Keywords Child - diabetes mellitus - insulin - insulin infusion systems - preschool - type
1
Introduction
The highest upsurge in the increasing incidence of Type 1 diabetes mellitus (T1DM)
is observed among preschoolers.[[1 ]] Parental treatment administration often causes psychological distress attributed
to children's normative misbehavior.[[2 ]],[[3 ]],[[4 ]],[[5 ]],[[6 ]] In addition, T1DM management is not mastered by most parents.[[7 ]],[[8 ]]
Continuous subcutaneous insulin infusion (CSII), also known as insulin pump therapy,
has become a progressively standard mode of treatment for this age group as it offers
some advantages over multiple dose injection (MDI) insulin therapy.[[2 ]],[[9 ]] Several better clinical outcomes of CSII over MDI have likewise been reported,
including improvement in HbA1c and lower cardiovascular mortality.[[10 ]],[[11 ]] However, some associated adverse events were also observed, thereby negatively
affecting its safety profile.[[12 ]],[[13 ]] Parental knowledge, in addition, is often noncomprehensive.[[14 ]]
This study primarily aims to investigate the benefits of CSII versus MDI in preschool
children in effectiveness, safety, and parental satisfaction. Effectiveness is described
as the outcome of glycemic control; safety is assessed by minimal hypoglycemic effect,
and parental satisfaction is defined as the enthusiasm to continue on CSII. Moreover,
this study aims to suggest criteria to help clinicians choose suitable patients for
CSII.
Materials and Methods
Electronic manual search to identify all related articles in the following online
databases CINAHL, MEDLINE (OVID), PubMed, Cochrane Library, Embase, Evidence NHS,
PsycINFO, Scopus, NICE, Best Practice (BMJ) since there is no single source that provides
all relevant literature.[15,16] The following search parameters were used: (1) study
objectives: randomized controlled trials (RCTs) and nonrandomized studies which discuss
the selected outcomes, (2) participants: children 0–6 years of age with T1DM, (3)
types of intervention: CSII in comparison with MDI, and (4) years of publication:
the period from 1984 to February 2019 was chosen because it is unlikely to find any
related studies before 1984.[[17 ]]
The full texts of potentially relevant studies were retrieved and thoroughly examined
by two reviewers. Forty-two records were identified. Thirty duplicates were removed,
and seven additional studies were excluded. Five studies were included in the qualitative
and quantitative analysis [[Figure 1 ]]. Data extraction was carried out independently by three reviewers. Discrepancies
between the reviewers were resolved by discussion.
Figure 1: Flow diagram of database searching
Results
Random effects meta-analyses were performed for effectiveness (5 studies) and safety
(3 studies) [[Table 1 ]], patients on MDI as the control group and patients on CSII as the intervention
group. Studies that did not have a control[[9 ]],[[18 ]],[[20 ]] and those with children in school age[[20 ]],[[21 ]],[[22 ]] were not included.
Table 1: Summary of the five controlled studies included in the meta-analyses including effect
sizes for group comparisons in effectiveness, safety, and parental stress variables
Table 1: Contd...
Effectiveness
The effect size for all of the included studies corresponds to the difference in glycemic
index measured as glucosylated hemoglobin (HbA1c) between intervention and control
groups. The mean and standard deviation (SD) of HbA1c measured at all time points
after baseline was grouped for the two study groups. It was deemed appropriate to,
where possible, combine the HbA1c scores taken at different time points during the
trials rather than use the final HbA1c outcome score, which was usually after 6 months.
This is because the purpose of the intervention is the management of the disease over
time.
Cohen's d effect sizes were calculated for each of the five studies included in the
meta-analysis. The mean and SD for the difference in HbA1c between the baseline and
follow-up were not reported in four of the five studies. This implies that the differences
at baseline were not accounted for in the effect sizes for these studies. However,
these studies reported no differences in HbA1c scores at baseline between randomized
control and intervention groups. Due to this, it was assumed that the effect size
reflects the difference between the two insulin treatment methods on HbA1c. Wilson
et al., 2005,[[24 ]] reported the mean difference in HbAlc between the baseline and follow-up stages
for participants who completed the study, and so this result was used in the meta-analysis.
