FormalPara Key Summary Points

Controlling diabetes in young patients (youth-onset diabetes) who are treated with insulin is challenging in low-resource settings where the treatment relies only on Neutral Protamine Hagedorn (NPH), regular inusulin and premixed biphasic human insulin formulations.

We hypothesized that metabolic control could be improved in these patients by providing better self-management education in combination with the use of sliding scales which provide for patient-adapted self-adjustment of premixed 70/30 insulin doses that account for missed self-monitoring of blood glucose (SMBG) tests and skipped meals.

The results showed that glycemic control significantly improved with this treatment protocol.

These data call for new strategies for better metabolic control in patients living in limited-resource settings where cutting-edge therapies are still lacking.


Glycemic control is the single most important predictor of an increased risk for diabetes-related complications [1]. Analog basal-bolus insulin regimens are the standard of care in high-income settings where increasingly smarter insulin delivery systems are being developed and used to best mimic physiologic insulin needs. Even with such cutting-edge therapies, achieving glycemic targets [2] in young patients with diabetes (youth-onset diabetes) is very challenging [3] due to multiple factors, including lifestyle, developmental, psychological and hormonal changes.

Patients of ethnic minorities and lower socioeconomic status are disproportionately at risk of poor glycemic control [3,4,5]. This is especially true in less-resourced settings, such as in low-income countries where analog insulins and insulin pumps are mostly inaccessible or unaffordable [6]. Instead, patients and healthcare providers must rely on Neutral Protamine Hagedorn (NPH), regular (R) and premixed biphasic human insulin formulations [6]. The pharmacodynamics of these insulins call for regular meals and snacks to “feed the insulin” when it peaks in order to avoid hypo- or hyperglycemia. In a context where only minimal care standards [6] are available, glucometers and glucose test strips are scarce and food insecurity is common, it is exceedingly difficult to achieve glycemic control among insulin-treated patients with youth-onset diabetes [6,7,8]. In Haiti, youth-onset diabetes is mainly treated with twice-daily human insulin injection regimens; for the most part these include premixed insulin consisting of 70% NPH and 30% R (70/30 insulin), with NPH and R insulins given as separate injections to a lesser extent. The reason for the higher use of premixed insulin among these young patients is essentially due to their low education level, which may be a limiting factor in the self-management of insulin mixing, and the greater availability of premixed insulin on the market compared with NPH insulin.

The few publications on Haitian youth with diabetes treated with insulin have shown poor glycemic control, with an average hemoglobin A1c (HbA1c) of > 11% [7, 8]. Given the economic challenge to obtain analog insulins and increase the number of daily self-monitoring of blood glucose (SMBG) tests, as well as the inability to eradicate food insecurity and irregular access to daily meals and snacks, it became necessary to seek a strategy by which to improve glycemic control despite these context-specific barriers. Therefore, we hypothesized that patient-adapted self-adjustment of premixed insulin doses that accounts for missed SMBG tests and skipped meals could improve glycemic control in youth with diabetes. The aim of this study was to determine the impact of a sliding scale-based insulin adjustment protocol that would account for missed SMBG tests and skipped meals, in young patients with diabetes. The primary objective was to improve HbA1c after the adoption of the new protocol. The secondary objective was to compare the outcomes of a group of patients treated with 70/30 insulin with those of another group treated with self-mixed NPH + R insulins under the same follow-up conditions.


Study Design

This was a prospective open-label randomized controlled trial of patients with insulin-treated youth-onset diabetes in Port-au-Prince, Haiti, from 21 October 2017 to 17 May 2018.

Sample Size

The sample size calculation was based on the primary outcome reduction in HbA1c in G70. Thirty-two patients (16 in each group) were required to detect a 1 percentage point reduction in HbA1c with 80% power and a 5% level of significance.


