Table 1

Summary of the 43 studies that were included in the final systematic review

First author, year (ref.)Intervention, brief descriptionN; study durationAge (years); diabetes duration (years)SH ratesHA scoreHbA1c
Educational intervention (before-and-after studies)
 de Zoysa, 2014 (28)DAFNE-HART: Psychotherapeutic group education (motivational interviewing and cognitive behavioral therapy), 6 sessions in patients with IAH and SH.N = 24; 1 year54.4 ± 7.9; 30.7 ± 11.9Decreased median SH from 3 (0–104) to 0 (0–3) events/patient/year. Mean SH 18.96 (27.3) to 0.478 (0.947) episodes/patient/year.9 of 20 (45%) regained HA with Gold/Clarke score <4. Gold score from 5.6 ± 1.4 to 4.5 ± 1.9 (P < 0.029). Clarke score from 5.4 ± 1.2 to 3.8 ± 1.8 (P < 0.001).No change in HbA1c: baseline 7.8 ± 1.2% to end 7.8 ± 1.1% (P = 0.80).
 Jordan, 2013 (4)Tayside insulin management course: Structured group education, 1 day of education per week for 4 consecutive weeks.N = 210; 24 months45 ± 13.9; median 16.5 (0.5–58.7)Decreased proportion with SH (20.3% to 5.4%). Number with ≥1 SH per year was 25 to 4 patients.Decrease in number of patients with HU (34.3 ± 47.8% to 8.6 ± 28%, P < 0.001).Median HbA1c reduction: 8.7% (7.8–9.7) to 8.2% (7.6–8.9) (P < 0.001).
 Hopkins, 2012 (21)DAFNE audit: Structured diabetes group education program, 5-day course in flexible insulin therapy.N = 1,163; 12 months41.5 ± 13.7; 18.0 ± 12.1Decreased mean SH: 1.7 ± 8.5 to 0.6 ± 3.7 episodes/person/year (P < 0.001, n = 539).Improved hypoglycemia recognition in those with HU in 43%, not defined according to scores but no symptoms at BG <54 mg/dL (n = 215).Improved HbA1c: 8.5 ± 1.4% to 8.2 ± 1.3% at 1 year (P < 0.001, n = 639).
 Hernandez, 2008 (29)Self-awareness education on body cues associated with various levels of glycemia. Eight 3-h, biweekly sessions, follow-up study of Hernandez, 2003.N = 29; 18 months50.5 ± 14.6; 26.5 (range 10–47)Decreased mean SH (requiring assistance): 13.33 (17.4) to 7.1 (11.59) (P = 0.478); mean SH (hospitalization): 0.76 (2.19) to 0.19 (0.40) (P = 0.370, n = 23).Improved detection of symptom cues of euglycemia and hypoglycemia.Improved HbA1c: 8.8 ± 1.5% to 8.0 ± 1.5% (P = 0.002, n = 23).
 Kubiak, 2006 (31)IG with hypoglycemia-specific education program (6 lessons, 45 min) vs. CG (conventional education, 2 lessons, 45 min).N = 207; 6 months34.3 ± 12.9 (CG) vs. 37.0 ± 14.1 (IG); 16.2 ± 9.3 (CG) vs. 16.4 ± 10.6 (IG)Decreased mean SH in IG: 1.1 ± 4.5 to 0.1 ± 0.5 (P < 0.01, n = 92); CG: 1.3 ± 4.5 to 0.9 ± 5.3 (P = NS, n = 85) but no difference in SH between groups (P = 0.26).Using modified Gold score: visual analog scale 18.5 cm (0 = total unawareness, 18.5 = perfect awareness), HA remained static in IG (12 ± 4.8 to 11.7 ± 5.1, P = NS) but deteriorated in CG (11.5 ± 5.6 to 10.0 ± 4.7, P < 0.01); between-group P = 0.06.Improved HbA1c in both groups; no difference between IG: 6.8 ± 1.6% to 6.3 ± 0.9% (P < 0.01) and CG: 6.8 ± 1.5% to 6.2 ± 1.3% (P < 0.05); between-group P = 0.67.
