s ubjective Symptoms, Blood Glucose Estimation, and Blood Glucose Concentrations in Adolescents With Diabetes ANNE FREUND, M.S., SUZANNE BENNETT JOHNSON, Ph.D., ARLAN ROSENBLOOM, M.D., BARRIE ALEXANDER, Ph.D., AND CAROLYN APPERSON HANSEN, M.S. Twenty-five adolescent campers with insulin-dependent diabetes mellitus (IDDM) completed a Symptom Rating Checklist and estimated their blood glucose (BG) immediately before having their BG assessed four times daily for 11 days. Consistent relationships between BG and symptoms were not identified when the data were analyzed for the group as a whole. However, when each camper's data were analyzed separately, 23 of the 25 adolescents had at least one significant glycemia-symptom (G-S) correlation. Each camper seemed to have a unique G-S pattern; only one symptom (hungry) was significantly related to BG for more than half of the youngsters studied. Almost all of the significant G-S correlations were indicative of low rather than high BG. However, when asked, few campers were able to accurately identify which symptoms were reliably associated with low or high BG. In this study, different measures of BG estimation error led to different results. The percent of estimates ±20% of the actual BG value (55% in this study) was strongly influenced by the actual BG reading because higher BG values have larger accuracy ranges than lower BG concentrations. When estimated BG was simply subtracted from actual BG, under- and overestimates canceled each other out, resulting in an unusually small estimated error (5 mg/dl in this investigation). The absolute difference score ignores the direction of estimation error, but may more accurately reflect patients' average estimation error (68 mg/dl in this study). When actual and estimated BG values were correlated for the group as a whole, the patients appeared to be highly accurate at estimating BG (r = .93, P < .0001). However, when the same correlational analysis was conducted for each individual camper, BG estimation accuracy appeared substantially reduced (mean r = .51) and varied greatly from camper to camper (range r = .06-.80). The type of estimation error was also strongly influenced by the actual BG value. Low (<70 mg/dl) anid middle (70-150 mg/dl) range readings were often overestimated, while higher BG values (>150 mg/dl) were frequently under- estimated. In the sample studied, campers with lower mean BG were better able to correctly identify hypoglycemic symptoms and made more accurate BG estimates than their more poorly controlled peers. However, BG estimation accuracy was unrelated to G-S patterns and to camper awareness of G-S patterns; this suggests that campers were not using internal symptoms to estimate BG. DIABETES CARE 1986; 9:236-43. F or youngsters with diabetes, the goal of treatment is nent of diabetes self-management because symptoms can pro- to maintain blood glucose levels as close to normal vide an important signal to act, e.g., to take a BG reading as possible. Since current treatment methods permit or eat a snack. Consequently, patients with diabetes are taught only a crude approximation of normal pancreatic signs and symptoms that are indicative of abnormally high or function, imperfect attainment of this goal is to be expected. low BG levels. However, little is known about patients' actual The detection of symptoms associated with high and low use of these signs and symptoms to accurately detect hyper- blood glucose concentrations (BG) is an important compo- or hypoglycemia. 236 DIABETES CARE, VOL. 9 NO. 3, MAY-JUNE 1986