We recruited young patients from the FHADIMAC (Haitian Foundation for Diabetes and Cardiovascular Diseases) outpatient clinic, which is the largest pediatric diabetes outpatient setting in Haiti, with a catchment area that includes the capital city of Port-au-Prince and the surrounding cities and suburbs. Inclusion criteria were age 11–28 years and treatment with 70/30 human insulin twice daily. Exclusion criteria were renal failure, severe visual impairment and pregnancy. Written informed consent was obtained from adult patients and from adults legally responsible for minors. The study protocol was reviewed and approved by the Haitian National Bioethics Committee of Haitian Ministry of Health (Ref: 1617-55). It was performed in accordance with the Helsinki Declaration of 1964, and its later amendments.

Baseline and Study Treatment Protocols

The baseline treatment protocol included twice-daily insulin doses, one or two SMBG tests per day, a monthly 10-minute visit and the recommendation to present the SMBG and insulin adjustment reports at each visit. The carbohydrate (CHO) counting method, which included Haitian meals, was refreshed.

Patients were provided with free SMBG supplies, insulin and syringes through the Life for a Child Program [9]. Hyperglycemic crises including diabetic ketoacidosis (DKA) and hyperglycemic hyperosmolar state (HHS) were defined according to the consensus guidelines of the International Society for Pediatric and Adolescent Diabetes (ISPAD) [10]. The biochemical criteria, which are generally associated with a variable clinical presentation, for the diagnosis of DKA were: blood glucose (BG) > 200 mg/dL (11 mmol/L), venous pH < 7.3 or bicarbonate < 15 mmol/L, ketonemia or ketonuria. The criteria for HHS included: BG > 600 mg/dL (33.3 mmol/L), venous pH > 7.25 or arterial pH > 7.30, serum bicarbonate >15 mmol/L, small ketonuria, absent to mild ketonemia, effective serum osmolality > 320 mOsm/kg, altered consciousness (e.g. obtundation, combativeness) or seizures. Severe hypoglycemia was defined in adulthood as an asymptomatic hypoglycemic event associated with severe cognitive impairment (including coma and convulsions) requiring external assistance by another person to actively take corrective actions. In childhood, it was defined as an event associated with severe neuroglycopenia resulting in coma or seizure and requiring parenteral therapy (glucagon or intravenous glucose) [11].

From the list of all patients who regularly attended the clinic during the 12 weeks prior to study start, we selected those who met the inclusion criteria. After removing the selected subjects who presented one or more exclusion criteria, 32 remained as study participants. They were assigned via fixed block randomization (block size 4) to remain on 70/30 (group 70 [G70], n = 16) or to switch to self-mixed NPH + R (group NR [GNR], n = 16), with twice-daily injections. In GNR, NPH insulin accounted for 50% of the total daily dose (TDD). Both groups were instructed to consume about 60 g of CHO at each of the three daily meals and up to two 15-g snacks per day at mid-morning and mid-afternoon, respectively. A carbohydrate-to-insulin ratio was calculated for each patient using the 450 rule (450 divided by TDD) to obtain an estimate of how many grams of CHO can be consumed for each 1 unit of insulin given [12]. Meal insulin doses (MID), representing the insulin R doses required for the CHO content in a meal, were calculated for 60 g and other amounts of CHO. An insulin sensitivity factor (that refers to the mg/dL drop in BG caused by 1 unit of rapid acting or regular insulin taken in a fasting or pre-meal state) was calculated by dividing the constant 1700 by the TDD [12,13,14]. Planned dose is defined as the previous dose adjusted, when needed, to its effect on BG.

A sliding scale was provided to all patients according to their randomization and TDD. For patients in G70, the front side displayed insulin correction doses (ICD) for different BG ranges (see Table 1) and were used to adjust or determine the 70/30 insulin doses shown in Table 2 when applicable (meal skipped and BG measure missed, or mid-day BG, when performed, > 240 mg/dL). The other cases were based on a retrospective algorithm (planned dose). For patients in GNR, the front side of the sliding scale displayed the ICD (Table 1) and the pre-meal insulin R doses (ICD + MID) related to 60 g of CHO for different BG ranges. However, adjustment of the twice-daily doses of NPH insulin was based on a retrospective algorithm (planned dose). Additional instructions were provided on pre-meal insulin R dose adjustment for estimated CHO loads other than 60 g. The corresponding part (G70 or GNR) of Table 2 was positioned on the back of the sliding scales to determine insulin doses based on the presence of pre-meal BG tests and meals.

