
hypoglycemia 
1691 
LAIA glargine or detemir, compared to NPH insulin for T1D, assessing glycated hemoglobin (A1C) and hypoglycemia . Data sources included Pubmed, Cochrane Library, EMBASE and handsearching. The methodological quality 

hypoglycemia 
2148 
evidence that LAIA are efficacious compared to NPH, with estimates showing a reduction in nocturnal hypoglycemia episodes (RR 0.66; 95% CI 0.57; 0.76) and A1C (95% CI 0.23; 0.12). No significance was found related 

hypoglycemia 
2272 
(RR 0.66; 95% CI 0.57; 0.76) and A1C (95% CI 0.23; 0.12). No significance was found related to severe hypoglycemia (RR 0.94; 95% CI 0.71; 1.24).ConclusionThis study design has allowed us to carry out the most comprehensive 

hypoglycemia 
5266 
longacting insulin analogues are effective and safe for T1D patients compared to NPH in reduction of hypoglycemia and maintenance or improvement of glycemic control.Materials and methodsWe performed an overview of 

hypoglycemia 
6331 
included at least one of the primary outcomes related to efficacy (A1C, and general, severe and nocturnal hypoglycemia ).In cases of eligible systematic reviews that included RCTs and other study designs, we included these 

hypoglycemia 
7361 
analyzed only rapidacting insulin analogues as the intervention, or assessed outcomes other than A1C and hypoglycemia .To update the included systematic reviews, we performed a separate search for RCTs assessing longacting 

hypoglycemia 
7716 
the end of study and baseline or the measures of the baseline and the end of study independently) and hypoglycemia , considering the categories: general, severe and nocturnal (in terms of number of episodes or rate of 

hypoglycemia 
12903 
in cases of no reply, we assumed SD = 0.05. For dichotomous variables (general, severe and nocturnal hypoglycemia ), we used the number of episodes per personweek, which is calculated by dividing the total number of 

hypoglycemia 
13109 
number of episodes in each group by the number of persons in each group adjusted by the time until a hypoglycemia occurs. When studies reported only the rate of episodes per persontime (time other than a week), we 

hypoglycemia 
13377 
episodes. For crossover trials, we considered the A1C mean difference in the first period and the sum of hypoglycemia episodes in both periods of analysis. For trials that analyzed two intervention groups (i.e.: the same 

hypoglycemia 
13595 
drug but different frequency or time of injection), we considered the best A1C results and summed the hypoglycemia episodes of both groups. This was different from Rossetti and colleagues [[19]], who yielded the hypoglycemia 

hypoglycemia 
13705 
hypoglycemia episodes of both groups. This was different from Rossetti and colleagues [[19]], who yielded the hypoglycemia results as episodes per persontime, thus we considered the best results, as it was impossible to sum 

hypoglycemia 
15017 
confidence intervals by DerSimonian & Laird method. For dichotomous data (general, severe and nocturnal hypoglycemia ) we calculated the relative risk as the effect measure. The measure for any type of hypoglycemia considered 

hypoglycemia 
15114 
nocturnal hypoglycemia) we calculated the relative risk as the effect measure. The measure for any type of hypoglycemia considered in the analysis was number of episodes per personweek, so the relative risk should be interpreted 

hypoglycemia 
16863 
insulin analogues; (f) if study design was crossover or not; (g) followup period; (h) definition of hypoglycemia (just for general hypoglycemia); and if study had provided standard deviation for A1C measures.We assessed 

hypoglycemia 
16894 
design was crossover or not; (g) followup period; (h) definition of hypoglycemia (just for general hypoglycemia ); and if study had provided standard deviation for A1C measures.We assessed the potential for publication 

hypoglycemia 
23314 
Concerning quantitative analysis, the outcomes that were most frequently analyzed were A1C and general hypoglycemia (25 trials).In relation to A1C, the longacting insulin analogue yielded a statistically significant 

hypoglycemia 
23859 
[[67]].10.1371/journal.pone.0194801.g002Fig 2Metaanalysis of outcomes assessed in this overview: A: glycated hemoglobin—A1C; B: general hypoglycemia ; C: severe hypoglycemia; D: nocturnal hypoglycemia.Regarding hypoglycemia, its definition varied among 

hypoglycemia 
23883 
outcomes assessed in this overview: A: glycated hemoglobin—A1C; B: general hypoglycemia; C: severe hypoglycemia ; D: nocturnal hypoglycemia.Regarding hypoglycemia, its definition varied among the analyzed trials. 