 Broers, 2005 (22)Dutch adaptation of BGAT-III (6 weekly 1.5–2 h sessions); group vs. individual setting. Psychoeducational intervention, follow-up study to Broers, 2002.N = 59; 12 months43.8 ± 9.4; 22.7 ± 10.7Decreased SH episodes/year: 7.9 ± 7.5 to 1.7 ± 2.4 in group vs. 6.6 ± 7.6 to 0.3 ± 8.5 in individual BGAT (P = 0.001, n = 26).Improved recognition of hypoglycemia in both groups (27.9 ± 24.6% to 42.1 ± 23.7%) and individual BGAT (35.3 ± 33.7% to 42.4 ± 25.6%) (P = 0.02).No change in HbA1c: 7.3 ± 1.2% to 7.3 ± 1.3% in group vs. 7.2 ± 0.9% to 7.5 ± 1.1% in individual BGAT (P = 0.30).
 Hernandez, 2003 (30)Refer to Hernandez, 2008.N = 29; 6 months50.5 ± 14.6; 23.8 ± 11.0Decreased mean SH requiring assistance: 13.3 ± 17.4 to 9.4 ± 14.8 (P = 0.454, n = 25); requiring hospitalization: 0.8 ± 2.2 to 0.1 ± 0.4 (P = 0.227).No increase in ability to detect hypoglycemia but subjects could identify normal BG more accurately.No change in HbA1c: 8.9 ± 1.5% to 8.6 ± 1.4% (P = 0.074).
 Broers, 2002 (23)Refer to Broers, 2005.N = 59; 6 weeks43.8 ± 9.4; 22.7 ± 10.7No measure of SH but improved decision on when not to drive when BG was low and to raise BG during hypoglycemia.Accuracy index of BG perception increased from 8.7 to 13.9% (P = 0.11); improved recognition of hypoglycemic episodes from 32.1 to 39.2% (P = 0.12).No change in HbA1c: 7.3 ± 1.2% to 7.2 ± 1.2% (P = 0.80, n = 42).
 Fritsche, 2001 (32)Hypoglycemia avoidance for 4 months, raised preprandial and bedtime BG targets, SMBG ≥5 times/day; contacted biweekly for insulin dose adjustment.N = 10; 4 months46 ± 16; 20 ± 10Baseline SH 2.0 ± 0.6 episodes in the 4 months before the study to 0. Reduced frequency of BG <70 mg/dL: 8.4 ± 0.9 to 1.4 ± 0.3 episodes/week (P < 0.001).Improved autonomic and neuroglycopenic symptom scores during hypoglycemic clamp. No change in epinephrine and norepinephrine responses.Increased HbA1c: 6.8 ± 0.3% to 7.7 ± 0.3% (P < 0.0001).
 Cox, 2001 (24)BGAT-II psychoeducational group program, follow-up study of Cox, 1995.N = 73; 12 months38.3 ± 9.1; 19.5 ± 10.5Decreased mean SH episodes/month: 1.6 ± 2.0 to 1.1 ± 2.0 (P < 0.002). Booster intervention did not affect these benefits.Improved percentage detection of low BG from 34 ± 29% to 44 ± 27% (P < 0.005).No change in HbA1c: 10.2 ± 2.0% to 10.2 ± 1.9% (P = NS).
 Dagogo-Jack, 1999 (33)Avoidance of hypoglycemia, 3-year follow-up study of Dagogo-Jack, 1994.N = 4; 3 years26.8 ± 4.7; 15.5 ± 4.4In 2 patients, BG <59 mg/dL was 2.6 ± 0.6% compared with 13.2 ± 1.4% at baseline and 4.8 ± 2.3% at reversal of HU at 1 year. No report on SH.Improvement in neurogenic and neuroglycopenic symptoms score at 1 year postreversal from preintervention.Slight increase in HbA1c: 7.4 ± 0.2% at baseline to 7.6 ± 0.8%.