Appendix 1 shows the information extracted from each study ahead of the meta-analysis,
including Cohen's d effect size. [[Figure 2 ]] is a forest plot showing the results of the fixed effect meta-analysis for the
five controlled studies. The effect in the meta-analysis showed a small, nonsignificant
positive effect of the CSII compared to the MDI insulin intervention method (mean
Cohen's d effect size = 0.25, SE = 0.18, P = 0.16 [n = 127]).
Figure 2: Random-effects meta-analysis of the effectiveness of continuous subcutaneous insulin
infusion versus multiple dose injection insulin administration techniques for children
6 years old and younger
A Cohen's d between 0.2 and 0.5 (such as the effect found in this meta-analysis) is
considered a small effect size, an effect that may not be apparent when visually inspecting
the data, but that nonetheless is clinically important. It has been suggested that
actual clinically significant results have effect sizes of more than 0.4, and the
effect size found in this study was smaller than the alternative value.[[25 ]]
Safety
Four of the controlled studies reported the number of incidences when blood sugar
levels dropped below a safe level (blood glucose below <60 mg/dL,[[26 ]],[[27 ]] <5.6 mmol/L,[[24 ]] and <70 mg/dL[[22 ]]). These outcomes were deemed as similar measures of hypoglycemic incidents and
therefore were included in the meta-analysis.
However, the results from Opipari-Arrigan et al., 2007,[[22 ]] were not included in the analysis because the groups were significantly different
at baseline, and the mean difference between baseline and follow-up was not reported.
Including this study would be misleading because the difference in blood glucose level
between the groups may be due to differences between participants within those groups
observed at baseline.
Cohen's d effect sizes were calculated for each of the three studies included in the
meta-analysis. The mean and SD difference in hypoglycemic effect between the baseline
and follow-up were not reported in these three studies. For these studies, the differences
at baseline were not accounted for in the effect sizes. However, studies were only
included if they reported no differences between participants at baseline, and participants
were randomized to the control and intervention groups. Due to this, it was assumed
that the effect size reflects the difference between the two insulin treatment methods
on the hypoglycemic effect.
[[Figure 3 ]] is a forest plot showing the results of the fixed effect meta-analysis for the
three controlled studies. The effect in the meta-analysis showed a small, nonsignificant
negative effect on safety of the CSII compared to the MDI insulin intervention method
(mean Cohen's d effect size = −0.26, SE = 0.36, P = 0.47 [n = 70]).
Figure 3: Random-effects meta-analysis of the safety of continuous subcutaneous insulin infusion
versus multiple-dose injection insulin administration techniques for infants 6 years
old and younger
Parental satisfaction
Only two controlled studies[[22 ]],[[28 ]] offered enough data about parental satisfaction to be included in the meta-analysis.
These studies used different questionnaires (parental stress index and diabetes quality
of life [QoL] questionnaire). One of the noncontrolled studies[[9 ]] also measured some aspects of parental satisfaction. However, since study participants
were unblinded and this study was uncontrolled, it may have been biased. Given that
only two controlled studies could be combined, the conclusions drawn from such a meta-analysis
would be limited.
Discussion
All of the studies seemed to have included the MDI component, although the frequency
may have differed. Questions may also be raised about whether the parents were already
using this administration method before the beginning of each study and if the pump's
introduction was always new. Patient similarity may have also been an issue since
there is a big difference between younger participants and older ones. Another factor
may be differences in severity. Fox et al., 2005,[[27 ]] had patients with low glycemic index, therefore probably less severe than the other
studies. Whether there are differences in the impact of CSII administration for more
severe cases is unsettled. These factors influence the comparability of the studies.
