Metabolic Bone Disease in Children with Intestinal Failure and Long-Term Parenteral Nutrition: A Systematic Review

Metabolic bone disease (MBD) is a possible complication of intestinal failure (IF), with a multi-factorial pathogenesis. The reduction of bone density (BMD) may be radiologically evident before manifestation of clinical signs (bone pain, vertebral compression, and fractures). Diagnosis relies on dual-energy X-ray absorptiometry (DXA). Incidence and evolution of MBD are not homogeneously reported in children. The aim of this systematic review was to define the prevalence of MBD in IF children and to describe risk factors for its development. A comprehensive search of electronic bibliographic databases up to December 2021 was conducted. Randomized controlled trials; observational, cross-sectional, and retrospective studies; and case series published between 1970 and 2021 were included. Twenty observational studies (six case-control) were identified and mostly reported definitions of MBD based on DXA parameters. Although the prevalence and definition of MBD was largely heterogeneous, low BMD was found in up to 45% of IF children and correlated with age, growth failure, and specific IF etiologies. Data demonstrate that long-term follow-up with repeated DXA and calcium balance assessment is warranted in IF children even when PN dependence is resolved. Etiology and outcomes of MBD will be better defined by longitudinal prospective studies focused on prognosis and therapeutic perspectives.


Introduction
Intestinal failure (IF) is the consequence of several gastrointestinal conditions that determine the necessity of parenteral nutrition (PN) to maintain adequate growth and fluids/energy balance [1]. Short bowel syndrome (SBS) is the principal cause of IF in children; more rarely, IF is the result of a congenital enteropathy or a severe motility disorder. SBS generally follows a bowel resection in the neonatal period for a gastrointestinal anomaly (atresia, gastroschisis, and omphalocele) or an acquired event (neonatal volvulus or necrotizing enterocolitis). More rarely, SBS can develop in older children because of a surgical resection for several conditions, including inflammatory bowel disease (IBD), volvulus, and ischemic events [2]. The reversibility of IF and the achievement of enteral autonomy depend on the underlying condition and the surgical outcomes but are also related to both nutritional and medical treatments. PN has a crucial role in the different phases of IF management; in fact, PN represents the mainstay of treatment in the very early phase of SBS, a medium-term solution in reversible IF cases (when the process of intestinal adaptation progressively encompasses the bowel loss), and a life-long strategy in the most severe forms of IF. Advancement in home PN (HPN) and enteral nutrition (EN), both in surgical techniques and in medical strategies and particularly the development of

Information Sources and Search Strategy
A systematic literature search was conducted through electronic bibliographic databases, in particular PubMed, Science Direct, Cochrane Library, and ERIC. The following keywords were used to include the more relevant studies: "parenteral nutrition" or "parenteral" AND "bone" or "bone health" or "osteoporosis" or "osteopenia" AND "intestinal failure" or "short bowel". Moreover, references of the selected studies were screened to identify other relevant articles. The search was performed up to September 2021, and a last access was repeated in December 2021 in order not to exclude any updated study.

Study Selection, Quality Assessment and Data Extraction
Titles were collected by two reviewers (S.G., A.P.) who removed duplicates. Abstracts and full articles were individually reviewed by three authors (S.G., A.P., and S.Q.). All full-text articles that satisfied eligible criteria underwent a quality assessment through the Academy of Nutrition and Dietetics Quality Criteria Checklist (QCC) for Primary Research. The QCC includes fourteen questions (four relevance questions that address applicability to practice and ten validity questions) based on the Agency for Healthcare Research and Quality (AHRQ) domains for research studies [15]. Each question can be answered with "yes", "no", "unclear", or "not applicable (N/A)", and a rating of positive, neutral, or negative can be assigned depending on the answers. Quality assessment and data collection process was independently performed by three reviewers (S.G., S.Q., and A.P.) who checked separately for accuracy and consistency of each selected study. An extraction form on a specific worksheet using the Microsoft Excel 2016 software was created and used to extract data. In particular, the following data were collected from the selected studies: study details (design, participants, country, period), IF-or PN-dependence definition, participants (number, age, gender, length of PN, prevalence of ongoing PN), bone disease definition (based on DXA or other radiologic techniques and clinical judgment, i.e., prevalence of fractures), DXA parameters (BMC in grams, BMD in grams/cm 2 , bone mineral apparent density (BMAD) in grams/cm 3 , and z-scores and/or adjusted z-scores), prevalence of bone disease and/or comparison with control group, serum levels of markers of calcium metabolism or prevalence of vitamin D insufficiency and/or deficiency (serum and urinary calcium and phosphorus, vitamin D, parathormone, alkaline phosphatases, other markers), risk factors (and type of analysis) correlated to MBD or to DXA parameters (i.e., age, type of IF, PN duration, quantity or composition, levels of bone markers and vitamin D), and follow-up results. Prevalence of MBD was estimated by summing all the patients with at least a DXA parameter ≤−2 z-score (at first DXA scan) divided by the total number of patients having a DXA scan in the same cohorts. If different studies clearly included patients from the same cohort, prevalence estimates were only registered once for each cohort.