 Fritsche, 1998 (25)5-day inpatient diabetes education program (DTTP), 25 60-min lessons on flexible insulin therapy, carbohydrate counting, correction and prevention of hypo- and hyperglycemia.N = 54; 3 months33.7 ± 11.7; 11.7 ± 9.3Trend toward lower frequency of low BG <70 mg/dL in patients with repeated SH vs. those with no history of SH.Improved accuracy index of BG estimation in patients with SH but no improvement in the group without SH.Decreased HbA1c: 8.0 ± 0.3% to 7.1 ± 0.2% (P < 0.05).
 Fanelli, 1997 (35)Avoidance of hypoglycemia for 6 months in patients with T1D (8 without diabetic autonomic neuropathy [DAN], 13 with DAN) and 15 subjects without T1D.N = 21 T1D (+15 healthy volunteers); 6 months36.9 ± 4.3; 22 ± 4Decreased frequency of hypoglycemia across all groups ∼20 to ∼2 episodes/patient-month. SH did not occur.Improved autonomic and neuroglycopenic symptoms in all groups. Responses were lower in DAN+ than DAN− patients. Plasma epinephrine responses to hypoglycemia improved in DAN− but not significant in DAN+. Responses remained lower than in subjects without T1D.Increased HbA1c in all groups but remained within therapeutic target: 6.2 ± 0.3% to 6.9 ± 0.2% (P < 0.05).
 Liu, 1996 (36)Avoidance of hypoglycemia with less strict glycemic control and higher BG targets, SMBG 4 times daily with modification of insulin doses.N = 7 T1D (+12 healthy volunteers); 3 months36 ± 3; 18 ± 4No SH during the study period, no baseline rate. Decreased number of episodes of hypoglycemia <54 mg/dL, from mean 4.7 (SE 1.3) to 1.9 (SE 0.5)/patient/week (P < 0.05).Improved symptoms scores for sweating and lack of concentration. Improved GH and epinephrine responses but no changes in glucagon, norepinephrine, and cortisol.Increased HbA1c: 6.9 ± 0.3% to 8.0 ± 0.3% (P < 0.05).
 Cox, 1995 (26)BGAT-II, refer to Cox, 2001.N = 100; 6 months38.2 ± 9.0; 19.3 ± 10.4Improved low BG index (predictor of future SH occurrence) in patients with HA but not significant for HU group. No report of SH.Better accuracy in detecting BG fluctuations and low BG levels. Those with reduced HA had improved detection of low BG.Not available.
 Davis, 1994 (27)Conventional insulin therapy vs. intensive insulin therapy.N = 5; 6–10 months27.6 ± 6.6; 9.6 ± 6.2Frequency of hypoglycemia during conventional therapy vs. intensive insulin therapy was 0.6 vs. 2.2 episodes/patient/week based on symptomatic events or BG <59 mg/dL reported to clinician.Reduction in total hypoglycemia symptom scores with intensive insulin therapy, with no reversal on reinstitution of conventional therapy. Lower plasma glucose to stimulate plasma epinephrine secretion during intensive therapy compared with initial conventional therapy without complete reversal on reinstitution of conventional therapy.HbA1c in conventional group: 9.5 ± 1.2% vs. 6.6 ± 0.2% in intensive insulin-treated group (P < 0.0001), then to 9.7 ± 0.9% with conventional treatment (P < 0.005).
 Dagogo-Jack, 1994 (34)Refer to Dagogo-Jack, 1999. Original group of 18 patients (6 HA, 6 HU, 6 healthy volunteers).N = 12 T1D (+6 healthy); 3 months26.5 ± 4.2; 16.9 ± 3.6Proportion of BG <59 mg/dL decreased from 13.2 ± 1.4% to 4.8 ± 2.3% (P < 0.02).Increase in total neurogenic and neuroglycopenic symptoms score responses to hypoglycemia. No significant increases in neuroendocrine responses (epinephrine, pancreatic polypeptide, glucagon, GH, and cortisol) after intervention.Increase in HbA1c: 7.4 ± 0.2% to 7.9 ± 0.2% (P < 0.001) after 3 months of avoidance of hypoglycemia in HU group.