hypoglycemia 
23910 
overview: A: glycated hemoglobin—A1C; B: general hypoglycemia; C: severe hypoglycemia; D: nocturnal hypoglycemia .Regarding hypoglycemia, its definition varied among the analyzed trials. Furthermore, authors have classified 

hypoglycemia 
23933 
hemoglobin—A1C; B: general hypoglycemia; C: severe hypoglycemia; D: nocturnal hypoglycemia.Regarding hypoglycemia , its definition varied among the analyzed trials. Furthermore, authors have classified hypoglycemia 

hypoglycemia 
24033 
hypoglycemia, its definition varied among the analyzed trials. Furthermore, authors have classified hypoglycemia into many categories, such as general, overall, minor, severe, major, symptomatic, allday, day and 

hypoglycemia 
24547 
the general or severe class, depending on the applied definition in the studies.Considering general hypoglycemia , the metaanalysis showed a statistically significant difference between the insulin analogue and NPH 

hypoglycemia 
24831 
effect was statistically significant, its magnitude and clinical significance were small.Data for severe hypoglycemia were usefully available from 16 studies (Fig 2C). Concerning this outcome, the pooled estimate of all 

hypoglycemia 
25158 
significant results compared to NPH, revealing a RR = 0.94 (95% CI 0.71; 1.24; I2 94.7%).Concerning nocturnal hypoglycemia , it was available from 20 clinical trials, and the metaanalysis results favor insulin analogues (RR 

hypoglycemia 
25688 
by duration of the study, but not statistically significant (p = 0.16). The heterogeneity in general hypoglycemia was explained in 26.13% by bolus insulin (p = 0.01), in 13.96% by hypoglycemia definition (p = 0.10), 

hypoglycemia 
25767 
heterogeneity in general hypoglycemia was explained in 26.13% by bolus insulin (p = 0.01), in 13.96% by hypoglycemia definition (p = 0.10), and in 7.16% by duration of study (p = 0.10). For nocturnal hypoglycemia, it 

hypoglycemia 
25863 
by hypoglycemia definition (p = 0.10), and in 7.16% by duration of study (p = 0.10). For nocturnal hypoglycemia , it was explained in 23.71% by crossover design (p = 0.08) and in 13.55% by intervention brand (p = 

hypoglycemia 
25994 
in 23.71% by crossover design (p = 0.08) and in 13.55% by intervention brand (p = 0.08). For severe hypoglycemia , heterogeneity was marginally explained by age in 3.84% (p = 0.29).Analysis of publication biasVisual 

hypoglycemia 
26149 
age in 3.84% (p = 0.29).Analysis of publication biasVisual inspection of Begg’s funnel plots for hypoglycemia endpoints (Fig 3A, 3B and 3C) showed a tendency of published trial results to have a small risk of publication 

hypoglycemia 
26383 
the Egger’s test confirmed this direction for general (p = 0.75), severe (p = 0.68), and nocturnal hypoglycemia (p = 0.74). However, A1C studies were very dispersed (Fig 3D), suggesting that there is publication 

hypoglycemia 
26678 
0.05).10.1371/journal.pone.0194801.g003Fig 3Publication bias analysis by Begg’s funnel plots.A. General hypoglycemia . B. Severe hypoglycemia. C. Nocturnal hypoglycemia. D A1C.DiscussionTo the best of our knowledge, this 

hypoglycemia 
26702 
0.05).10.1371/journal.pone.0194801.g003Fig 3Publication bias analysis by Begg’s funnel plots.A. General hypoglycemia. B. Severe hypoglycemia . C. Nocturnal hypoglycemia. D A1C.DiscussionTo the best of our knowledge, this study is the first overview 

hypoglycemia 
26729 
bias analysis by Begg’s funnel plots.A. General hypoglycemia. B. Severe hypoglycemia. C. Nocturnal hypoglycemia . D A1C.DiscussionTo the best of our knowledge, this study is the first overview of systematic reviews 

hypoglycemia 
27270 
insulin analogues are efficacious compared to NPH, with estimates showing a reduction in nocturnal hypoglycemia episodes (RR 0.66) and general hypoglycemia episodes (RR = 0.95). Further results showed a nonclinically 