The random-effects meta-analysis model considers that each study is different and
probably estimates a distinct “true” effect size. Therefore, the random-effects estimate
(in this case, d = 0.25) estimates the mean of a normal distribution of true effect
sizes.
The findings of Shehadeh et al., 2004;[[9 ]] Jeha et al., 2005;[[19 ]] and Litton et al., 2002[[18 ]] were not included in analyses because these did not directly specify control conditions
such as learning factors or placebo effect. For the study of Nabhan et al., 2009,[[28 ]] only the results of up to 6 months were included because although the control group
started using CSII after this duration, the CSII group did not cross over to intensive
insulin therapy (IIT). The standard error and control condition means had to be extracted
from the figure as no values were reported in the text. This may impact the precision
of the results of this study. Moreover, the baseline PSI results were not reported
by the group that the parents were in. Thus, while comparing parents' stress in IIT
versus CSII groups as possible, there seemed to be no method to compare their baseline
levels of stress. Therefore, the difference, or lack thereof, of PSI scores between
parents in the intervention and control groups was impossible to interpret without
assuming their similarity or difference at the beginning of the trial. Thus, this
study was not included in the meta-analysis with other controlled studies due to a
lack of control on this aspect. Fox et al., 2005,[[27 ]] likewise failed to report the actual scores for the parental stress index; therefore,
only two sets of results were left, making meta-analysis unbeneficial.
Below targets baseline of glucose concentrations was used in the study of Wilson et
al. 2005,[[24 ]] combined with two other studies that measure the proportion of blood sugar level
<60, the measurement of the same parameters (e.g., the proportion of times participants
dropped below safe blood sugar levels) was assumed. The glucose events from the study
of Opipari-Arrigan, et al., 2007,[[22 ]] were not included in the meta-analysis because the groups were significantly different
at baseline, and the mean difference between baseline and follow-up was not reported.
When performing a meta-analysis, it is common to use and analyze a funnel plot to
judge whether or not there has been publication bias. This was not done because only
three and five studies were included in the meta-analyses, and the two studies showed
a negative effect in each analysis.
The insulin pump has probably been considered less safe for preschool children because
studies have not used the latest insulin pump, which has a feature of “before low
suspend.” With the new technology, which has hypoglycemia alerts, alarms and insulin
suspend feature, the insulin pump may be safer for preschool children than MDI.
An obvious limitation of this review is the small sample size which may not represent
the general population of preschool children with T1DM. Only a few RCTs studied this
specific age group (preschool children) separately. Despite that they are recent,
other RCTs mixed older children with preschool children, which made it impossible
for us to separate them from the rest of the group. This also reduced the power of
statistical analyses to detect significant differences.
It is well known that improvement of HbA1c is mainly related to the effectiveness
of glucose monitoring; however, there was no mention of glucose monitoring devices
used in the RCTs.
Moreover, time in range (TIR) would have been more informative, but the studies included
in our analysis were carried out before we started using TIR.
Conclusions
As there are no CSII selection criteria for this age group, a few conclusions emerge
from this review. First, parents need to have realistic expectations of CSII on optimizing
basal/bolus therapy, reducing recurrent severe hypoglycemia, and reducing recurrent
ketoacidosis; second, regarding psychological aspects, the need to agree between parent
and the team, strong parental motivation, and capacity to use the pump; and third,
concerning educational factors, parents need to be educated on the insulin pump to
learn the basics of dealing with an insulin pump, safety and basics of carbohydrate
counting, dose adjustments, frequency of blood sugar testing, and how to deal with
daily changes. Further research for a longer duration and the use of a multicentre
approach to study larger populations of preschool children with T1DM may be the best
way to provide more significant results and to assess long-term outcomes of CSII use.
A standard QoL questionnaire should also be developed for parents of preschool children
with T1DM.
Authors' contribution
All authors contributed substantially to the study by literature search, data analysis,
drafting and revising of the manuscript and approval of its final version.
Compliance with ethical principles
No formal ethical approval is required for systematic review–meta-analysis type of
study.