Study Selection
The systematic research identified a total of 650 articles. A total of articles were removed after title screening, and 114 abstracts were selected for eligibility. Seventy-three articles were full text screened and fifty-four excluded. One further article was included by reference snowballing. Twenty studies were included in the final review [16][17][18][19][20][21][22][23][24][25][26][27][28][29][30][31][32][33][34] (Figure 1). Nutrients 2022, 14, x FOR PEER REVIEW 4 of 24 All the selected papers underwent the QCC assessment and were considered suitable for data collection process and summary measures (Table 1). Table 2 summarizes the main characteristics of the included studies. All the selected papers underwent the QCC assessment and were considered suitable for data collection process and summary measures (Table 1). Table 2 summarizes the main characteristics of the included studies.        At univariate analysis, patients with BMD z-score ≤ −2 had lower WAZ (p = 0.01), lower Ca (p = 0.04), and higher PTH levels (p = 0.006).
At the multivariable logistic regression analysis, WAZ (R = 1.8, p = 0.03) and serum Ca (R = 3.8, p = 0.02) were independent factors of a low BMD z-score.
Thirteen studies reported definitions of bone disease based on DXA parameters (compared to reference values extrapolated from healthy reference populations matched for age, ethnicity, and gender). Four studies defined MBD as one of the DXA parameters including BMC, BMD, and BMAD below −1 z-score [20,23,25,26]. Seven studies indicated as MBD one of the previous parameter inferiors to −2 z-score [22,24,27,29,30,33,34], while two studies considered both the cut-off values [31,32]. An adjustment of DXA parameters for bone characteristics (bone age or bone size) or for anthropometric measures (ideal weight or height or statural age) was considered by seven studies [22,23,25,29,30,33,34].

Prevalence of MBD and Comparison with Control Population
Data on prevalence of MBD in children on long-term PN were largely heterogeneous, reflecting the different definitions used by the researchers. Prevalence of MBD (defined as BMD z-score ≤ −2) varied from 12.5%, reported by Ubesie [22] (n = 80), to 45% of IF children, reported by Poinsot [30] (n = 31). A total estimated prevalence of MBD of 28.8% (100/347 IF children) was calculated considering data from the nine studies using the cutoff of −2 z-score in at least one of the DXA parameter. A BMD z-score ≤ −1 was described in the vast majority of IF patients, with the highest prevalence reported by Diamanti [20] (83%, n = 24) and Olszewska [31] (87.5%, n = 17). BMD was found significantly reduced in IF children compared to controls in three studies [20,24,34], in a total of 55 IF children compared to 93 controls. This difference was not found by Dellert [19] in 1998, when BMC values (from 18 IF children and 36 controls) were adjusted for height and weight. Evidence of osteopenia based on standard radiography varied from 37.5% in the small study by Cannon [16] (n = 8), 59% in the study by Wozniak [28] (n = 27), and to 85.7% in the case series by Larchet [17] (n = 7).
Dependent on different cut-off and definitions, prevalence of vitamin D deficiency varied between 33% [29] and 63.8% [26] of IF children. Interestingly, no difference in serum levels of vitamin D were found in IF children compared to controls in one study [34], and a lower prevalence of vitamin D insufficiency in IF subjects was shown in another study [24].