 Cranston, 1994 (9)Avoidance of hypoglycemia for 3 weeks in 2 groups of patients: A) good glycemic control (HbA1c <7%) and B) poor glycemic control (HbA1c 8.2% with wide fluctuations of BG).N = 12; 4 monthsAge range 28–55; duration range 11–32 yearsFrequency of hypoglycemia fell from 21 per month in group A and 14 per month in group B to none in either group.Improved symptoms scores after 3 weeks of no hypoglycemia. Improved glucose threshold for recognition of hypoglycemia in group A from glucose threshold of 2.3 ± 0.18 to 3.4 ± 0.23 mmol/L (P = 0.0005) and from 2.4 ± 0.25 to 3.3 ± 0.14 mmol/L (P = 0.015) in group B. Improved counterregulatory hormone (adrenaline, noradrenaline, GH) responses.No significant change in HbA1c during intervention period; group A: 6.5 ± 0.17% to 6.9 ± 0.25% (P = 0.32) and group B: 8.2 ± 0.18% to 8.7 ± 0.32% (P = 0.26).
 Fanelli, 1994 (38)Intensive insulin therapy (physiologic insulin replacement and continuous education) with avoidance of hypoglycemia.N = 21 T1D (16 in IG, 5 in CG) (+20 healthy volunteers); 12 months33 ± 2; 11.4 ± 1.8Decreased frequency of hypoglycemia (<72 mg/dL) in IG from 0.52 ± 0.05 to 0.05 episodes/patient/day vs. no decrease in frequency of hypoglycemia in CG. Baseline 9 patients had at least 1 SH during the year before study to no episodes of SH during study.Improvement in autonomic symptoms in IG, glucose threshold for autonomic symptoms at baseline from 2.4 ± 0.06 to 3 ± 0.06 mmol/L at 3 months and maintained at 1 year (P < 0.05). No change in CG. Improved counterregulatory hormones (adrenaline, cortisol, GH) responses in IG maintained at 1-year follow-up, but not normalized to healthy volunteers. No changes in CG.Increased HbA1c in IG but still within target (5.8 ± 0.2% to 6.9 ± 0.1, P < 0.05, n = 16). CG: HbA1c showed no increase over 3 months.
 Fanelli, 1993 (37)Avoidance of hypoglycemia with adjustment of doses of insulin aiming for higher fasting, preprandial, and bedtime BG targets.N = 8 T1D (+12 healthy volunteers); 3 months26 ± 2; 5 ± 0.6Decreased frequency of hypoglycemia from 0.49 ± 0.05 at baseline to 0.04 ± 0.03 episodes/patient/day at 3 months (P < 0.05). Baseline 2 patients had at least 1 SH in the year preceding study to no SH during study.Improved neuroendocrine and symptom responses with no difference in autonomic glycemic thresholds compared with healthy volunteers. Epinephrine responses increased from baseline but still lower than in healthy volunteers.Increased HbA1c: 5.8 ± 0.3% to 6.9 ± 0.2% at 3 months (P < 0.05).
Education intervention (RCT studies)
 Hermanns, 2013 (42)Group education program PRIMAS (IG) vs. DTTP (CG).N = 160 (n = 81 in IG, 79 in CG); 6 months45.4 ± 13.6; 19.5 ± 13Reduction of SH in both groups. SH requiring 3rd party assistance/patient/year: CG: 0.31 ± 1.5 to 0.01 ± 0.1 (P = 0.096) vs. IG: 0.29 ± 0.9 to 0.06 ± 0.2 (P = 0.04); no between-group difference (P = 0.179). SH requiring medical assistance/patient/year: CG: 0.09 ± 0.3 to 0.01 ± 0.1 (P = 0.058) vs. IG: 0.19 ± 0.7 to 0.06 ± 0.2 (P = 0.125) (P = 0.214 for between-group difference).Improvement in HA (modified Clarke score) in both groups: CG: 1.5 ± 1.6 to 1.2 ± 1.3 (P = 0.010) vs. IG: 1.8 ± 1.7 to 1.3 ± 1.2 (P = 0.003) but no between-group difference (P = 0.981).Improved HbA1c in PRIMAS group: 8.3 ± 1.1% to 7.9 ± 1.0% (P = 0.004) vs. no change in CG: 8.1 ± 1.0% to 8.1 ± 1.0% (P = 0.571) (P = 0.012 between groups).
