Cumulative diagnostic imaging radiation exposure in premature neonates

Abbreviations

BW

Birth weight

1 Introduction

There have been remarkable advances in the care of extremely premature (PT) and extremely low birth weight (ELBW) infants over the past few decades. These advances have led to a significant improvement in the survival rate of premature infants and a lowering of the age of viability to as early as 22 weeks gestation in most of the very specialized neonatal intensive care units worldwide. Unfortunately, premature infants typically have extended hospital stays and often suffer multiple medical and surgical complications of prematurity that are directly proportional to the degree of prematurity and birth weight [1]. A premature infant’s hospital course almost always includes exposure to ionizing radiation in the form of diagnostic imaging in order to help establish a diagnosis and to monitor for complications of prematurity. The growing use of ionizing radiation imaging procedures in the United States has raised concerns about radiation dose in the Neonatal Intensive Care Unit (NICU) given the stochastic no-threshold nature of radiation-induced carcinogenesis and genetic damage to premature infant’s cells [1–4]. In addition, there is a unique developmental effect from radiation exposure during early life that can hinder complete maturation or development of tissues or organs [4, 5].

There are, to date, no available long-term follow up clinical studies on this topic. In addition, the frequency of ionizing imaging procedures performed on NICU patients varies by medical practice patterns, hospital setting and by the acuity of the patient population. As such, an understanding of the total cumulative effective ionizing radiation dose (cED) may help justify policies and practice procedures moving forward. Our main objective was to provide an estimate of the total cED received by PT infants ≤32 weeks gestation from diagnostic imaging radiation studies performed throughout their hospital stay in a level IV NICU in the United States, and to explore potential risk factors and predictors of such exposure.

2 Methods

Human Investigation Committee approval was obtained for this study. The NICU at Yale-New Haven Children’s Hospital is a 54-bed level IV NICU, with approximately 800 to 900 NICU admissions per year. Data were collected retrospectively on all inborn infants who were ≤32 weeks gestation admitted to the NICU from 2004 to 2011. Subjects were identified through our NICU admission database. Data collection included demographics, gestational age (GA), birth weight (BW), sex, method of delivery, hospital course, problem list, length of stay (LOS) and all imaging studies involving ionizing radiation that were performed during their entire hospital stay. We excluded infants transferred to another hospital prior to readiness for hospital discharge due to the lack of details concerning diagnostic imaging radiation studies performed after transfer. We also excluded infants who did not survive to hospital discharge to avoid underestimating the radiation dose received by the infant cohort. Infants who required transfer to different services within the hospital, such as the surgical unit, Pediatric Intensive Care Unit, and the Respiratory Care Unit were included in the study until their first hospital discharge.

Diagnostic studies involving ionizing radiation performed during our study period included computed radiographs (CR) of chest, abdomen, skull, spine, pelvis, extremity, soft tissue, shunt series, and skeletal surveys. Computed Tomography (CT) scans of the brain, chest and abdomen were included. Fluoroscopy studies included: upper and lower gastrointestinal contrast studies, airway fluoroscopy, esophagogram, cystogram, fistulagram, and fluoroscopy for central line placement. Radioisotope scans of the lungs, thyroid, hepatobiliary, renal, and gastric emptying scans were included.

The effective dose is a measure designed to represent the overall stochastic effects of a radiation exposure and is calculated by applying a weighting factor to each exposed organ during the exposure, with radiosensitive organs (e.g. lungs, bone marrow, stomach, breast) receiving the highest weighting [6].

The estimated effective dose from each radiological study performed was obtained from the published literature [6–19] and used to calculate the cED for each infant (Table 1). The sum of radiation doses across all organs and tissue represents the total radiation dose to the exposed person. The literature, however, has limited or no references to the specific neonatal diagnostic procedures listed in Table I. For instance, pediatric nuclear medicine procedure guidelines [20] for injected doses are given for age groups from newborn to 5 years but not for neonates, so we used the recommended injection activities in the guidelines for newborns to calculate the effective doses for neonates and assumed an average of 1 kg neonatal weight. Comments in Table I explain how effective doses were calculated when a reference does not address the specific procedure in the list.

3 Results

A retrospective chart review was conducted on all infants ≤32 weeks gestation born at Yale University Hospital and admitted to the NICU from 2004 to 2011. A total of 1045 patients met our inclusion criteria.