Evolution of MBD and Risk Factors for MBD
Longitudinal studies based on repeated DXA studies reported data on 105 IF children (representing the 58% of the baseline cohorts of those studies) assessed after a period ranging from 1 to 6.2 years [20,24,26,29,30]. The three studies by Diamanti [20], Neelis [29], and Poinsot [30] (total n = 53) described an improvement of DXA parameters at follow-up. The study by Demehri [26] (n = 17) did not report significant changes in DXA status at a follow-up of 2 years, while the study by Pichler [24] reported a reduction of BMD at 1 and 2 years in 35 children with repeated DXA and a reduction of BMAD only at 1 year.
Several factors, including age, anthropometric data, the underlying gastrointestinal condition, dietetic factors (duration of PN, degree of dependency, calcium and vitamin D intakes), and bone markers level were found to be associated with MBD in the different studies.
Older age at DXA assessment was directly correlated with MBD in the studies by Ubesie [22] (n = 123) and Neelis [29] (n = 46) at a regression analysis. The longitudinal study of Poinsot [30] identified a younger age (at final DXA assessment, n = 31) as a factor associated with an increased prevalence of MBD at follow-up; however, effect of age was not analyzed at baseline. No correlation of BMD with age was found by Derepas [25] (n = 13) and Olszewska [31] (n = 17).
Increased weight and height for age were positively correlated to a higher BMD in the study by Diamanti [20] (n = 24) at a correlation analysis and in the study by Khan [27] (n = 65) for the weight z-score only at a multiple regression analysis. Patients with lower height z-score and higher weight for height (WFH) z-score had lower BMD z-scores in the study by Neelis [29] (n = 46), and this result was confirmed only for WFH z-score at a multivariate analysis. Lastly, no correlation with anthropometric measures was described by Olszewska [31] (n = 17) and by Derepas [25] (n = 13).
Some IF etiologies were correlated to a major risk of MBD, and this was reported for motility disorders and congenital enteropathy by Poinsot [30], for surgical IF (no SBS) by Neelis [29], and for non-SBS etiology by Diamanti [20]. Both the studies by Pichler [24] and Nader [33] showed a risk of lower BMD in children with congenital enteropathy, while only the study by Pichler [24] found higher risk in congenital motility disorders, but all these data were not confirmed at the logistic regression or multiple variate analysis. Duration of PN was found to be negatively correlated with BMD at multivariate analysis in the studies by Mutanen [23], Demehri [26], and Neelis [29] (total n = 123, PN duration, mean or median= 3.4-9.3 years). Dellert [19], Diamanti [20], Olszewska [31], and Nader [33] did not report a significant correlation between DXA parameters or risk of MBD and PN duration (total n = 99, PN duration, mean or median= 0.6-12.4 years). The study by Poinsot [30] did not find a correlation between PN duration and risk of MBD at baseline (n= 31, median PN duration= 2.7 years); however, at last DXA (n = 31, median PN duration = 9.2 years), children with MBD had shorter PN duration. The longitudinal analysis in the same study showed a reduced risk of MBD per year of PN (OR: 0.9 per year).
Dependence on PN at bone health evaluation was associated with reduced BMD in the study by Ubesie [22] at univariate analysis (data not confirmed at logistic regression analysis). The studies by Pichler [24] and Nader [33] did not confirm the dependency from PN or the entity (measures= number of PN days/week, PN dependency index) to be related to MBD. Duration of EN after weaning PN was a significant predictor for a lower BMD z-score in the population of 41 IF children, described by Mutanen [23]. Appleman [24] (n = 20) found no association between aluminum concentration in PN and reduced BMC or BMD.
A negative correlation between serum vitamin D levels and prevalence of MBD was found by Demehri [26] at univariate analysis but not confirmed at the multiple linear regression model. Levels of vitamin D or vitamin D deficiency were not related to the risk of MBD in the studies by Pichler [24] and Nader [33]. Lower calcium levels and higher PTH levels were more prevalent in patients with MBD in the study by Khan [27], but no influence of mineral status was found by Diamanti [20]. Calcium supplementation resulted as a predictor of MBD in the study of Mutanen [23].

