Interventions That Restore Awareness of Hypoglycemia in Adults With Type 1 Diabetes: A Systematic Review and Meta-analysis

Hypoglycemia remains the major limiting factor toward achieving good glycemic control (1). Recurrent hypoglycemia reduces symptomatic and hormone responses to subsequent hypoglycemia (2), associated with impaired awareness of hypoglycemia (IAH). IAH occurs in up to one-third of adults with type 1 diabetes (T1D) (3,4), increasing their risk of severe hypoglycemia (SH) sixfold (3) and contributing to substantial morbidity, with implications for employment (5), driving (6), and mortality. Distribution of risk of SH is skewed: one study showed that 5% of subjects accounted for 54% of all SH episodes, with IAH one of the main risk factors (7). “Dead-in-bed,” related to nocturnal hypoglycemia, is a leading cause of death in people with T1D <40 years of age (8).

Although small research studies have shown that meticulous avoidance of hypoglycemia can improve awareness of hypoglycemia (9), achieving this in clinical practice is difficult and hard to sustain. Strategies used include educational approaches, using biopsychosocial or behavioral therapies; technological interventions, such as continuous subcutaneous insulin infusion (CSII), continuous glucose monitoring (CGM), and sensor-augmented pumps (SAP); and pharmacotherapies.

This systematic review assessed the clinical effectiveness of treatment strategies for restoring hypoglycemia awareness (HA) and reducing SH risk in those with IAH and performed a meta-analysis, where possible, for different approaches in restoring awareness in T1D adults. Interventions to restore HA were broadly divided into three categories: educational (inclusive of behavioral), technological, and pharmacotherapeutic.

A systematic literature search in the databases of The Cochrane Library, MEDLINE, Embase, Science Citation Index Expanded, Social Sciences Citation Index, PsycINFO, and CINAHL was performed from inception until 1 October 2014. Additional studies were identified by hand-searching reference lists of included trials and systematic reviews and contacting experts in the field. Search terms and their synonyms used were type 1 diabetes mellitus, hypoglycemia, low glucose, hypoglycemia unawareness, impaired awareness of hypoglycemia, avoidance of hypoglycemia, and awareness (Supplementary Table 1).

The recommendations of the Centre for Reviews and Dissemination for Systematic Reviews (10) were followed. All randomized controlled trials (RCTs), nonrandomized controlled trials, and before-and-after studies that assessed interventions to restore HA were included. Case series and case reports were excluded.

Inclusion criteria were adults (age ≥18) with T1D using accepted criteria (11). Studies must have described HA status at baseline by validated scoring systems such as the Clarke (12) or Gold (13) scores. In studies that did not use these scores, accuracy of blood glucose (BG) estimate was allowed as a surrogate measure of awareness status. Studies performed in people with type 2 diabetes, involving fewer than five participants, those aged <18 years, or who had <1 month follow-up were excluded.

Islet and pancreas transplantations were excluded because intractable recurrent severe hypoglycemia is a proven indication for these interventions (14,15).

Two authors (E.Y. and M.N.) independently assessed abstracts and titles for eligibility and extracted data, with differences in interpretation resolved by a third reviewer (P.C.) and consensus after discussion. Full texts of studies that fulfilled inclusion criteria were obtained and data extracted using a standardized data extraction table. Relevant missing information was sought from article author(s).

Interventions were classified into patient education (including diabetes education classes, psychological interventions, behavioral therapy); use of technology (CSII, CGM sensors, retrospective or real-time [RT]), and pharmacological therapies (insulin analogs and other pharmacological agents). For studies with more than one intervention (e.g., combining education and technology), the technological intervention was considered the primary intervention with a further description of all interventions.

Outcome measures were categorized into SH rates (defined as events requiring external assistance [16]), restoration of HA (Gold [13] or Clarke [12] scores), subjective recognition of low BG by participants or improved autonomic or neuroglycopenic symptoms, responses to hypoglycemia assessed by symptom scores (17), counterregulatory hormone responses, and changes to glycemic control measured by HbA_1c.