The median GA for the study subjects was 30.0 weeks (SD = 2.7, range 22.0–32.6 weeks) and the median BW was 1340.0 grams (SD = 445.4, range 420–2470 grams). The ethnic representation of study subjects was 50% White, 31% African American, 16% Hispanic and 3% Asian. Fifty-three percent of our study patients were male, and 47% were female (Table 2). Clinical diagnoses included the common neonatal problems found in the NICU. Forty-eight percent of the patients had surfactant deficiency requiring surfactant administration, 33% had retinopathy of prematurity (ROP), 27% had an oxygen requirement at 36 weeks, 18% had neonatal sepsis, 15% had intraventricular hemorrhage (IVH), 8% had necrotizing enterocolitis (NEC), 2% had periventricular leukomalacia (PVL), and 2% had spontaneous intestinal perforation.

There was a total of 22,117 images done on the 1,045 study patients, with an estimated sum total cED of 544,224μSv across all study patients. The vast majority (98%) of images were either chest or abdominal X-rays. Although chest X-rays accounted for 70% of the images done, they only constituted 41% of the total cED (Table 3). On the other hand, lower gastrointestinal contrast studies (Barium enemas) and upper gastrointestinal fluoroscopy images combined accounted for less than 1% of the images done, yet they represented 18% and 11% of the total cED, respectively.

The number of radiographic studies per infant ranged from 0 –210 with a median of 9 images and a standard deviation of 28.5. The median total cED was 162μSv (range 0 –9248μSv). The distribution of the total cED was highly right-skewed due to the presence of high outliers (Fig. 1). The total cED exceeded 500μSv in 283 (27%) of patients and exceeded 1000μSv in 167 (16%) of patients. The cED was positively associated with length of stay (r_s = 0.759, p < 0.001) as would be expected, but the cED was inversely correlated with GA (r_s = –0.620, p < 0.001) and BW (r_s = –0.599, p < 0.001). The median cED by GA was 1132μSv for GA 22–24 weeks, 462μSv for GA 25–28 weeks and 82μSv for GA 29–32 weeks.

OPEN IN VIEWER

Fig. 1

Distribution of the 1045 study patients’ total Cumulative effective radiation dose (cED) in microSieverts (μSv).

The median cED by co-morbidity was 434μSv for infants who received surfactant at any time vs. 71μSv for infants who did not (Z = 17.0, p < 0.001), and 714μSv for infants who received surfactant at the delivery room vs. 115μSv for infants who did not (Z = 12.8, p < 0.001). The median cED was 1,065μSv for infants with oxygen requirement at 36 weeks gestation vs. 119μSv for others (Z = 15.0, p < 0.001) and 1,233μSv for infants who received steroids for bronchopulmonary dysplasia (BPD) vs. 132μSv for others (Z = 12.4, p < 0.001). Infants who had NEC received a median cED of 1228μSv vs. 136μSv for those without NEC (Z = 10.0, p < 0.001), and 990μSv for infants who were diagnosed with PVL compared to 162μSv for others (Z = 3.7, p < 0.001). The median cED in infants diagnosed with sepsis was 607μSv vs. 72μSv for others (Z = 2.9, p = 0.002), 796μSv for infants diagnosed with ROP vs. 165μSv for others (Z = 13.8, p < 0.001), and 622μSv for infants with IVH compared to 164μSv for others (Z = 7.8, p < 0.001). The highest median cED was noted in infants diagnosed with intestinal perforation with a median cED of 1661μSv compared to 162μSv for others (Z = 10.0, p < 0.001).

Forty (3.8%) of our subjects had at least one CT scan during their hospital stay, of whom 32 (80%) had a CT scan of the brain, 8 (20%) had a CT scan of the chest, 3 (8%) had a CT scan of the abdomen and 3 (8%) had a CT scan of the neck.

