglycemia in type 1 diabetes are preceded, and followed, within 48 hours by measurable disturbances in blood glucose. J Clin Endocrinol Metabol 85:4287-4292, 2000.


Table 1: Comparison of No Severe Hypoglycemia with Severe Hypoglycemia Subjects


No SH Mean (SD)

SH mean (SD)

T (2-tail P)*

Autonomic Symptom Awareness



3.3 (.001)

Neuroglycopenic Symptom Awareness



2.3 (.027)

Self-Treatment Behavior



- 3.6 (<.001)

Low Blood Glucose index



- 4.2 (<.001)

*All t-tests have 94 of freedom.

SD = standard deviation; SH = severe hypoglycemia.

Table 2: Prediction of Future Severe Hypoglycemia (r2 = 52%)

Variable in the Model

Partial Correlation

T (P Value)

History of Severe Hypoglycemia


4.0 (<.001)

Autonomic Symptom Awareness Index

- 0.49

- 5.0 (<.001)

Neuroglycopenic Symptom Awareness Index


2.4 (<.01)

Self-Treatment Behavior Index


0.4, nonsignificant

Low Blood Glucose Index


3.4 (<.001)



Figure 1. Biobehavioral model of hypoglycemia.


Figure 2. Risk for future severe hypoglycemia is determined by two linearly independent factors quantified by the Low Blood Glucose/Treatment Behavior Indices and Autonomic/Neuroglycopenic Symptom Awareness Indices.