To assess the quality of included studies, Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines were used for RCTs (18) and Meta-analysis of Observational Studies in Epidemiology (MOOSE) guidelines for observational studies (19). Instead of a score allocated to each study for quality assessment, we assessed the strength of evidences using the four domains suggested by the Agency for Healthcare Research and Quality guidelines (20): risk of bias, consistency of effect sizes, directness of link between interventions and outcomes, and precision or the certainty of effect in relation to a specific outcome. Additional Agency for Healthcare Research and Quality domains include assessment of a dose-response association, existence of confounders, strength of association, and publication bias. The first three of these are more relevant to observational studies than RCTs. The strength of evidence was based on a global assessment of all of these domains and studies graded as high, moderate, low, or insufficient. A study was considered of high quality if it was a well-conducted RCT, prospective, with a low risk of bias, and accounted for confounders such as age and diabetes duration.

If interventions studied had sufficient data, MedCalc software was used to perform a meta-analysis pooling the standardized mean difference (SMD). If studies did not directly report the mean and SD for change from baseline to 12 months for the outcomes of interest, these were calculated. Where outcomes were measured on different scales, SMDs were combined, where possible. Studies reporting outcomes in a measure that was not suitable for inclusion into the meta-analysis are presented as a summary of findings and analyzed in a narrative synthesis.

In the meta-analysis, heterogeneity was assessed using the I² statistic. Effect sizes were pooled by using fixed-effects and random-effects models. The two models used different assumptions. The former assumes there is one true effect size that is shared by all of the included studies, and the latter, by contrast, assumes that the true effect could vary from study to study.

The database search identified 1,683 abstracts until 1 October 2014 (Fig. 1). Review of titles and abstracts identified 57 full-text articles meeting the inclusion criteria. A further two articles were retrieved from reference lists of included articles, of which one met inclusion criteria. Forty-three studies were included in the final systematic review, summarized in Table 1.

Patient education was the primary intervention in 27 included studies (8 RCTs); 11 (6 RCTs) were based on technology, and 5 (all RCTs) examined pharmacotherapeutics. In four studies combining multiple interventions (e.g., education with technology with/without pharmacotherapy), one intervention was classified as primary and the other the comparator. Studies with long-term follow-up or different outcomes were analyzed separately.

Thirteen studies were conducted in North America and Canada, 10 in the U.K., 19 elsewhere in Europe, and 1 in Australia. Numbers of subjects enrolled were generally <100, with 10 studies involving ≥100 subjects. Follow-up ranged from 1 to 12 months, with long-term follow-up (>1 year) reported for eight studies.

This represented the largest intervention group: 27 relevant studies used an educational approach to restore HA.

There were 20 studies in this category: 8 in unselected populations with T1D (4,21–27) and 12 in participants with IAH at baseline (9,28–38). Eight studies (22,23,24,26,29,30,33,34) were long-term, durations ranging between 1 and 3 years.

In unselected populations with T1D, some educational programs were based on well-established structured education in insulin self-management, such as the German Diabetes Treatment and Training Program (DTTP), designed in Dusseldorf (39) and adopted elsewhere in Germany (25); Dose Adjustment For Normal Eating (DAFNE) (21); and the Tayside insulin management course (4), an adaptation from Bournemouth type 1 intensive education (BERTIE), whereas others were based on psychoeducational programs, such as Blood Glucose Awareness Training (BGAT) (22–24,26), delivered in individual or group settings. All of these approaches, based on adult learning principles, resulted in significant reductions in SH rates of between 15% and 75% in four studies (4,21,22,24), with one showing a trend toward a lower frequency of low BG readings that was not statistically significant (25). BGAT showed improved ability and accuracy index in recognizing symptoms of low BG (22). The DAFNE (21) and Tayside (4) studies were large-scale registry data with more than 1,000 patients, followed up for a year, showing generalizability of the interventions. One year after DAFNE, the rate of IAH had fallen from 39.9% to 33%, with awareness restored in 43% of participants reporting IAH at baseline; the Tayside registry showed improvement in HA in 25% of participants, although in both studies IAH was defined without a known validated score. Glycemic control improved or was maintained at target level, uninfluenced by the method of delivery of the structured education (individual or group setting).