For multiple linear regression analysis, the 13 initial candidate variables considered for inclusion in the regression model to predict cED were length of stay, GA, BW, surfactant at any time, surfactant at the delivery room, oxygen requirement at 36 weeks gestation, steroids for BPD, NEC, PVL, sepsis, ROP, IVH and intestinal perforation. After stepwise variable selection, the 7 variables remaining in the model were LOS, GA, oxygen requirement at 36 weeks, steroids for BPD, NEC, PVL and intestinal perforation. These 7 variables as a group exhibited statistically significant predictive accuracy (F_7,931 = 128.1, p < 0.001) and explained 49.1% of the variation in cED. After controlling for the other variables in the regression model, each additional day of LOS increased the predicted cED by 10.7μSv (p < 0.001), and higher GA was associated with a higher predicted cED by 29.8μSv (p = 0.018). However, a highly significant inverse correlation between GA and LOS was noted (r_s = –0.822, p < 0.001), so before interpreting the counter-intuitive positive association between GA and cED in this model, we should point out that a strong correlations between independent variables causes problems when fitting a multiple linear regression model.³¹ When LOS was removed from the regression model to address collinearity, the intuitive inverse relationship between GA and cED was manifested, with each additional week of gestational age resulting in a predicted decrease in cED of 70.0μSv, which was a highly significant negative association (p < 0.001). After controlling for the other variables in this regression model, infants requiring oxygen at 36 weeks GA had a predicted cED that was 605.1μSv higher than infants who did not require oxygen at 36 weeks GA (p < 0.001). Infants receiving steroids for BPD had a predicted cED that was 474.1.6μSv higher than infants not receiving steroids for BPD (p < 0.001). Infants with PVL had a predicted cED that was 471.4μSv higher than infants without PVL (p = 0.009). Infants with NEC had a predicted cED that was 1119.4μSv higher than infants without NEC (p < 0.001). Lastly, infants with intestinal perforation had a predicted cED that was 1273.9μSv higher infants without intestinal perforation (p < 0.001). However, examination of the residual quantile-quantile plot from the fitted regression model for regression diagnostics revealed that the residuals were not normally distributed, and the scatter plot of the residuals vs. predicted cED indicated marked heteroscedasticity, with larger errors as the predicted cED increased. In an effort to satisfy the linear regression assumptions of homoscedasticity and normally distributed errors, a Box-Cox analysis was performed to iteratively search for the optimal power transformation, which resulted in a transformation of cED^1/4. When the regression model was refit using cED^1/4 as the response variable, the model fit improved with the coefficient of determination increasing from R² = 0.491 when using the untransformed cED as the response variable to R² = 0.560 using cED^1/4 as the response. In checking the linear regression assumptions, it was found that the quantile-quantile plot of the residuals was greatly improved after the transformation and the skewness coefficient went from a value of 2.835 for the untransformed cED to a value of 0.006 for the transformed cED^1/4. (A normal distribution has a skewness coefficient of zero.) Likewise, the heterogeneity of the residuals did not increase with increasing values of the predicted cED after the transformation, as was the case before the transformation, thereby, indicating the assumption of homoscedasticity was satisfied. Multicollinearity was not an issue after the transformation, with the highest Variance Inflation Factor (VIF) value being 3.4, well below the threshold of 10 considered indicative of a problem [23], and none of the tolerance values were below 0.10 [24]. Steroids for BPD was no longer significantly associated with the cED received after the transformation, but the other predictors in the regression model were significant both before and after the transformation (Table 4).

There is growing concern for potentially significant public health problems accompanying the ever-increasing use of diagnostic imaging studies involving ionizing radiation in the expanding population of vulnerable preterm infants. As NICUs across the country increasingly care for a growing population of extremely premature patients, the use of diagnostic imaging procedures becomes inevitable to save infants’ lives. Unfortunately, the youngest most vulnerable preterm neonates are found to have the highest radiological exposure dose during the most sensitive time of their development. Although not entirely unexpected, this issue remains a grave concern for those of us caring for these patients.

Ionizing radiation exposure from medical procedures now represents the major source from which individuals in the USA are exposed [25, 26]. This is particularly so for premature infants who typically have extended stays in the neonatal intensive care unit. During this relatively short time they typically require multiple diagnostic imaging studies that involve ionizing radiation exposure such as CR, contrast studies and CT scans. The inherent vulnerability of premature infants to environmental insults, the higher radiosensitivity of their developing organ tissues, up to 15 times as radiosensitive to radiation as adults [27], their extended stay in the NICU and their longer lifespan following exposure all increase their risk from the cumulative effects of ionizing radiation exposure. In addition, their smaller size results in a higher exposure to the effective dose conversion factor per radiograph than older children or adults [28].

