Article Text

Download PDFPDF

Early postnatal hyperglycaemia is a risk factor for treatment-demanding retinopathy of prematurity
  1. Carina Slidsborg1,
  2. Louise Bering Jensen1,
  3. Steen Christian Rasmussen2,
  4. Hans Callø Fledelius1,
  5. Gorm Greisen3,
  6. Morten de la Cour1
  1. 1 Department of Ophthalmology, Copenhagen University Hospital, Rigshospitalet, Copenhagen, Denmark
  2. 2 Department of Clinical Microbiology, Copenhagen University Hospital, Hvidovre Hospital, Hvidovre, Denmark
  3. 3 Department of Neonatology, Copenhagen University Hospital, Rigshospitalet, Copenhagen, Denmark
  1. Correspondence to Carina Slidsborg, Department of Ophthalmology, Rigshospitalet, Nordre Ringvej 57, 2600 Glostrup, Denmark; carinaslidsborg{at}hotmail.com

Abstract

Background To investigate whether neonatal hyperglycaemia in the first postnatal week is associated with treatment-demanding retinopathy of prematurity (ROP).

Methods This is a Danish national, retrospective, case–control study of premature infants (birth period 2003–2006). Three national registers were searched, and data were linked through a unique civil registration number. The study sample consisted of 106 cases each matched with two comparison infants. Matching criteria were gestational age (GA) at birth, ROP not registered and born at the same neonatal intensive care unit. Potential ‘new’ risk factors were analysed in a multivariate logistic regression model, while adjusted for previously recognised risk factors (ie, GA at birth, small for gestational age, multiple birth and male sex).

Results Hospital records of 310 preterm infants (106 treated; 204 comparison infants) were available. Nutrition in terms of energy (kcal/kg/week) and protein (g/kg/week) given to the preterm infants during the first postnatal week were statistically insignificant between the study groups (Mann-Whitney U test; p=0.165/p=0.163). Early postnatal weight gain between the two study groups was borderline significant (t-test; p=0.047). Hyperglycaemic events (indexed value) were statistically significantly different between the two study groups (Mann-Whitney U test; p<0.001). Hyperglycaemia was a statistically independent risk factor (OR: 1.022; 95% CI 1.002 to 1.042; p=0.031).

Conclusion An independent association was found between the occurrence of hyperglycaemic events during the first postnatal week and later development of treatment-demanding ROP, when adjusted for known risk factors.

  • Preterm birth, Retinopathy of Prematurity
  • Risk factors
  • Hyperglycemia.

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Introduction

During the last decade, independent studies have documented that growth retardation (intrauterine/postnatal) is an important risk factor for retinopathy of prematurity (ROP).1–4 A recent Swedish screening algorithm (WINROP) predicted development of treatment-demanding ROP based on postnatal growth retardation and insulin-like growth factors I (IGF-I) levels.1 To combat growth retardation, international guidelines currently advise that neonatal intensive care units (NICUs) should administer intensified therapeutic feeding regimens for the preterm infants.3

Neonatal hyperglycaemia is a common problem in extremely preterm infants.5 6 The pathophysiology of neonatal hyperglycaemia is very complex.7 The condition is related to a developmental immature hepatic glucose production, an inadequate pancreatic response and insulin insensitivity.8 There is also growing concern about the programming of the cells involved in glucose regulation.9–11 Cell programming in preterm infants, in the extrauterine environment, possibly affects the long-term health of the infants.9–11 In fact, a few recent studies documented that neonatal hyperglycaemia is associated with development of ROP.12–18 It remains unclear whether neonatal hyperglycaemia also predicts the development of the severe stages of ROP.13 15 18

In Denmark, during the last decade, an increase in the incidence of treatment-demanding ROP occurred.2 Neither known risk factors, nor a shift in indication for treatment, could fully explain the increase.2 19 During the same birth period, in Denmark, as in other countries, an intensified therapeutic feeding regimen was introduced to the preterm infants. This was to improve the long-term growth and neurodevelopment of the infants. This new feeding approach possibly induced more metabolic instability and more hyperglycaemic events and finally more treatment-demanding ROP cases during the study period.

This study is based on a large national study sample of ROP-treated infants born during the birth period where most cases of ROP-treated infants were born. The hypothesis of the study was that neonatal hyperglycaemic events in the first postnatal week are associated with treatment-demanding ROP.

Materials and methods

This is a national, retrospective, case–control study of premature infants born in Denmark in the period 2003–2006. Three national registers were involved in this study: the National Register of Rigshospitalet (NRR), the National Birth Register (NBR) of the National Board of Health and the National Patient Register (NPR) of the National Board of Health. Data from the different registers were linked through a unique Civil Registration Number.