 Hermanns, 2010 (43)HyPOS (IG) vs. standard education (CG), long-term follow-up study of Hermanns, 2007; 85.6% were HU at baseline.N = 164 (n = 84 in HyPOS, n = 80 in CG); 31 months46 ± 12.5; 21.4 ± 10.9Lower incidence of SH in HyPOS vs. CG: 0.1 ± 0.2 vs. 0.2 ± 0.4 episodes/patient/year (P = 0.04); 26.5% of patients had 1 SH episode in CG vs. 12.5% in HyPOS (OR 0.4, 95% CI 0.2–0.9, P = 0.04).Not reported.No difference in glycemic control: CG: 7.3 ± 1.1% vs. HyPOS: 7.1 ± 0.9% (P = 0.18).
 Hermanns, 2007 (44)Refer to Hermanns, 2010.N = 164 (n = 84 in HyPOS, n = 80 in CG); 6 months46 ± 12.5; 21.4 ± 10.9No difference in rates of SH in CG vs. HyPOS group, number of SH episodes/patient/year in CG: 3.6 ± 3.6 to 1.2 ± 2.0 vs. IG: 3.5 ± 3.6 to 0.9 ± 1.9 (P = 0.264); reduced SH in both groups (81.1 to 37.7% vs. 78.3 to 34.8%), no difference between groups (P = 0.119).Greater improvement in HyPOS group on modified Gold score (0–10; 10 = fully HA): CG: 4.3 to 5.3; IG: 4.3 to 6.1 (P = 0.015). Improved detection of low BG and treatment of low BG. Increased intensity of hypoglycemia symptoms scores in HyPOS group.HbA1c improved in CG (7.4 ± 1.1 to 7.1 ± 0.9, P = not reported) and remained unchanged in HyPOS (7.2 ± 0.9 to 7.2 ± 0.8, P = 0.21).
 Schachinger, 2005 (45)Randomized to BGAT–III (IG) vs. physician-guided self-help control intervention (CG).N = 111 (n = 56 in BGAT, n = 55 in CG); 12 months46.4 ± 13.8; 22.9 ± 12.1Reduced frequency of SH (episodes/6 months): BGAT: 1.61 ± 3.49 to 0.13 ± 0.33 vs. CG: 1.76 ± 3.71 to 1.78 ± 4.56 (P = 0.04).Improved recognition of low, high, and overall BG in BGAT vs. CG. Detection of low BG improved in BGAT: 52.7 ± 21.8% to 65.2 ± 25.2% but deteriorated in CG: 53.5 ± 28.0% to 48.0 ± 25.5% (P = 0.005).No change in HbA1c: 6.9% maintained in both groups.
 Cox, 2004 (46)Randomized to SMBG + HAATT (7-week group psychoeducational program vs. SMBG (CG).N = 60 (n = 30 in each group); 18 months38.1 ± 9.3; 13.9 ± 8.5Reduced SH (2.0 to 0.4/subject in HAATT vs. 1.8 to 1.7 in CG, P = 0.03).Improved detection of low BG (52 to 70% for HAATT vs. 58 to 55% in CG, P = 0.005).No change in HbA1c: HAATT group 8.1 to 9.0% and CG 8.0 to 8.1% (P = NS).
 Kinsley, 1999 (47)BGAT vs. cholesterol awareness (CG) in patients enrolled into an intensive diabetes treatment program.N = 47 (n = 25 in BGAT, n = 22 in CG); 4 months34.8 ± 8; 9 ± 3Increased frequency of hypoglycemia BG <3.9 mmol/L in both groups, 0.50 ± 0.08 to 0.68 ± 0.06 episodes/day (P < 0.05) in CG vs. 0.45 ± 0.06 to 0.69 ± 0.07 episodes/day (P < 0.001) in BGAT (P = NS between groups). No data on SH.Increased neurogenic and neuroglycopenic symptom scores but did not differ between CG and BGAT groups before or after 4 months of intensive diabetes therapy. Increased epinephrine response in BGAT group to hypoglycemia.Improved HbA1c in both groups: 9.0 ± 1.1% to 7.8 ± 0.8% (P < 0.001) in CG and 9.1 ± 1.4% to 7.9 ± 1.1% in BGAT (P < 0.001) (P = NS between groups).