Table 1 Insulin correction doses
Table 2 Insulin dose adjustment according to presence of pre-meal blood glucose tests and meals

SMBG and insulin were routinely administered before morning and evening meals. The pre-meal glycemic target was 80–130 mg/dL. Any BG test not done in the morning or evening with or without a meal was considered missed. Any meal not taken among the three standard daily meals was considered skipped. Patients were encouraged to perform SMBG at the time of any skipped meal, before the next meal following BG < 80 mg/dL or if needed (symptoms of hypo- or hyperglycemia, or unexplained malaise).

All patients visited the diabetes clinic twice a month for a 14-week intervention period. The visit duration was also liberalized to allow for better communication on SMBG, insulin adjustment rules and meal CHO load. All visits were timed to track duration. At the end of this 14-week period and up to 29 weeks post-enrollment, while the current insulin types in each group were maintained, the use of sliding scales was discontinued and the previous 10-min-long, once-monthly visit was resumed.

Data Collection

Clinical and demographic information was collected from the patients’ records. At 14 weeks, patients were re-evaluated in terms of glycemic outcomes, insulin needs, number of meals taken and number of BG tests performed. HbA1c was performed with a point-of-care DCA Vantage machine (Siemens Healthineers AG, Erlanden, Germany) on study day 1 (day of training and provision of all necessary supplies) and at 14 and 29 weeks.

Statistical Analysis

The Epi Info version software package was used for data entry and descriptive statistics, and SPSS Statistics version 17.0 (SPSS IBM Corp., Armonk, NY, USA) was used for non-parametric tests and multivariate analysis. We used the Mann–Whitney U test to compare the median between the groups and the chi-squared test for categorical variables. A p value of < 0.05 was considered to be significant. All analyses were intention-to-treat.


The complete analysis was performed on 30 participants because two participants left the study during the intervention period for personal reasons. All patients could be assessed as belonging to the lower social class of the general population. The baseline and 14-week characteristics are shown in Table 3. At baseline, the median age at evaluation and diagnosis, rate of patients diagnosed after age 14 years, proportion of females, body mass index (BMI) Z-score, duration of diabetes, HbA1c, history of DKA at any stage and TDD of insulin did not differ between the two groups. G70 has a median HbA1c value of 9.5% (interquartile range [IQR] 8.8%, 11.1%) (80.3 mmol/mol) and GNR has a median HbA1c value of 10.6% (IQR 8.1%, 14.4%) (92.4 mmol/mol); the difference was not statistically significant (p = 0.59). A wide dispersion of HbA1c values was noted in GNR, with six patients with HbA1c > 12% versus one patient in G70. The HbA1c target of < 8% (63.9 mmol/mol) was reached in two patients in each group (13.3%). None of the patients were obese. None of the patients presented retinopathy nor microalbuminuria. Cataract was found in two patients (one in each group), and one patient in G70 presented suspected glaucoma.