Identification of cases/comparison infants

In Denmark, ROP treatment is centralised to the ophthalmology department at Rigshospitalet, Copenhagen University Hospital. The treated infants were, therefore, identified in the hospital registry, the NRR. Combinations of the ROP ICD10 code, the codes for laser/cryo treatment, as well as codes for vitrectomy, and scleral buckling surgery were used to detect the infants. Local surgery lists from the ophthalmology department were also surveyed to verify sample completeness (100%). A sample of 106 infants having had ROP treatment was obtained.

In Denmark, physicians and midwives are obliged to report every live-born child to the NBR, and reporting is considered almost complete (99.7%).2 19 20 For all preterm infants born during the study period, information on gestational age (GA) at birth, birth weight (BW), gender, multiple births (MB) and, when relevant, date of death, whether diagnosed with ROP were obtained from the NBR. Information on hospital departments where the infants were born and subsequent admissions to other hospital departments was obtained from the NPR. Each of the 106 preterm cases was matched randomly with two comparison infants having a similar GA (preferably within 1 week of birth), no diagnosis of ROP registered and being born at the same NICU. If the comparison infant had moved to another hospital department within the first day after birth, the second department was used as a matching criterion. Records on eight comparison infants (8/212; 3.8%) were missing. The remaining comparison group consisted of 204 preterm infants, and the final study sample consisted of 310 preterm infants.

Data retrieval and registration

All 17 NICUs in Denmark involved in handling the infants from the study sample were contacted in order to retrieve relevant data from the local medical records. Relevant data were entered into an ACCESS database. GA, BW, MB, gender, regain of BW (postnatal days), head circumference (cm), birth length (cm), death (<3 postnatal months of life), postnatal weights (g), nutrition (energy kcal/kg; protein g/kg) and hyperglycaemia (number of events/day) were registered. Nutritional data and data on hyperglycaemia were obtained from the medical records where they are routinely registered by the nurse at the moment that the nutrition is given and the blood glucose level is measured. Both enteral and parenteral nutrition were given from the day of birth. The enteral nutrition consisted of breast milk, donor milk, milk fortifier or infant formula through a nasogastric tube. Although very preterm infants are breastfed from early age, they usually do not receive significant amounts of breast milk during the first postnatal weeks. So this parameter was disregarded. Two NICUs represented 57.0% (32/56) of the missing medical records. The remaining NICUs lacked only a few medical records.

As in other comparable studies, neonatal hyperglycaemia is herein defined as blood glucose of more than 8.5 mmol/L (or more than 150 mg/dL).12–14 17 18 21 For each of the infants, a SD score of BW for a given GA was calculated.22 If the recorded BW was <76% of the estimated mean BW; the entry was labelled small for gestational age (SGA).22

Data analyses

All data sets were investigated for normality to determine whether parametric or non-parametric methods should be applied. To compare two groups, if the data were normally distributed, then the unpaired t-test was used; if not, the Mann-Whitney test was used. Potential ‘new’ risk factors were analysed in a multivariate logistic regression model, while adjusted for known risk factors such as GA, SGA, MB and male sex. The significance level of the estimated parameters of the multiple logistic regression analysis was assessed by the use of the Wald Statistic. Throughout we used a CI of 95% and a significance level of 5%. Statistical analysis was performed by using Sigma Stat (Systat Software, San Jose, California).

Results

Descriptive data

Hospital records of 310 preterm infants (106 treated; 204 comparison infants) were available. Rarely data were missing when the record was available. The treated infants and comparison infants were matched on their GA and the final hospital admission within the first 24 hours after birth. Table 1 shows the study characteristics of preterm population, divided into ROP-treated group and the comparison group.

Table 1

Study characteristics

It is apparent from table 1 that matching on immaturity was incomplete. The sample of treated infants was generally more immature, smaller, had smaller head circumference and smaller birth length and compared with the comparison infants. The regain of birth weight was faster in the treated infants compared with the comparison infants. A larger proportion of the treated infants were intrauterine growth retarded, males and part of MB. In nearly all cases, the hospital on which the infant was born was used as matching criterion. One preterm infant died before the first 12 postnatal weeks. This infant had ROP treatment, and exclusion was not needed.