Only one study assessed effect of education on counterregulatory hormone responses (27), comparing hypoglycemia rates in intensively treated (defined as CSII therapy or multiple daily insulin [MDI] injections, 4–7 BG tests, and weekly contact with the treatment team) against conventional insulin therapy (defined as twice-daily insulin injections, 1 to 2 daily BG tests, and monthly clinic visits, in five subjects, four being switched to CSII). This showed that intensive therapy was associated with improved HbA_1c but resulted in a reduction in epinephrine and symptom responses to experimental hypoglycemia, neither of which was fully restored on return to conventional therapy, despite worsening of HbA_1c.

Twelve studies in people with IAH were small: between 5 and 30 people in 11 studies (9,28–30,32–38), although 1 study had >100 people (31). Follow-up was 3–12 months, with longer-term follow-up data of 18 months (29) and 3 years (33) was available for two studies. In seven studies (9,32,34–38), the educational content comprised strict avoidance of hypoglycemia by using dose adjustment, raised glucose targets, and/or intensive professional contact in frequent clinic visits and regular (even daily) telephone contact. All studies were able significantly to reduce/eliminate SH and improve awareness, providing proof of principle of reversibility of IAH. Although autonomic and neuroglycopenic symptom scores improved to levels seen in control subjects without diabetes, counterregulatory hormone responses did not improve in two of seven studies (32,34). Dagogo-Jack et al. (33) looked at the sustainability of HA restoration at 3 years, monitoring four of six original participants. Reversal of hypoglycemia unawareness was sustained beyond the period of active intervention despite no regular contact with participants, suggesting that skills acquired under supervision for hypoglycemia prevention may become ingrained. In three studies (32,34,36), improvement in HA was at the expense of worsened glycemic control, with HbA_1c rising significantly to suboptimal values, whereas it remained within therapeutic targets in two (35,37) and showed no significant change in one (9).

Four of 12 studies with baseline IAH used a psychoeducational approach to restore awareness (28–31). BGAT, based on work by Cox et al. (40), focuses on increasing self-awareness of personal cues for hypoglycemia. The DAFNE-Hypoglycemia Awareness Restoration Training (HART) pilot study (28) incorporates motivational interviewing and cognitive behavioral therapies to address behavioral issues found to promote and sustain IAH (41). All of these approaches successfully reduced SH and improved awareness. Although the BGAT studies did not report any prior structured education, the DAFNE-HART program, in particular, recruited people with very high rates of SH, despite having had structured education, and demonstrated success of psychotherapeutic approaches in these people without deterioration in glycemic control. Neither study assessed counterregulatory hormone responses.

Of seven education RCTs (42–48), four recruited unselected patients [BGAT (45,47,48) and Program for Diabetes Education and Treatment for a Self-Determined Living With Type 1 Diabetes (PRIMAS) (42)] and three recruited those with IAH [HyPOS (43,44) and Hypoglycemia Anticipation, Awareness and Treatment Training (HAATT) (46)]. The longest follow-up was 4.9 years in the BGAT programs (48). In the U.S.-based BGAT studies, BGAT was compared against general diabetes education (49), a stress-management training program (48), and education unrelated to glycemic control (cholesterol education) (47). These studies did not report baseline SH rates (47,48) and people with SH in the preceding 2 years of the study were excluded from the Kinsley et al. (47) study. In the European-BGAT study (45), people with recurrent SH were encouraged to participate: SH fell by 88% in the BGAT group compared with the physician-guided self-help control group, which showed no reduction in SH rates. A head-to-head comparison between PRIMAS (42), a new German education program with additional aspects of goal-setting, motivation, and greater hypoglycemia focus, against the well-established DTTP (39) showed equivalent reductions in SH in both intervention and control groups. There was no consistent finding in improvement in HA status, with PRIMAS and BGAT-Kinsley showing improvements with control and intervention but no between-group differences (42,47), whereas other BGAT studies had an additional 15–30% improvement in HA in the intervention versus control group (45,48). These differences can be attributed to the different comparator arms: the PRIMAS study (42), in particular, compared the new educational method with the DTTP, a well-established program (39). Of note, the DTTP structured education program on flexible insulin therapy includes education on hypoglycemia avoidance. Observational studies of the DTTP with longer-term follow-up showed a marked reduction in SH rates by ∼50% in unselected patients with T1D (50) and an even greater reduction of 70–80% in SH rates in those with three or more episodes of SH in the year before DTTP (51). Their principles were adapted into several of the programs included in this analysis.