 Cox, 1994 (48)Long-term follow-up of BGAT patients with a proportion of patients receiving BGAT booster training.N = 41 (n = 14 in BGAT, n = 14 in BGAT+booster, n = 13 in CG); 4.9 years42.9 ± 3.5; 16.3 ± 2.8BGAT subjects had fewer automobile crashes than control subjects: 15% in BGAT had at least 1 automobile crash vs. 42% in CG. SH not reported.BGAT patients had better estimation of BG levels than control subjects. Percentage low BGs (<50 mg/dL), detected by BGAT+booster, BGAT, and CG was 85, 50, and 43%, respectively (P < 0.02). BGAT+booster was more aware of hypoglycemia than BGAT alone.Improved HbA1c over time: BGAT: 12.3 to 10.2% and CG: 11.4 to 9.9%.
Technological intervention (before-and-after studies)
 Choudhary, 2013 (55)Retrospective audit of RT-CGM use: 33 patients were on CSII before starting CGM, 1 on MDI, 1 converted to CSII within 2 months of starting CGM.N = 35; 12 months43.2 ± 12.4; 29.6 ± 13.6Decreased median SH rate from 4.0 (IQR 0.75–7.25) episodes/patient-year to 0.0 (0.0–1.25, P < 0.001); mean 8.1 ± 13 to 0.6 ± 1.2 episodes/year (P = 0.005).19 patients (54%) reported subjective improvement in awareness, 13 no change, 3 slight worsening in awareness. Paired Gold scores unchanged for 19/34 subjects: 5.0 ± 1.5 vs. 5.0 ± 1.9 (P = 0.67).Improved HbA1c: 8.1 ± 1.2% to 7.8 ± 1.0% (P = 0.007).
 Giménez, 2010 (52)CSII use in patients with >4 nonsevere hypoglycemia events per week (in the last 8 weeks) and >2 SH events in last 2 years.N = 20; 24 months34.0 ± 7.5; 16.2 ± 6.6SH fell from baseline of 1.25 ± 0.44 per subject/year to 0.05 ± 0.22 (P < 0.001).Improved Clarke score, baseline 5.45 ± 1.19 to 1.6 ± 2.03 (P < 0.001). At baseline, 19 subjects were HU according to Clarke test, and at 24 months, 3 of 20 were HU.No change in HbA1c: 6.6 ± 1.0% at baseline to 6.3 ± 0.9%.
 Leinung, 2010 (56)Retrospective study on CGM use on HbA1c and SH rates.N = 104; 2.3 years43.2 ± 12.8; 24.9 ± 12Reduction in SH with OR 0.40 (95% CI 0.24–0.65).64.4% with IAH at baseline not tracked over time.Improved HbA1c: 7.6 ± 1.1% to 7.2 ± 0.8% (P < 0.001).
 Ryan, 2009 (54)CGM use in patients with SH.N = 18; 2 months52.0 ± 2.3; 29.4 ± 2.8SH dropped from 16 episodes in baseline month to 3 during study month when on CGM (P = 0.064).Modified-HYPO score dropped from 857 ± 184 to 366 ± 86 (P = 0.023).No change in HbA1c: 8.4 ± 0.3% to 8.2 ± 0.3%.
 Hübinger, 1991 (53)Patients started on CSII with changes in HA.N = 16; 6 months29.5 ± 9.5; 12 ± 6No SH reported.7 of 16 patients on CSII developed HU after 6 months of CSII.Improved HbA1c in HU group: 8.4 ± 2.3% to 7.7 ± 1.0%; HA group: 8.2 ± 0.9% to 7.8 ± 0.9%.