Table 3 Characteristics of participants at baseline and 14 weeks

At 14 weeks, the duration of the clinic visit had increased by 7 min compared to the baseline average visiting time of 10 min. Five patients in GNR had persistent difficulty in fully mastering the insulin mixing technique. The HbA1c in G70 patients decreased by 1.5%, reaching a median of 8.0% (IQR 7.1%, 9.2%) (63.9 mmol/mol) (p < 0.01); in comparison, the HbA1c in GNR patients decreased by 1.6%, reaching a median of 9.0% (IQR 7.6%, 9.9%) (74.9 mmol/mol) (p = 0.10) (Fig. 1). There was no statistically significant between-group difference in HbA1c reduction (p = 0.12). The proportion of patients with HbA1c < 8% (63.9 mmol/mol) reached 46.7% (7/15) in G70 (p = 0.04) and 33.3% (5/15) in GNR (p = 0.39); this between-group difference was not statistically significant (p = 0.45). Large fluctuations in BG levels were not found in any patient. The median number of missed SMBG tests per week was significantly higher in GNR than G70 (0.5 [IQR 0.3, 2.8] vs. 0.3 [IQR 0.0, 0.8]; p = 0.04). The median number of skipped meals per week did not differ between the two groups (4.3 [IQR 0.9, 6.2] in G70 vs. 5.0 [IQR 0.0, 7.4] in GNR; p = 0.60). No serious acute complications, including severe hypoglycemia, were reported in either group. However, all patients reported at least one episode of mild hypoglycemia. No patient had to take additional insulin injections. The TDD (in units) increased in both groups from baseline, without statistical significance (G70: 8 [IQR 3.5, 14], p = 0.11; GNR: 7 [IQR − 0.5, 12] ;p = 0.44). In the multivariate regression analysis for all patients, age at evaluation, age at diagnosis, diabetes duration, TDD/kg, type of insulin regimen, BMI Z-score and number of missed SMBG and skipped meals did not predict HbA1c values.

Fig. 1
figure 1

Change in glycated hemoglobin (HbA1c) during the study period. In both treatment groups, HbA1c decreased significantly after 14 weeks, and re-increased to values not significantly different from baseline after 29 weeks

At 29 weeks, HbA1c increased in both groups, reaching 8.4% (IQR 6.9%, 10.2%) (68.3 mmol/mol) and 9.2% (IQR 7.8%, 12.0%) (77.0 mmol/mol) in G70 and GNR, respectively (p = 0.19) (Fig. 1). These HbA1c values were not statistically different from those at week 14 (G70: p = 0.60; GNR: p = 0.26) nor from those at baseline (G70: p = 0.08; GNR: p = 0.32).


The results of our study demonstrate that in a resource-limited setting where SMBG is infrequent, food insecurity is common and glycemic control is extremely poor, the use of a simplified sliding scale for the twice-daily administration of premixed insulin 70/30 in combination with increased frequency and duration of clinic visits significantly lowers HbA1c. The increase in HbA1c to values not significantly different from baseline at week 29 was due to the removal of the sliding scales and the return to the previous schedule of a once-monthly visit to the diabetes clinic, thereby confirming the effectiveness of the evaluated treatment protocol. The improvement in glycemic control during the intervention period was obtained without increasing the risk of hypo- or hyperglycemic events or other acute complications. However, the decrease in HbA1c with the sliding scale-based self-mixed NPH + R insulin regimen and the biweekly visits to the diabetes clinic did not reach statistical significance, probably because of the wide dispersion of baseline HbA1c values in the GNR, the short period of follow-up and, possibly, the persistent difficulty for some GNR patients to properly self-mix insulins, resulting in dosing errors.

While glycemic control was overall very poor and remained above the recommended target range for the vast majority of patients even during the intervention period, median HbA1c in both groups was in the expected range of mean HbA1c for an “intermediate care” environment, which is 8–9.5% (64–80 mmol/mol) [6]. Interestingly, this result is achieved without the use of all “intermediate care” resources, which include multiple daily injections (“basal-bolus regimen”). This improvement in glycemic control, if maintained, should lead to a reduction in long-term complications and mortality, as has been shown in some studies [6, 15]. In a recent population-based cohort study, Marcus Lind found that HbA1c > 8.6% (> 70 mmol/mol) was a risk factor for proliferative retinopathy and macroalbuminuria in 10,398 children and adults with type 1 diabetes [16]. This finding highlights the 8% HbA1c level at week 14 in G70 in terms of reducing the risk of severe diabetic complications.