Evaluation of ‘new’ risk factors for treatment-demanding ROP

Figure 1A,B show box-plots of the amount of nutrition in terms of energy (kcal/kg/week) and protein (g/kg/week) given to the preterm infants during the first postnatal week. Neither of these parameters were significantly different between the treated and the untreated group (Mann-Whitney U test; p=0.165/p=0.163). Figure 1C shows a box-plot of the early postnatal weight gain (g/week) between the two groups. There is borderline statistical significance between the two groups (t-test; p=0.047). Figure 1D shows a box-plot of the hyperglycaemic events occurring within the first postnatal week divided into two study groups. The number of hyperglycaemic events was statistically significant between the two study groups (Mann-Whitney U test; p<0.001).

Figure 1

(A and B) Box-plots of the amount of nutrition in terms of energy (kcal/kg/week) and protein (g/kg/week) given to the preterm infants during the first postnatal week. No difference is found between the two groups. (C) Box-plot of the early postnatal weight gain (g/week) between the two groups. There is a tendency towards an increased weight gain in the treated compared with the untreated group. Finally (D) shows a box-plot of the hyperglycaemic events occurring in the two study groups. The number of hyperglycaemic events was higher in the treated compared with the untreated group.

Figure 2A shows the mean number of days of missing blood sugar measurements, divided into two study groups, and represented per day during the first postnatal week. There is no statistically significant difference in the number of days with missing blood glucose measurements, between the two study groups (Mann-Whitney U test; p=0.660). Figure 2B shows the mean number of blood sugar measurements performed per day, divided into two study groups and represented per day during the first postnatal week. During this postnatal period, the treated infants had more blood glucose measurements performed, compared with the comparison infants. The difference was statistically significant (Mann-Whitney U test; p=0.007).

Figure 2

(A) The mean number of days of missing blood sugar measurements, divided into two study groups and represented per day during the first postnatal week. There is no significant difference between them (B) show the mean number of blood sugar measurements performed per day, divided into two study groups and represented per day during the first postnatal week. During this postnatal period, the treated infants had more blood glucose measurements performed, compared with the comparison infants.

The number of hyperglycaemic events per infant was transformed into a hyperglycaemic index that takes into account the different number of blood glucose measurements performed in the individuals (during the first postnatal week): the individual number of hyperglycaemic events during the first postnatal week/total number of blood measurement performed in the same infant during the first postnatal week. The hyperglycaemic index remained statistically significantly different between the two study groups (Mann-Whitney U test; p<0.001).

Multivariate logistic regression

Table 2 presents the results of the multivariate logistic regression analysis performed to identify whether hyperglycaemia is an independent predictor of treatment-demanding ROP.

Table 2

Multivariate logistic analysis

All statistically significant risk factors were included in the final model, that is, GA, SGA, males and presence of hyperglycaemia. Estimates are reported as odd ratios with 95% CI. This model shows that the initial matching on GA was incomplete.

However, after adjustment for known risk factors, the neonatal hyperglycaemic index remained a statistically independent risk factor for development of treatment-demanding ROP (OR: 1.022; 95% CI 1.002 to 1.042; p 0.031). Estimates for MB were excluded in the final model as the results were shown to be insignificant.

Discussion

There is rising evidence that the presence of neonatal hyperglycaemia increases the risk of mortality and certain morbidities, such as necrotisizing enterocolitis, brain damage and ROP.12–18 21 23 24 A few retrospective previous studies found that neonatal hyperglycaemia during the first postnatal month predicted the development of ROP.12–15 17 18 A recent prospective study by Mohsen et al 21 found a similar early association. However, possibly due to the small sample size, the study was unable to conclude on any association between neonatal hyperglycaemia and ROP treatment. To date, it is still controversial whether there is an association between hyperglycaemia and development of treatment-demanding ROP.13 15 18 25

This study further shows that neonatal hyperglycaemia occurring as early as the first postnatal week predicts much later development of treatment-demanding ROP. The statistically significant association persists even when adjusted for known risk factors (ie, GA, SGA, MB and male sex). This is important as hyperglycaemia might be dealt with by medical intervention. However, at present, no evidence-based agreement exists on whether or not to treat neonatal hyperglycaemia with insulin.17

Neonatal hyperglycaemia and the occurrence of ROP share quite a few risk factors such as sepsis (especially Candida sepsis), necrotisizing enterocolitis, acute intracerebral haemorrhage, postnatal steroid, some vasoactive drugs and theophylline.7 26 These risk factors could all be confounders, and neonatal hyperglycaemia the most important risk factor of ROP.