Technological intervention (RCT studies)
 Little, 2014 (63)HypoCOMPaSS: Optimized MDI vs. CSII with or without RT-CGM in SH (2 × 2 factorial design). All patients received structured diabetes and hypoglycemia education, weekly telephone contact, and monthly clinic visits.N = 96; 6 months48.6 ± 12.2; 28.9 ± 12.3Overall study population, decreased SH from 8.9 ± 13.4 to 0.8 ± 1.8 episodes/patient/year (P < 0.001); no between-group differences.Overall study population decreased Gold score: 5.1 ± 1.1 to 4.1 ± 1.6 (P < 0.001); no between-group differences.No change in HbA1c: 8.2 ± 1.2% to 8.1 ± 1.0%.
 Ly, 2013 (58)SAP + LGS vs. CSII only in patients with HU.N = 95; 6 months18.6 ± 11.8; 11.0 ± 8.9Reduced mean SH in CSII: 1.42 ± 3.05 to 0.54 ± 1.66 (P = 0.346); total SH events: 23.5 to 7. Mean SH in LGS: 1.46 ± 2.18 to 0.62 ± 1.19 (P = 0.076); total events: 20 to 8.Improvement in Clarke score in both groups: CSII: 6.46 ± 1.71 to 5.54 ± 1.71 (P = 0.053); LGS: 6.33 ± 1.72 to 4.17 ± 1.40 (P = 0.002) (P = 0.04 for between-group significance for end HU score). No difference in epinephrine response to hypoglycemia between groups.HbA1c was similar in both groups at baseline and did not change at end of study. CSII: 7.41 ± 0.73% to 7.20 ± 0.57% (P = 0.249); SAP: 7.49 ± 0.61% to 7.33 ± 0.77% (P = 0.266).
 Leelarathna, 2013 (59)HypoCOMPaSS clamp study (refer to Little, 2014).N = 18; 6 months50 ± 9; 34.9 ± 10.8Annualized SH rates were lower during study period: 4 (IQR 0–7) vs. 0 (0–0) (P = 0.001).Decreased Gold scores: baseline 5.2 ± 0.2 vs. 4.3 ± 0.4 (P = 0.009); 7 of 18 participants showed HU reversal. Glucose threshold at which subjects felt hypoglycemic improved: 2.6 ± 0.1 mmol/L at baseline to 3.1 ± 0.2 mmol/L (P = 0.017). Improved autonomic and neuroglycopenic symptoms scores. Improved metanephrine response.No change in HbA1c: 8.1 ± 0.2% baseline vs. 8.2 ± 0.2% (P = 0.66).
 Kovatchev, 2011 (62)SMBG with HHC device providing feedback, randomized to different sequences: 1-2-3 or 2-3-1 (1: routine SMBG, 2: added estimated HbA1c, hypoglycemia risk and glucose variability, 3: estimates of symptoms potentially related to hypoglycemia).N = 120; 12 months39.2 ± 14.4; 20.3 ± 12.9Reduced incidence of symptomatic moderate/SH from 5.72 to 3.74 episodes/person/month (P = 0.019), especially in those with HU (6.44 to 3.71 episodes, P = 0.045).Not reported on follow-up.Improved HbA1c: 8.0 to 7.6%.
 Thomas, 2007 (60)Randomized to optimized MDI (preprandial insulin lispro and pre-evening meal glargine), CSII, or education.N = 21; 6 months43 ± 10; 25 ± 10Further SH was prevented in 5 of 7 (71%) participants in each group (P = 0.06). Incidence of SH was 0.6 (analog), 0.9 (CSII), 3.7 (education) episodes/patient/year.Improved HA: education group: 7 with HU to 2 (P = 0.06); analog group: 7 to 4 (P = 0.25); CSII group: 7 to 3 (P = 0.13). Restoration of HA in 3 analog (43%), 4 CSII (57%), and 5 education (71%) patients.No change in HbA1c in education group: 8.5 ±1.1% to 8.3 ± 1.0% (P = 0.54) vs. improved HbA1c in analog group: 8.6 ± 1.1% to 7.6 ± 0.7% (P = 0.04) vs. improved HbA1c in CSII: 8.5 ± 1.9% to 7.4 ± 1.0% (P = 0.06).