Counterregulatory hormones to hypoglycemia were only measured in one study, with improved epinephrine response to hypoglycemia in the BGAT group despite no between-group differences in hypoglycemia symptom scores (47). Glycemic control improved in the U.S.-based BGAT studies (47,48) in both control and intervention groups, but showed no change in the European-BGAT (45). PRIMAS showed improvement in HbA_1c only in the intervention arm (42).

Three studies specifically recruited participants with IAH: HyPOS (43,44) and HAATT (46). Similar to PRIMAS, HyPOS compared a biopsychosocial education program with a standard education program, specifically in patients with previous SH. Both groups showed similar reductions in SH at 6 months, but the reduction in SH in the HyPOS group was greater compared with the control group in long-term (2.5 years) follow-up (43,44). There was no difference in long-term glycemic control. The HAATT study compared a psychoeducational program to self-monitoring of blood glucose (SMBG) in Bulgaria, where SMBG was not routinely available. This study showed a reduction in SH and improved detection of low BG in the intervention group despite no significant change in HbA_1c between the two groups, implying that the psychoeducational component of the study was vital in reducing SH and improving awareness (46).

Of five studies that used technology as the primary intervention, two evaluated CSII (52,53) and three RT-CGM prospectively (54) and retrospectively (55,56). Four of these studies recruited people with SH and IAH at baseline and showed significant reductions in SH postintervention (52,54–56). In contrast, an earlier study by Hübinger et al. (53) recruited those starting CSII for poor metabolic control and evaluated the effect of CSII on HA status. IAH developed in 43% of these participants after 6 months on CSII, with an improvement of HbA_1c of 0.5%. In Giménez et al. (52), people with IAH and repeated nonsevere or SH were started on CSII. There was improvement in Clarke score for HA, with participants scoring similarly after treatment to controls without IAH, with no deterioration in glycemic control.

Of two RT-CGM studies, an observational study of 35 participants with recurrent SH showed no change in awareness status using Gold scores, despite a significant reduction in SH, although 54% of participants reported subjective improvement in awareness status (55). The other study (54) showed improvement in hypoglycemia awareness but used a different score, the HYPO-score (57). There was improvement in HbA_1c in the RT-CGM studies that were studied retrospectively with longer duration (≥12 months) (55,56), but no change in HbA_1c in the prospective study with a much shorter duration of 2 months (54).

Six RCTs evaluated technology as the primary intervention (58–63). All except one (62) included IAH or prior episodes of SH in their inclusion criteria. Kovatchev et al. (62) did not specify SH or IAH in their inclusion criteria, but 39.2% of their subjects had IAH. SH was defined according to American Diabetes Association definitions (16) except in the Ly et al. (58) study, which defined SH as hypoglycemia resulting in seizures or coma and moderate hypoglycemia as hypoglycemia requiring assistance for treatment. Their population included adolescents, and the authors were contacted for data restricted to participants aged ≥18 years, reported here.

Baseline assessments of HA status were performed using Clarke (58,60) or Gold scores (59,63) except in Kanc et al. (61), which was based on patient-reported inability to perceive autonomic hypoglycemia warning symptoms, and in one study where assessment was not specified (62).

The Comparison of Optimised MDI versus Pumps with or without Sensors in Severe hypoglycaemia (HypoCOMPaSS) trial (59,63) and Thomas et al. (60) compared CSII against optimized MDI and education, with additional RT-CGM, against SMBG. Both studies included a high degree of support and education from the researchers. Ly et al. (58) compared sensor-augmented pump (SAP) with a sensor-driven automated insulin suspension (low-glucose suspend [LGS]) against conventional CSII, and Kanc et al. (61) compared nighttime CSII with bedtime NPH insulin. In studies that provided structured education or feedback in addition to technology to all participants, SH was reduced and HA status improved in all intervention arms, with technology (CSII or RT-CGM) having no additional benefit (59,60,63). In the adults from Ly et al. (58), there was an equivalent reduction in SH incidence in SAP and in conventional CSII but greater improvement in HA status with SAP. Despite greater frequency of visits compared with routine care, the follow-up of the participants in all of these studies did not differ between arms. Kovatchev et al. (62) investigated the utility of a handheld computer (HHC) device that provided continual education feedback associated with predicted risk of hypoglycemia and glucose variability data. Use of the HHC and predictive data were associated with reduction in SH, greater in those with hypoglycemia unawareness at baseline, with an increase in the BG estimation accuracy index.