hypoglycemia 
27314 
to NPH, with estimates showing a reduction in nocturnal hypoglycemia episodes (RR 0.66) and general hypoglycemia episodes (RR = 0.95). Further results showed a nonclinically significance reduction in A1C.Our study 

hypoglycemia 
27667 
[[28],[30]]. Of the 11 reviews included in this overview, only three [[28],[30],[31]] reported the measure of hypoglycemia as episodes per persontime, as recommended by methodological guidelines [[15],[68]]. The other reviews 

hypoglycemia 
27827 
methodological guidelines [[15],[68]]. The other reviews [[10],[11],[26],[29],[32]–[34]] reported hypoglycemia as the risk of having at least one event during the study (number of people who had hypoglycemia / number 

hypoglycemia 
27924 
reported hypoglycemia as the risk of having at least one event during the study (number of people who had hypoglycemia / number of people in the study group). This is a misconception, given that each person could experience 

hypoglycemia 
28651 
and collaborators [[30]], in a Cochrane review, yielded comparable results to our overview concerning hypoglycemia . Results showed the reduction was not statistically significant in general hypoglycemia (OR 0.93; 95% 

hypoglycemia 
28739 
concerning hypoglycemia. Results showed the reduction was not statistically significant in general hypoglycemia (OR 0.93; 95% CI 0.8; 1.08) (30). For severe hypoglycemia, for which overview results were not statistically 

hypoglycemia 
28797 
not statistically significant in general hypoglycemia (OR 0.93; 95% CI 0.8; 1.08) (30). For severe hypoglycemia , for which overview results were not statistically significant, Vardi and colleagues showed a reduction 

hypoglycemia 
28968 
significant, Vardi and colleagues showed a reduction of 27% (OR 0.73; 95% CI 0.61; 0.87) and for nocturnal hypoglycemia , the results were in the same direction, with a significant reduction favouring insulin analogue (OR 

hypoglycemia 
29204 
0.79) [[30]]. The results of Singh and colleagues’ review [[31]] are very similar to ours. Nocturnal hypoglycemia was reduced in insulin analogues in 35% (OR 0.65; 95% CI 0.55;0.77) [[31]].Regarding clinical trials, 

hypoglycemia 
29465 
are usually misunderstood. Considering dose titration phases, the proportion of patients experiencing hypoglycemia episodes during the titration period was higher for insulin analogues than for NPH in the first trials 

hypoglycemia 
29835 
analogues. Thus, a substantial number of hypos in the analogues group was noted. In conclusion, neither hypoglycemia episodes nor A1C in the titration period should be included in the results, according to Food and Drug 

hypoglycemia 
30276 
studied” [[69]].Two other major challenges when analyzing RCTs in T1D are the different ways of measuring hypoglycemia and the short followup of studies. Regarding the definition of hypoglycemia, this varies across the 

hypoglycemia 
30353 
different ways of measuring hypoglycemia and the short followup of studies. Regarding the definition of hypoglycemia , this varies across the studies (<2.0 mmol / L, <2.8 mmol / L, <3.1 mmol / L, <4.0 mmol / L), contributing 

hypoglycemia 
30509 
/ L, <2.8 mmol / L, <3.1 mmol / L, <4.0 mmol / L), contributing 14% to the heterogeneity of general hypoglycemia in our analysis. Recommendations from the American Diabetes Association Workgroup on Hypoglycemia [[72]] 

hypoglycemia 
30729 
and the International Society for Pediatric and Adolescent Diabetes [[73]], which stratify and define hypoglycemia appropriately, have been published. However, many of the studies, published later than these recommendations, 

hypoglycemia 
32560 
primary role in the algorithm for treating T1D is to improve glucose control and decrease the risk of hypoglycemia . Therefore, since A1C is an average of the glycaemia of the last three months, we can see the fact that 

hypoglycemia 
32719 
glycaemia of the last three months, we can see the fact that these insulins decrease the occurrence of hypoglycemia (lower limit of the mean) as positive, providing maintenance or even little improvement of A1C [[74]], 

hypoglycemia 
33087 
most explained the heterogeneity were bolus insulin (26%) and definition of hypo (14%) for general hypoglycemia , crossover design (23%) and intervention brand (13%) for nocturnal hypoglycemia, and Jadad score (16%) 

hypoglycemia 
33168 
(14%) for general hypoglycemia, crossover design (23%) and intervention brand (13%) for nocturnal hypoglycemia , and Jadad score (16%) for A1C. Two other factors have influenced the inconsistency of the results, 