A twice-daily rather than multiple-daily insulin injection regimen was chosen for two reasons. First, glucose test strips are often unaffordable in Haiti; second, culturally, our lower-class young Haitian patients have low compliance with midday SMBG and insulin injection outside the home. They argue that the material is too cumbersome to have with them and that performing the BG test and injecting themselves with insulin injection with the knowledge of their peers is stigmatizing. A retrospective adjustment algorithm is actually among the best insulin dosing adjustment methods in terms of reducing glycemic fluctuations. However, it would be error-prone, subjective, and very difficult to apply to those receiving 70/30 insulin when meals are skipped and BG measurements missed. This is the main reason for choosing a non-typical sliding scale for this study. It is also worth noting that taking into account the insulin sensitivity factor, the CHO ratio and CHO content of the meals in the calculations establishing our sliding scale insulin regimens may have helped reduce large fluctuations in BG levels (“roller coaster” phenomenon).

Premixed human insulin regimens have been shown to be associated with a more important risk of hypoglycemia than self-mixed or basal-bolus regimens due to the impossibility of separately adjusting the two types of insulin to BG levels or CHO load [11, 12]. They also provide poorer metabolic control than a self-mixed regimen when used in adolescents [4], although selection bias in these studies is possible since adolescents may have been put on premixed insulin because of non-adherence. In addition, glycemic control in teenagers is generally more difficult to achieve, likely due to a combination of social, hormonal, dietary and activity factors—even in high-income countries. For example, in the USA-based type 1 diabetes exchange registry, mean HbA1c in teenagers aged 15–18 years was 9.3% (78 mmol/mol) as compared to 8.1% (65 mmol/mol) in their 5-year-old peers [3].

Our findings suggest that in our population of youth who reside in a resource-limited, food-insecure setting, the use of sliding scales adjusted for missed SMBG and skipped meals can be a key factor in achieving better glycemic control, without severe hypoglycemia, at least in patients treated with premixed 70/30 human insulin. The beneficial effect was achieved in the presence of a minimum but essential clinical support package that included two rather than one clinic visits per month and a moderate increase in the length of each visit for better self-management education. In resource-limited settings where children and adolescents are considered autonomous at a much younger age and patients’ families have a lower education level, these findings can provide important information to diabetes care teams who are often called upon to substitute for families in terms of providing the required care support to youth with diabetes. Further, healthcare providers may encourage the transfer from a self-mixed regimen to a premixed one when the former appears to be a factor contributing to poor treatment adherence [17].

Our study has a number of limitations. The small size of the sample and the short duration of the follow-up period may have prevented us from finding significant differences between and within groups or detecting significant predictors of poor glycemic control. The high baseline HbA1c level may have made patients more prone to improving their metabolic control. Our protocol, which prevented some patients from taking insulin at midday and night in cases of missed BG tests and skipped meals, was a potential risk factor for DKA, particularly on sick days, but this risk was lessened by the systematic administration of insulin in the morning and the patients’ awareness to signs and management of DKA. Similarly, failure to administer 70/30 insulin for a BG level < 80 mg/dL (4.4 mmol/mol) in the absence of meal could result in DKA but, based on our long-standing experience at FHADIMAC, we hypothesized that rapid-acting insulin, even after ingestion of fast-acting carbohydrates, might increase the risk of hypoglycemia in the following hours. SMBG before the next meal, as recommended, then might help decrease the risk of DKA. On the other hand, the limitation of only one to two SMBG tests per day could have under-detected rates of hypoglycemia. Lastly, it was not possible to dissociate the glycemic effect of the sliding scale from that of the change in the characteristics of the visit, since the two formed a package.


The use of sliding scales adjusted for missed SMBG tests and skipped meals, and biweekly clinic visits that focus on patient self-management education significantly improved glycemic control in youth with diabetes treated with a premixed 70/30 human insulin regimen. This information may further contribute to motivating the search for new strategies to improve metabolic control in patients on premixed biphasic insulins in socioeconomic contexts where advocacy efforts to access analog insulins and greater availability of SMBG supplies remain unsuccessful.