Of studies conducting hyperinsulinemic-hypoglycemic clamps, one showed an increase in plasma metanephrine responses to hypoglycemia (59), and two showed no significant differences in hormone responses (58,61).

In all studies comparing CSII with insulin analog therapy, there was no deterioration or differences in glycemic control in any of the intervention arms when compared with control despite reductions in SH and improvements in HA status. In the Kovatchev et al. (62) study of the HHC, there was significant reduction in HbA_1c, especially in those with baseline HbA_1c >8% (64 mmol/mol).

Five studies were identified, all of which were conducted more than 10 years ago. Four studies compared short-acting and long-acting analog insulin against conventional soluble (SI) or NPH insulin (64–67). One noninsulin study was identified, investigating propranolol to restore HA (68). There was no mention of any change in education between the arms.

SH did not occur in three of these studies (64,66,68); two had no statistically significant change in SH rates between study arms (65,67). There was no consistent finding in changes in hypoglycemia symptom scores during hypoglycemic clamp studies between comparator arms in the insulin studies (64–67). The study on propranolol reported increased sweating during hypoglycemia with propranolol (68). There were no significant differences in counterregulatory hormones responses between lispro and SI (64) or SI before meals and NPH at bedtime versus the premix formulation of lispro, HM insulin (75% lispro and 25% neutral protamine lispro [NPL]) with meals and NPL at bedtime (67). There was, however, a higher peak plasma epinephrine response when NPH was delivered separately at bedtime compared with a combined SI and NPH with dinner (66). Counterregulatory hormones were not measured in the remaining two studies (65,68).

There were no significant changes in glycemic control in the three lispro studies (64,65,67). HbA_1c was lower at the end of the treatment period in the split-NPH dosing (66). Changes in HbA_1c were not reported in the propranolol study, which lasted only 1 month.

A meta-analysis for educational interventions on change in mean SH rates per person per year was performed. Combining before-and-after and RCT studies, six studies (n = 1,010 people) were included in the meta-analysis (Fig. 2) (21,28,31,42,43,45). We evaluated the active interventions used in the RCTs as individual before-and-after trials, because all included some educational component, a structured curriculum, and information around causes and prevention of hypoglycemia. For Schachinger et al. (45), the control group was not structured education, so it has not been included (there was a slight increase in SH in the control group).

A random-effects meta-analysis revealed an effect size of a reduction in SH rates of 0.44 per patient per year with 95% CI 0.253–0.628. From the RCT studies (Hermanns et al. [42,43]), which compared new structured education interventions PRIMAS and HyPOS against the established DTTP flexible insulin therapy program, we can conclude that any form of structured educational intervention in flexible insulin self-management has a beneficial and equivalent effect in reducing SH rates. Heterogeneity between studies was significant, with I² statistic of 68.58% (95% CI 34.2–85; P = 0.0023). Supplementary Table 2A and B lists the SMD, the 95% CI of the individual studies, and the test for heterogeneity.

Most of the educational interventions were observational and mostly retrospective, with few RCTs. The overall risk of bias is considered medium to high and the study quality moderate. Most, if not all, of the RCTs did not use double blinding and lacked information on concealment. The strength of association of the effect of educational interventions is moderate. The ability of educational interventions to restore IAH and reduce SH is consistent and direct with educational interventions showing a largely positive outcome. There is substantial heterogeneity between studies, and the estimate is imprecise, as reflected by the large CIs. The strength of evidence is moderate to high.