 Kanc, 1998 (61)Randomized crossover study to 2 groups: A) bedtime NPH vs. B) nighttime CSII.N = 14 T1D (+12 healthy volunteers); 4 monthsMedian 31.5 (20–45); median 12.5 (7–20)Episodes with BG <63 mg/dL during the last 6 weeks of both treatment periods was lower in CSII than bedtime NPH group: 16.1 ± 3.1 vs. 23.6 ± 3.3 (P = 0.03). SH outcome not reported.Autonomic symptoms appeared earlier at higher BG levels in CSII than in NPH group: 3.1 ± 0.1 mmol/L vs. 2.8 ± 0.2 mmol/L (P = 0.02). No differences between CSII and NPH for hypoglycemic thresholds for neuroglycopenic symptoms.No differences in end HbA1c between CSII and NPH: 7.2 ± 0.2 vs. 7.1 ± 0.2% (P = 0.2).
Pharmacological intervention (all RCT studies)
 Heller, 2002 (64)Randomized crossover trial, lispro with NPH vs. human soluble insulin (SI) with NPH.N = 13; 4 months in each study arm33 ± 3; 12 ± 2No difference in rates of symptomatic hypoglycemia, no SH in both groups.No significant differences in total symptom scores or counterregulatory hormone responses during hypoglycemia clamp.HbA1c not different between SI (6.6 ± 0.3%) and lispro (6.1 ± 0.2%) and from baseline (6.1 ± 0.3%; P = 0.077).
 Fanelli, 2002 (66)Randomized crossover trial, 2 different insulin regimens: A) split regimen of 4 daily insulin injections (3 bolus plus bedtime NPH) vs. B) mixed regimen of 3 daily insulin injection (3 bolus plus mixed regular insulin and NPH at dinner).N = 22; 4 months in each study arm29 ± 3; 14 ± 2Reduced nocturnal hypoglycemia with split regimen (0.10 ± 0.02 vs. 0.28 ± 0.04 episodes/patient-day for mixed regimen, P = 0.002). No SH in either group.Autonomic symptom scores increased earlier with split regimen than with mixed regimen (BG threshold: 3.0 ± 0.1 mmol/L vs. 2.9 ± 0.1 mmol/L, P = 0.010). Similar neuroglycopenic symptoms threshold in both groups.Better HbA1c with split vs. mixed insulin regimen (7.0 ± 0.11% vs. 7.5 ± 0.15%, P = 0.004).
 Ferguson, 2001 (65)Randomized crossover trial: insulin lispro vs. regular human insulin in patients with HU and history of frequent SH.N = 40; 6 months in each study arm46 ± 11; 25.8 ± 9.8Trend toward lower SH (55 in lispro vs. 84 in regular insulin, P = 0.087).Initial Gold score 4.6 ± 1.8 but no follow-up Gold score postintervention.No differences in HbA1c: 9.3 ± 1.0 in regular insulin vs. 9.1 ± 0.83% in lispro group (P = NS) from 9.0 ± 1.1% at baseline.
 Janssen, 2000 (67)Mix insulin (75% lispro, 25% neutral protamine lispro [NPL] insulin–HM insulin) before meals and NPL insulin at bedtime vs. human regular insulin before meals and NPH at bedtime.N = 35; 5–6 months31.1 ± 8.6; 13.7 ± 8.1No differences in hypoglycemia frequency. SH occurred in 1 patient in each group.HM therapy associated with slightly lower total epinephrine response, and autonomic symptom response occurred at a lower BG level during experimental hypoglycemia.No differences in HbA1c: 7.2 ± 0.5% to 7.2 ± 0.7% (HM) vs. 6.7 ± 0.5% to 6.7 ± 0.6% (regular insulin) (P = 0.5; adjusted for baseline period).
 Chalon, 1999 (68)Propranolol: 20 mg twice a day for first 2 weeks, followed by 30 mg twice a day for the next 2 weeks vs. placebo.N = 16; 4 weeks39.3 ± 3.6; data not availableNo difference in number of hypoglycemic episodes in placebo group (12.6 ± 1.6) vs. propranolol (10.7 ± 1.4) over 4 weeks (P = NS).More sweating in propranolol group during biochemical hypoglycemia compared with placebo.Not reported.
  • CG, control group; GH, growth hormone; HU, hypoglycemic-unaware; IG, intervention group; NS, not significant; OR, odds ratio.