Several international studies have shown that postnatal growth retardation is related to certain morbidities, such as ROP development.1 3 4 Therefore, in Denmark, to fasten postnatal weight gain, the physicians of the NICUs intensified therapeutic feeding regimens of the preterm infants. In this study, an association was not found between the intake of nutrition (energy/kcal; protein/g) during the first postnatal week and the development of treatment-demanding ROP. However, we observed a borderline tendency towards increased weight gain in the treated group when compared with the comparison group. The amount of missing data on nutrition (18.1%; 56/310) and early weight gain (7.1%; 22/310) makes establishing firm conclusions on these parameters difficult.

It is a strength of this study that the entire national sample of infants treated for ROP during a 4-year birth period (2003–2006) was available. The overall quality of data from the involved registers is considered good. The hospital lists of treated infants retrieved from the ophthalmology department database and at the RH database corresponded fully. The NBR also is considered complete and highly reliable.2 The register holds nearly all live-born infants in Denmark (ie, >99.7%). A previous study found that only 3.7% of the birth-related entries contained obvious errors. Data on the NPR are recognised to be used for research. In Denmark, it is demanded by law that only qualified hospital staff report diagnosis/treatment/surgery codes to the NPR. The registration to the register usually occurs within a few days after registration to the local hospital registers. The NPR which is administered by the National Board of Health (NBH) holds a research unit on which experts assist the researcher to obtain all the relevant information.2 19 20 27–29 The linking of data from the different registers through an individual civil registration number ensured data consistency.

The retrospective design of this study has the following inherent limitations: (1) hospital records on eight comparison infants were not available and therefore excluded (3.8%). This small amount of missing data is not expected to have any effect on the conclusions reached; (2) for practical reasons, hyperglycaemic events, rather than the duration of the hyperglycaemic event, were herein investigated. Duration of exposure to hyperglycaemia might, in fact, be a more relevant parameter to investigate. The impact that hyperglycaemia have on ROP development might turn out to be more important than was shown herein. (3) During the first postnatal week, only a few days had missing measurements of blood glucose, and no statistically significant difference was found between the two study groups. From the second postnatal week and onward, there was an increasing difference between the two groups, and these data were therefore not included in the present study. (4) It appears that the treated infants that had significantly more blood glucose measurements performed compared with the comparison infants. This might be a result of more events of hyperglycaemia detected in the treated group. To adjust for this potential bias of the study results, a hyperglycaemic index was used (cf. Materials and methods section). (5) The inability to obtain a complete data set on weight (7.1% missing) and especially nutritional (energy/protein) components (18.1% missing) could affect the conclusions reached on these two parameters. (6) This national study was at its outset designed to evaluate associations between growth, nutrition and hyperglycaemia and ROP while adjusting for the few risk factors that there are general consensus about. Due to the study design and the sample size, it was not possible to adjust for other potential risk factors for ROP in the statistical model.7 26 30

In conclusion, an independent association was found between the occurrence of hyperglycaemic events during the first postnatal week, and later development of treatment-demanding ROP, when adjusted for known risk factors. Future studies, preferably prospective, should confirm the conclusions herein reached while adjusting for other potential risk factors.7 26 It would be interesting to evaluate whether blood glucose fluctuation, or the total duration of hyperglycaemia, are more closely associated to ROP development than counted hyperglycaemic events.

References

Footnotes

  • Contributors CS contributed towards the conception and design of the work; the acquisition, analysis and interpretation of data for the work; drafting the work; final approval of the version to be published and agreed to be accountable for all aspects of the work. LBJ contributed towards the acquisition; drafting the work; final approval of the version to be published and agreed to be accountable for all aspects of the work. SCR contributed towards acquisition of data for the work; drafting the work; final approval of the version to be published and agreed to be accountable for all aspects of the work. HCF contributed towards the conception and design of the work; drafting the work; final approval of the version to be published and agreed to be accountable for all aspects of the work. GG contributed towards the conception and design of the work; analysis and interpretation of data for the work; drafting the work; final approval of the version to be published and agreed to be accountable for all aspects of the work. MdlC contributed towards the conception and design of the work; analysis and interpretation of data for the work; drafting the work; final approval of the version to be published and agreed to be accountable for all aspects of the work.

  • Funding This work was supported by grants from Danish Eye Health Society, Bagenkop Nielsens Myopi- and Eye Foundation, Velux foundation, Aase and Ejnar Danielsens Foundation, Dagmar Marshalls Foundation, Direktør Jacob Madsen and Hustru Olga Madsens Foundation, P. A. Messerschmidt and Hustrus Foundation.

  • Competing interests None declared.

  • Patient consent Not obtained as patients were not directly involved.

  • Ethics approval This study was approved by the Danish Data Protection Agency.

  • Provenance and peer review Not commissioned; externally peer reviewed.