There were approximately equal numbers of observational and RCTs of technological interventions. These trials were well conducted, with two RCTs of almost 100 patients selected for hypoglycemia unawareness. The overall risk of bias was considered low to medium, with moderate study quality. Double blinding was not possible, and there was lack of information on concealment in the RCTs. Combining all of these studies into a single meta-analysis was not appropriate because CSII, RT-CGM, and SAP are all different categories of technological interventions, with variable reporting of outcomes in each category. Furthermore, provision of education at baseline provides a degree of confounding. In CGM studies, the ability of CGM to reduce SH is consistent and direct, with all included studies showing a positive outcome and reduction in SH rates. The strength of evidence is thus moderate to high. However, the ability to improve or restore hypoglycemia unawareness is uncertain and the strength of evidence is low. The strength of evidence for the ability of CSII to reduce SH and restore hypoglycemia awareness is moderate to high, with a generally positive effect of CSII. However, when patients were provided education and optimized MDI therapy, CSII appeared not to provide any additional benefit.

All of the pharmacological intervention studies were RCTs. Lack of information on concealment was present, but the overall risk of bias was considered low to medium and the study quality was high. However, the strength of evidence for insulin analogs to reduce SH was low because SH was an exclusion criterion for many of the included studies. The strength of evidence of insulin analogs to restore hypoglycemia awareness was low, with no consistent outcome seen.

To our knowledge, this study represents the first systematic review and meta-analysis of the different interventions available for reversing IAH in T1D and includes a comprehensive and expansive literature search. Despite this, there are still limitations. A large proportion of studies did not report the type of diabetes education subjects received before the study intervention, and it is possible that a proportion of patients would have received previous structured education and that some may have had ongoing education given the duration of diabetes in most studies. Another limitation is study heterogeneity and the inconsistent reporting of outcome measures, in particular, in SH rates and measures of HA status, in noneducation studies, preventing a more comprehensive meta-analysis. SH rates were reported as mean (SD), median (interquartile range [IQR]), odds ratios, and proportion of subjects with reduced SH. HA was reported as Gold and Clarke scores, and BG estimation accuracy and the proportion of subjects who had improved awareness was often subjectively assessed. Some studies reported a modified Gold score with a score from 0 to 10 on a visual analog scale. In studies reporting Gold and Clarke scores, we used Clarke scores as the main reporting outcome. In studies that reported Gold scores only, we grouped the outcomes, because Gold and Clarke scores have been shown to be well correlated (69). Even so, it was not possible to perform a meta-analysis due to study heterogeneity. There were also large differences in participant numbers, variable follow-up durations, and differing baseline prevalence rates of SH/IAH.

In an unselected population with no prior diabetes education, structured education or BGAT can reduce SH and improve glycemic control. There is early evidence that such programs can also achieve these outcomes when provided as reeducation some years after the initial exposure (70). In patients with established IAH, BGAT and other psychotherapeutic programs, such as HyPOS and HAATT, are also effective. There was no difference between structured education programs in flexible insulin therapy and programs with a psychological approach when compared head to head, and this may be because in teaching users the basics of insulin pharmacodynamics and how to adjust their insulin regimens around their lifestyles to achieve glucose targets that exclude hypoglycemia, hypoglycemia exposure is lessened. There is perhaps a need to seek the common factors in successful programs to distill the essential elements of any new programs. Meanwhile, DAFNE-HART had a much higher baseline level of SH than any of the other studies and was the only study that took people who were IAH despite prior education. Although a small nonrandomized study, it demonstrated that a psychobehavioral therapeutic approach can have a sustained effect on SH and nonsevere hypoglycemic episodes in people whose IAH seems resistant to other interventions (28).

Thus, in unselected populations with T1D, structured education in flexible insulin usage reduces SH and may reduce the proportion of people with IAH and SH. In those with IAH, further education or BGAT reduces SH, with the greatest reductions seen in programs with a behavioral component.

CSII can reduce SH with greater reductions in those with greater SH at baseline (52), although there was evidence that in an unselected population, CSII and improved control may cause some deterioration of awareness (53). In observational studies, CGM showed a reduction in SH, even in those who remained in IAH despite education and CSII (55). In RCTs of technology, HypoCOMPaSS showed that in the presence of frequent contact, CSII, CGM, and SAP resulted in similar and significant reductions in SH by 57% with a reduction in Clarke scores by ∼2 points compared with optimized MDI and SMBG (63). A RCT of LGS compared with CSII in young people with IAH showed improved awareness and reduced SH with LGS-enabled SAP (58).

