Study Selection

Six-hundred-eighty studies were identified for inclusion. Full-text review was performed on 51, of which 11 were included for analysis (Figure 1). One non-English language study was translated from German to English by a volunteer. Five identified studies were ultimately excluded because corresponding authors were unable to provide the missing data required to include each study at the full-text phase. Reviewer kappa was 0.55619 for title and abstract screening between authors (MH and JS). For full-text screening, the senior author (JAG) double checked all inclusion decisions and the kappa at this stage was 1.00 between (JS and JAG) and 0.74194 between (MH and JAG).

Figure 1.

PRISMA Flow Diagram of Included and Excluded Studies.

Abbreviation: PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-Analysis.

Study Characteristics

The review included 11 studies: three RCTs^{Reference Carstens and Sprehn16–Reference Patrick, Collier, Hepburn, Steedman, Clarke and Robertson18} and eight observational studies^{Reference Hern, Kiefer, Louie, Barger and Alter5,Reference Chinn and Colella19–Reference Weant, Deloney, Elsey, Combs and French25} with a total of 1,462 patients. Of the 11 studies, eight were conducted in the prehospital setting^{Reference Moore and Woollard17,Reference Chinn and Colella19,Reference Hoffman, Schriger, Votey and Luo20,Reference Howell and Guly22–Reference Weant, Deloney, Elsey, Combs and French25} and originated from six countries (Denmark, United States, Wales, England, Germany, and Scotland; Table 1).

Table 1.

Characteristics of Included Studies

Abbreviations: IV, intravenous; BGL, blood glucose level; D10, 10% dextrose; D40, 40% dextrose; D50, 50% dextrose.

Three studies investigated the effects of D10,^{Reference Hern, Kiefer, Louie, Barger and Alter5,Reference Moore and Woollard17,Reference Weant, Deloney, Elsey, Combs and French25} nine studies investigated D50,^{Reference Carstens and Sprehn16–Reference Hoffman, Schriger, Votey and Luo20,Reference Howell and Guly22–Reference Weant, Deloney, Elsey, Combs and French25} and one study investigated a 40% dextrose (D40) solution.^{Reference Hohenstein21} Two studies compared D10 directly to D50:^{Reference Moore and Woollard17,Reference Weant, Deloney, Elsey, Combs and French25} Moore was an RCT where both interventions were used, and the dose could be titrated to the clinician’s discretion in response to drug effect;^{Reference Moore and Woollard17} Weant was an observational before-and-after study where a new protocol supported clinician discretion on dextrose concentration selection.^{Reference Weant, Deloney, Elsey, Combs and French25} The remaining studies did not have a comparator^{Reference Hern, Kiefer, Louie, Barger and Alter5,Reference Hoffman, Schriger, Votey and Luo20,Reference Hohenstein21,Reference Lerner23,Reference Mechem24} or used another standard hypoglycemia treatment as the comparator (one milligram of intramuscular glucagon; Table 1).^{Reference Carstens and Sprehn16,Reference Patrick, Collier, Hepburn, Steedman, Clarke and Robertson18,Reference Howell and Guly22}

Hypoglycemia was defined as point-of-care capillary measurement of less than 3.0 millimole/liter (mmol/L) at the lowest threshold^{Reference Carstens and Sprehn16} and up to 4.4mmol/L at the higher threshold,^{Reference Lerner23,Reference Mechem24} or by the paramedic’s “empiric impression” in one study (Hoffman). All studies that recorded history of diabetes reported 100% diabetes etiology, except Moore, which reported 84% (43/51) of patients had insulin dependent diabetes.^{Reference Moore and Woollard17} One study reported tissue extravasation.^{Reference Chinn and Colella19}

Data for each outcome of interest were extracted. Ten studies investigated symptom resolution;^{Reference Hern, Kiefer, Louie, Barger and Alter5,Reference Carstens and Sprehn16–Reference Patrick, Collier, Hepburn, Steedman, Clarke and Robertson18,Reference Hoffman, Schriger, Votey and Luo20–Reference Weant, Deloney, Elsey, Combs and French25} six investigated time to symptom resolution;^{Reference Carstens and Sprehn16–Reference Patrick, Collier, Hepburn, Steedman, Clarke and Robertson18,Reference Hohenstein21,Reference Howell and Guly22,Reference Weant, Deloney, Elsey, Combs and French25} glycemic profile post-treatment was investigated by five studies;^{Reference Hern, Kiefer, Louie, Barger and Alter5,Reference Carstens and Sprehn16–Reference Patrick, Collier, Hepburn, Steedman, Clarke and Robertson18,Reference Weant, Deloney, Elsey, Combs and French25} need for subsequent doses by four studies;^{Reference Hern, Kiefer, Louie, Barger and Alter5,Reference Patrick, Collier, Hepburn, Steedman, Clarke and Robertson18,Reference Mechem24,Reference Weant, Deloney, Elsey, Combs and French25} adverse events by seven studies;^{Reference Hern, Kiefer, Louie, Barger and Alter5,Reference Carstens and Sprehn16,Reference Moore and Woollard17,Reference Chinn and Colella19,Reference Lerner23–Reference Weant, Deloney, Elsey, Combs and French25} resolution of hypoglycemia by five studies;^{Reference Hern, Kiefer, Louie, Barger and Alter5,Reference Carstens and Sprehn16–Reference Patrick, Collier, Hepburn, Steedman, Clarke and Robertson18,Reference Lerner23} and time to resolution of hypoglycemia was reported in two studies^{Reference Carstens and Sprehn16,Reference Patrick, Collier, Hepburn, Steedman, Clarke and Robertson18} (Table 2).