Ionizing radiation can cause DNA damage or cell death directly and indirectly via the formation of free radicals and damage to essential cell enzymes [29]. In addition, radiation exposure is a well-known inducer of cellular stress responses [30], and the capacity of these cell pathways to resolve damage in neonates is largely unknown.

The probability of occurrence of radiation side effects depends on the total radiation dose received over time, genetic predisposition and other environmental risk factors [27, 31]. The studies of atomic bomb survivors concluded that the risk of all solid cancers had a linear relationship with the radiation dose received and showed that children were much more radiosensitive than adults [31, 32].

Our results showed that the BW and GA are inversely correlated with cED, and the length of hospital stay is positively associated with the cED. In our regression model, each additional week of GA is negatively associated with the predicted cED, and each additional day of LOS increased the predicted cED. The co-morbidities with the highest estimated cED were intestinal perforation and NEC. Infants with intestinal perforation and NEC typically get subjected to serial X rays to monitor the progression of the disease, need for surgical intervention, or the resolution of radiological findings.

Our data simply highlights that radiation exposure in the NICU is highest among the most premature and the extremely low birth weight neonates, who are at the highest risk to suffer from the severe complications of prematurity and have the longest hospital stay. Only forty neonates (3.8% of our subjects) had one or more CT scans during their hospital stay, however, the cED of repeated exposures to relatively low dose CR studies equaled or exceeded that of a CT scan in 19 neonates (1.8% of patients), with a median GA of 30 weeks. The results raise a question of whether or not an acute exposure to relatively higher radiation doses from CT scans would have similar biological effects as frequent exposure to relatively low dose ionizing radiation from CR.

In the past, there has been only limited work undertaken on diagnostic radiation exposure to premature infants and of the few observational studies performed to date, most have concluded that the radiation dose received was in compliance with recommendations made by the Commission of European Communities and International Commission on Radiological Protection (ICRP) [1, 33].

Radiation exposure in the NICU is, of course, only the beginning of the radiation exposure those infants will receive during their lifetime. NICU graduates, especially those who suffer multiple complications during their NICU stay, often present to emergency rooms with ongoing problems such as acute life threatening events (ALTE), pneumonia, ventriculoperitoneal shunt complications, intestinal problems etc., requiring multiple additional ionizing radiation studies including frequent CT scans. It should also be borne in mind that doses associated with a given radiograph or CT scan may vary considerably between different protocols, machines, and institutions.

Adopting new imaging technologies and radiation dose managing methods will help reduce radiation doses. For instance, the use of specific pediatric and neonatal CT protocols at lower beam energy (kVp), tube current (mAs), higher pitch values and advanced reconstruction algorithms demonstrated tangible patient dose reduction [34]. The use of image gently recommendations specific to pediatric safety (www.imagegently.org) [35] with regard to gentle use of imaging have helped reduce pediatric doses. In general radiography, the use of tight collimation, and proper adjustment of kVp and mAs through auto-exposure methods should help lower doses as well.

5 Conclusion

Imaging procedures are the major source of ionizing radiation exposure to the premature neonatal population and have the potential, in certain patient groups, to result in high cED. Our data shows that this would appear to be increasingly the case in the NICU, and that this finding is most marked in extremely premature infants

The stochastic (neoplastic) effects of ionizing radiation exposure are believed not to have a dose threshold, i.e. the severity of the effect is independent of the dose, but the likelihood of their occurrence simply increases with increasing dose. It is important, therefore, to consider this when exposing infants to even a single low dose imaging study.

Without a doubt, diagnostic imaging studies can help save infants’ lives; nevertheless, radiation exposure should always operate under the “As Low As Reasonably Achievable” (ALARA) principle and the evaluation of the risk: benefit ratio should be continually applied in order to justify such exposures. As such, our data show that efforts to minimize ionizing radiation exposure in the NICU and further education as to the need to use non-ionizing radiation imaging studies (MRI or ultrasound) are warranted. Further investigation into the lifetime oncology risks in this relatively new group of PT and ELBW patients would be a fascinating realm for future research.

Funding Source

Dr. Khattab was supported by the National Institute of Child Health and Human Development (NICHD) grant number T32 HD 07094.

Disclosure Statement

All authors Dr. Khattab, Dr. Hagan, Dr. Staib, Dr. Mustafa, and Dr. Goodman have no financial disclosures or conflict of interest to disclose relevant to this study.