Most studies with technology, such as CSII or CGM, were done in patients who had received prior education. Thus, in people with IAH despite prior education, CSII, CGM, and, in particular, sensor-augmented pump therapy with LGS provide additional benefits. The HypoCOMPaSS study (63) is in keeping with earlier studies by Cranston et al. (9) and Fanelli et al. (37) highlighting the importance of close and frequent contact, suggesting that this has a larger effect than any of the technological components tested. HypoCOMPaSS clearly illustrates the value of a holistic approach to the management of people with IAH, using structured education as a core foundation combined with optimized MDI and the use of CSII in selected individuals, to provide far greater advantages than one intervention alone.

We thus propose a stepped-care algorithm that may guide the health care professional in choosing the appropriate intervention when faced with a person with IAH (Fig. 3). We would argue that step one—provision of structured education in flexible insulin therapy—should be available to any person with T1D but that additional resources for individuals with higher care needs may be focused in centers where the more intensive interventions combining psychoeducational and technological interventions are available, to which people with IAH and SH posteducation can be referred.

For future research, we would recommend that outcome measures such as SH rates and HA scores should be reported in a standardized manner to allow future systematic reviews and meta-analyses. Because incidence and prevalence of SH rates are not normally distributed, the median (IQR) SH rate may be more appropriate than the mean (SD) commonly used. Measures of assessment of HA should also be standardized using Gold or Clarke scores because these have been shown to correlate well with clinical and clamp findings and each other. The proportion of patients with baseline IAH and then improved awareness should be reported as well as Gold or Clarke scores and their change.

Future research may be needed to compare structured education, possibly using psychotherapeutic techniques, and optimized MDI using insulin analogs, with comparisons against new diabetes technologies such as LGS-enabled SAP.

In summary, although research-based 1:1 intensive professional support can restore awareness and impaired counterregulation of IAH, group-based educational interventions can also improve hypoglycemia awareness and reduce SH rates in up to 45% of people with IAH, without deteriorating overall glycemic control. Psychotherapeutic techniques may provide additional benefit, in particular in improving HA status, and large RCTs using this approach should be conducted. Use of technology in diabetes, either better warning systems through CGM or through improved insulin delivery via CSII, can reduce SH rates and improve HA without worsening glycemic control, but without restoring counterregulatory hormone responses. A stepped approach is recommended in the management of people with IAH.

Acknowledgments. The authors thank the authors of the original cited studies who were contacted for sharing the information required from their studies.

Funding. E.Y. received fellowship funding as part of the Health Manpower Development Plan award from Khoo Teck Puat Hospital, Alexandra Health Pte, Ltd. (Singapore). M.N. received PhD funding as part of a Diabetes UK project grant. S.A. is partially supported by the National Institute for Health Research (NIHR) Biomedical Research Centre based at Guy’s and St Thomas’ NHS Foundation Trust and King’s College, London. None of the funding or supportive agencies were involved in the design or conduct of the study; collection, management, analysis, or interpretation of the data; or preparation, review, or approval of the manuscript. The views expressed are those of the author(s) and not necessarily those of the funding agencies.

Duality of Interest. P.C. has been on advisory boards and received speaking honoraria/travel support and performed studies for pump manufacturers (Medtronic, Roche, Animas Inc, Cellnovo). M.N. has received travel support from Roche and Lilly UK. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. E.Y., P.C., and M.N. undertook the literature search and reviewed the abstracts and full articles. E.Y. wrote the manuscript. S.A. performed and supervised the statistical analysis. S.A.A. conceived the idea for the review. All authors designed the study, contributed to the discussion, and critically reviewed the final manuscript.

Prior Presentation. Parts of the study were submitted in abstract form to the 8th International Conference on Advanced Technologies & Treatments for Diabetes, Paris, France, 18–21 February 2015, and to the Diabetes UK Professional Conference 2015, London, U.K., 11–13 March 2015.