Table 2.

Studies Investigating Each Outcome of Interest

Abbreviation: BGL, blood glucose level.

Risk of Bias within Studies

Using the JBI risk of bias assessment tools, two studies were identified as high risk: one RCT and one retrospective case series.^{Reference Patrick, Collier, Hepburn, Steedman, Clarke and Robertson18,Reference Hoffman, Schriger, Votey and Luo20} The Patrick study was deemed high risk due to lack of clarity in reporting of its methods, resulting in concerns regarding valid randomization strategy, blinding and appropriate design execution.^{Reference Patrick, Collier, Hepburn, Steedman, Clarke and Robertson18} The second high risk of bias study to be assessed was the Hoffman retrospective case series. This study was deemed high risk due to its limited reporting of patient results, patient test group similarities, and patient condition prior to treatment.^{Reference Hoffman, Schriger, Votey and Luo20} The remaining studies were all assessed as low risk of bias, included the RCT comparing D10 directly to D50 (Appendix A; available online only).^{Reference Hern, Kiefer, Louie, Barger and Alter5,Reference Carstens and Sprehn16,Reference Moore and Woollard17,Reference Chinn and Colella19,Reference Hohenstein21–Reference Weant, Deloney, Elsey, Combs and French25}

Results of Individual Studies

Resolution of Symptoms—All studies reported on symptom resolution after dextrose treatment. Evidence from nine pooled studies (three RCTs and six observational studies), enrolling 367 hypoglycemic patients treated with IV D50, reported symptom resolution in 88.8% (95% CI, 88.7%-89.1%) of patients (Figure 2).^{Reference Carstens and Sprehn16–Reference Hoffman, Schriger, Votey and Luo20,Reference Howell and Guly22–Reference Weant, Deloney, Elsey, Combs and French25} Evidence from three studies (one RCT and two observational), enrolling 1,057 patients treated with IV D10, reported 95.9% resolution of symptoms (95% CI, 95.8%-95.9%).^{Reference Hern, Kiefer, Louie, Barger and Alter5,Reference Moore and Woollard17,Reference Weant, Deloney, Elsey, Combs and French25} The RCT comparing D10 to D50 reported a non-statistically significant difference in symptom resolution of four percent favoring D50 (-12 to +22%; P = .360).^{Reference Moore and Woollard17} Symptom resolution was 96% (25/26 patients) in the D50 arm and 92% (23/25) the D10 arm.^{Reference Moore and Woollard17} When the results for symptom resolution were pooled across studies, both D10 and D50 had high resolution success with a non-clinically significant difference of 7.1% favoring D10.^{Reference Carstens and Sprehn16–Reference Hoffman, Schriger, Votey and Luo20,Reference Howell and Guly22–Reference Weant, Deloney, Elsey, Combs and French25} One single case reported a patient received a D40 infusion and recovered to normal level of consciousness.^{Reference Hohenstein21}

Figure 2.

Symptom Resolution.

Abbreviation: D50, 50% dextrose.

Time to Symptom Resolution—Evidence from four studies (three RCTs and one observational), enrolling 60 patients treated with D50, had a mean symptom resolution time of 4.1 minutes (95% CI, 3.46-4.78 minutes; Figure 3).^{Reference Carstens and Sprehn16–Reference Patrick, Collier, Hepburn, Steedman, Clarke and Robertson18,Reference Hohenstein21} Evidence from one RCT, enrolling 51 patients treated with dextrose, reported a median time to resolution of 8.0 minutes in the D10 arm (n = 25) and 8.0 minutes in the D50 arm (n = 26; P = .733).^{Reference Moore and Woollard17} The mean difference between these treatments may be clinically meaningful at 3.9 minutes. One single case reported a patient received a D40 infusion had symptom resolution within two minutes.^{Reference Hohenstein21}

Figure 3.

Time to Symptom Resolution.

Abbreviation: D50, 50% dextrose.