Discussion
This is the first systematic review to define the prevalence and risk factors for development of MBD in children with IF. All the studies included in the review reported cases of MBD in IF children (based on radiologic techniques, mainly DXA), with percentages varying from 12.5% [22] to 87.5% [17].
The findings of this systematic analysis are largely heterogeneous. Reasons for heterogeneity include small and variable study sizes and control groups and large variability in the definition of intestinal failure and metabolic bone disease. Techniques of bone disease assessment were variable, and older studies were based on inaccurate and subjective methods (i.e., X-ray), while DXA is currently considered the gold standard for osteoporosis evaluation both in children and adults. Limiting the data to the studies based on DXA and in accordance with the current international pediatric definition [36], a "low bone mineral mass or density" (defined as BMC, BMD, or BMAD less than or equal to −2 z-score) was found in up to 45% of IF children [30]. As expected, reported prevalence increases when a less stringent cut-off (z-score ≤ −1) was adopted. Furthermore, considering the discrepancy of the results, data retrieved by this review clearly support the necessity of a bone assessment in the follow-up of children with IF. Prevalence data of MBD in IF children are similar to figures detected in adult populations. A large, multicenter study in adults with IF reported a prevalence of MBD between 31 and 41% according to the different definitions [10]. Prevalence up to 56% were reported by other studies [37,38]. Although these results support ours, a precise comparison is not possible considering the following: (1) The adult IF population differs largely from children for PN indications with a large proportion of IBD patients; (2) adult MBD has different definitions and different DXA site assessment; and (3) no systematic review on adult data has been conducted so far.
The studies included in the review pointed out several factors correlated to the risk of MBD, using different methodologies. Older children and subjects with growth failure and with some specific IF etiologies seem more at risk of developing MBD compared to other IF subgroups.
An inappropriate development of skeletal tissue or bone loss can be the consequence of an insufficient availability of micro and macronutrients in a malnourished patient; however, few data in this systematic review support the link between a poor nutritional status and a reduced BMD. The only two studies documenting this correlation did not introduce any correction of DXA measures for anthropometric parameters or for bone age, while several other studies corrected the DXA findings for measures of poor growth, retrieving more accurate figures of bone mass or density [22][23][24]29,30]. Some degree of overestimation of MBD in children with faltering growth cannot be excluded; therefore, current guidelines [36] recommend in cases of short stature or poor growth an adjustment for the height z-score or the calculation of the BMAD as a parameter that considers the bone size. Interestingly, in the two controlled studies [18,24], differences in BMC or BMD between cases and controls were no longer significant after correction for weight and height. Patients with IF are at high risk of growth failure during PN [35,39,40] and after PN cessation [19,29]; therefore, an adjustment of bone density parameters is strongly recommended.
Other factors have been advocated in the risk of reduced bone mass, such as the medical cause of IF and some PN-related factors, including nutrient and mineral deficiencies, excessive urinary calcium excretion, metabolic acidosis, and high aluminum concentrations in PN [41].
In our research, the underlying gastrointestinal condition leading to IF was found to be relevant by several studies. Specifically, patients with motility disorders (congenital intestinal pseudo-obstruction, long-segment Hirschsprung's disease, and total or neartotal aganglionosis) and congenital or early-onset enteropathies were at higher risk of developing MBD. The association with other conditions, such as nephropathy in some motility disorders, and use of steroids in some early-onset inflammatory enteropathies (autoimmune) can contribute to bone loss and ultimately to the severity of bone disease in these subgroups of IF patients, and therefore, a particular attention should be reserved to these subjects. Motility disorders and congenital enteropathies are also at risk for elevated fluid and electrolyte losses and consequent imbalance with the possibility of metabolic acidosis, another factor potentially impacting the bone mineralization process. In fact, a chronic acidotic state can directly impair vitamin D metabolism or compromise the bone buffering system while favoring the leaching out of calcium and phosphorus from the bone [35]. The impact of episodes of acidosis has been not specifically analyzed by the included studies.