Blood Glucose Level Post-Treatment—Evidence from four pooled studies (three RCT and one observational), enrolling 196 total patients treated with D50, resulted in a mean post-treatment glycemic profile of 8.45mmol/L (95% CI, 8.4-8.5mmol/L; Figure 4).^{Reference Carstens and Sprehn16–Reference Patrick, Collier, Hepburn, Steedman, Clarke and Robertson18,Reference Weant, Deloney, Elsey, Combs and French25} Evidence from three pooled studies (one RCT and two observational), enrolling 1,057 patients treated with D10, had a mean post-treatment glycemic profile of 6.2mmol/L (95% CI, 6.24-6.16mmol/L).^{Reference Hern, Kiefer, Louie, Barger and Alter5,Reference Moore and Woollard17,Reference Weant, Deloney, Elsey, Combs and French25} A statistically significant difference of 3.2 minutes (P = .003) was reported by the RCT comparing D50 directly to D10.^{Reference Moore and Woollard17} This RCT reported a median post-treatment BGL of 9.40mmol/L in the D50 arm and 6.20mmol/L in the D10 arm.^{Reference Moore and Woollard17}

Figure 4.

Blood Glucose Post-Treatment.

Abbreviations: BGL, blood glucose level; D50, 50% dextrose.

Need for Subsequent Doses—Evidence from three pooled studies (one RCT and two observational), enrolling 259 patients treated with D50, reported 8.1% (95% CI, 7.9%-8.3%) of patients needed subsequent treatment (Figure 5).^{Reference Patrick, Collier, Hepburn, Steedman, Clarke and Robertson18,Reference Mechem24,Reference Weant, Deloney, Elsey, Combs and French25} Results from two observational studies, reporting on 1,032 patients treated with D10, found 19.5% (95% CI, 19.4%-19.6%) required additional D10 treatment.^{Reference Hern, Kiefer, Louie, Barger and Alter5,Reference Weant, Deloney, Elsey, Combs and French25}

Figure 5.

Need for Subsequent Doses.

Abbreviation: D50, 50% dextrose.

Resolution of Hypoglycemia—Evidence from pooled four studies (three RCTs and one observational), enrolling 78 patients treated with D50, reported 98.7% (95% CI, 98.4%-99.0%) resolution of hypoglycemia (Figure 6).^{Reference Carstens and Sprehn16–Reference Patrick, Collier, Hepburn, Steedman, Clarke and Robertson18,Reference Lerner23} Evidence from two pooled studies (one RCT and one observational), enrolling 896 patients treated with D10, reported 99.2% (95% CI, 99.2%-99.2%) resolution from hypoglycemia.^{Reference Hern, Kiefer, Louie, Barger and Alter5,Reference Moore and Woollard17} The RCT comparing D50 to D10 reported 100% resolution in both groups: 26 of 26 and 25 of 25 patients, respectively.^{Reference Moore and Woollard17}

Figure 6.

Resolution of Hypoglycemia.

Abbreviation: D50, 50% dextrose.

Time to Resolution of Hypoglycemia—Time to resolution of hypoglycemia was reported on in two RCTs investigating D50 administration in 20 patients.^{Reference Carstens and Sprehn16,Reference Patrick, Collier, Hepburn, Steedman, Clarke and Robertson18} The RCT by Patrick enrolling 14 patients receiving D50 reported a BGL of approximately12.5mmol/L at the first check five minutes post-treatment.^{Reference Patrick, Collier, Hepburn, Steedman, Clarke and Robertson18} The RCT by Carstens enrolling six patients receiving D50 reported a BGL of approximately 9.5mmol/L at the first check 10 minutes post-treatment. No data were available for this outcome using D10 treatment.^{Reference Carstens and Sprehn16}

Adverse Events—Seven studies, including 1,367 total patients, investigated adverse events (Table 3 and Table 4). Evidence from six studies (two RCTs and four observational studies), enrolling 310 patients treated with D50, reported a 4.2% (13/310) adverse event rate.^{Reference Carstens and Sprehn16,Reference Moore and Woollard17,Reference Chinn and Colella19,Reference Lerner23–Reference Weant, Deloney, Elsey, Combs and French25} These consisted of three patients reporting mild hypoglycemia requiring only self-treatment, eight rebound calls to 911 within 48 hours, one patient who did not recover who was later diagnosed in-hospital with encephalopathy as the cause of the hypoglycemia, and one case of extravasation. There were no (0/1,057) adverse events reported in the three studies investigating adverse events for D10 treatment.^{Reference Hern, Kiefer, Louie, Barger and Alter5,Reference Moore and Woollard17,Reference Weant, Deloney, Elsey, Combs and French25} The RCT reported no adverse events in either treatment group (0/25 D10 group and 0/26 D50 group).^{Reference Moore and Woollard17} One observational study reported an overall 27.0% (84/311) of patients had hyperglycemia upon hospital arrival with no statistical difference between the D50 and D10 groups. Hyperglycemia had not been considered as an adverse event in this review, but an untoward effect of IV dextrose administration accounted for in the post-treatment BGL results.^{Reference Weant, Deloney, Elsey, Combs and French25} Mortality had not been listed as an a priori outcome but, of note, the Weant study reported a non-significant difference for in-hospital mortality of 4.7% in the D50 group versus 6.2% in the D10 group (P = .623).^{Reference Weant, Deloney, Elsey, Combs and French25}

Table 3.

Results for Each Intervention by Outcome

Abbreviation: BGL, blood glucose level.

Table 4.

Adverse Events