Effects of mineral deficiencies on the BMD, mainly calcium and phosphorus, considered essentials for bone structure, were sporadically assessed in the studies included in the review. The negative correlation between the necessity of calcium supplementation [23] and a direct correlation between calcium serum levels and low BMD [27] identified by two studies suggest the importance of avoiding calcium deficiency and/or hypocalcemia. Furthermore, enteral calcium and phosphorus intakes were significantly lower in IF patients compared to healthy children [32], suggesting the necessity of supplementation. The right amount of parenteral supplementation of calcium and phosphorus has long been debated considering the risk of calcium-phosphate precipitation in PN solution, and no specific guidelines on enteral calcium and phosphorus intakes are available.
The relationship between vitamin D and PN-associated MBD is less evident. If vitamin D deficiency is clearly associated with risk of bone loss and rickets, patients with IF have several reasons to be at risk of hypovitaminosis D (malabsorption, reduced sun exposure, etc.); conversely, the risk of vitamin D toxicity has been raised in patients on long-term PN. The excess of vitamin D would increase bone reabsorption in these patients and consequently contribute to bone disease. Several studies in adults on PN demonstrate an improvement in BMD after vitamin D reduction or withdrawal without consequences on vitamin D serum levels or calcium balance [41,42]. One study in IF children showed no effect of discontinuing vitamin D in the serum levels of 1,25-(OH2)D and calcium/phosphorus balance, with no clinical consequences [17]. According to the results of our review, vitamin D deficiency was common in IF patients; however, no significant differences were seen with the control population, and no clear evidence of a relation between vitamin D levels and risk of MBD emerged. Overall, vitamin D requirement does not seem to be increased in patients with IF (unless other comorbidities are present, such as renal or liver-cholestatic disease), and an adjustment based on serum levels is recommended.
The role of specific PN mixtures or components in the risk of MBD was not systematically investigated in the included studies. The contribution of aluminum toxicity on bone metabolism and MBD development has been advocated for several years [35]. In the last years, low-containing PN solutions have been developed, reducing such risk, which remains relevant in patients with concomitant kidney impairment. Only one study [24] evaluated this association, showing, despite an elevated serum aluminum concentration in IF subjects, a lack of correlation with BMC or BMD.
Overall results from the studies included in the review suggest the risk of MBD in children both during and after PN cessation. Unexpectedly, controversial results were found on the association between degree of PN dependence and/or duration of PN with the severity of bone disease, not permitting to draw firm conclusion on this argument. This reinforces the multifactorial pathogenesis of MBD in intestinal failure. Despite the evidence of bone disease in IF children several years after withdrawing PN, the reduction of the risk of MBD in the longitudinal study by Poinsot [30] is encouraging. However, overall results of our review emphasize the importance of a regular and long-term monitoring. The most recent guidelines on pediatric HPN suggest measurement of BMD using DXA on a 2-3 yearly basis or annually if previously abnormal [43]. Strategy of primary and secondary prevention are fundamental also considering that treatment options are restricted and sporadically reported in IF children [44,45].
Some limitations to this systematic review need to be mentioned. First, the quality of the studies varied considerably; most of the studies were small and retrospective, and the statistical analysis not always included a multivariate or regression analysis. Studies were heterogeneous in patient selection (definition of the IF population) and outcome measures, particularly in the definition of bone disease and vitamin D deficiency. For these reasons, the possibility to perform a metanalysis was not considered. Studies were limited to European and U.S. centers; therefore, data could be not representative of other countries. All these limitations highlight the need for prospective, multicenter studies on this topic based on standardised DXA measures, corrected for growth failure, and compared with population nomograms.

Conclusions
In conclusion, data from our review indicate an increased prevalence of metabolic bone disease in children with intestinal failure. The risk is higher in older children with growth failure and longer PN duration; however, the possibility of MBD remains significant even when the enteral autonomy is achieved. A long-term monitoring of bone health through DXA assessment (with adequate correction for anthropometric measures) and a careful evaluation of calcium balance and vitamin D requirement is warranted in children with IF. Given the importance of the topic and the limitations of the published studies, there is a need for prospective population multicenter research on this argument. Specifically, the impact of single PN components and the effects of treatment strategies should be assessed in a longitudinal, prospective pediatric studies.
Author Contributions: Conceptualization, S.G. and S.Q.; methodology, S.G. and C.C.; data search and analysis: S.G., A.P. and S.Q.; writing-original draft preparation, S.G., S.Q., A.P. and G.N.C.; writing-review and editing, S.G., M.E.L. and C.C. All authors have read and agreed to the published version of the manuscript.