Trends of INR and Fecal Excretion of Vitamin K During Cholestasis Reversal: Implications in the Treatment of Neonates with Intestinal Failure Associated Liver Disease

Abstract

Background:

Vitamin K is a fat-soluble compound that plays important roles in coagulation. In children with intestinal failure associated liver disease (IFALD), the disrupted enterohepatic circulation can lead to intestinal loss of vitamin K. Fish oil-based lipid emulsion (FOLE) has proven effective in treating IFALD. As biliary excretion is restored during cholestasis reversal, the accelerated vitamin K loss can pose a risk for deficiency.

Methods:

Ten neonates with IFALD receiving FOLE monotherapy were prospectively enrolled in the study from 2016 to 2018. In addition to weekly measurements of international normalized ratio (INR) and direct bilirubin (DB), ostomy output was collected for determination of fecal concentrations of phylloquinone (PK) with mass spectrometry. Trends of DB, INR, and fecal PK concentrations were summarized with locally estimated scatterplot smoothing.

Results:

The median time (interquartile range) from FOLE initiation to cholestasis reversal was 59 (19–78) days. During cholestasis reversal, INR remained relatively unchanged while the mean (95% confidence interval; CI) daily fecal excretion of PK increased from 25.1 (5.0, 158.5) ng at the time of FOLE initiation to 158.5 (31.6, 1000.0) ng at complete reversal. Examination of individual trends in fecal PK excretion and INR also revealed little correlation between the two measurements (r = −0.10; P=0.50).

Conclusion:

Children with IFALD are at risk for vitamin K deficiency during cholestasis reversal. Close monitoring and quantified supplementation of vitamin K may be warranted during this period. However, this should not be guided by INR alone as it is a poor indicator of vitamin K status.

Introduction:

Short bowel syndrome (SBS) occurs when the absorptive capacity of the small intestine is compromised following extensive loss caused by surgical resection, or from underlying functional gastrointestinal diseases. Patients with SBS usually require parenteral nutrition (PN) to sustain life. Prolonged PN dependence in children puts them at risk for intestinal failure associated liver disease (IFALD), a condition characterized by hepatic cholestasis, and even cirrhosis¹. The soybean oil present in the standard intravenous lipid emulsion is thought to be a major contributor for the development of IFALD². Its substitution with a fish-oil based intravenous lipid emulsion (FOLE) is an effective strategy for the treatment of this condition and reversal of cholestasis³.

Vitamin K is one of the fat-soluble vitamins. It occurs as phylloquinone (PK), or vitamin K1, and a family of similar vitamers collectively called menaquinones (MKs), or vitamin K2. Vitamin K is an essential cofactor that plays a central role in maintaining the coagulation cascade, calcium homeostasis, and prevention of vascular calcification⁴. The liver is the major site of metabolism of vitamin K. Patients with acute liver failure may display prolonged clearance of PK after intravenous administration of vitamin K⁵. Almost half of all excreted PK is found in feces, of which 15–23% is unmodified PK^6–8. Additionally, when bile flow is diverted via a T-tube, no vitamin K metabolites are found in the feces, demonstrating that biliary excretion is the only route for metabolized vitamin K to enter the intestine⁶. Prothrombin time (PT) and the activated partial thromboplastin time have been used as traditional screening tests for vitamin K deficiency but a prolonged PT is not specific to vitamin K deficiency and is not useful in diagnosing subclinical deficiency⁹. The most useful confirmatory test is restoration of a normal PT only after supplementation with vitamin K. The international normalized ratio (INR) can also be used as an indirect measurement of vitamin K status In states of deficiency, a decrease in levels of the vitamin K-dependent coagulation factors is reflected in increased INR values¹⁰.

The concern for omega-3 fatty acid-induced coagulopathy continues to persist, despite the lack of clinical evidence^11,12. This study aimed to investigate the trend of INR and fecal excretion of vitamin K in a prospective cohort of IFALD children treated with FOLE. We hypothesize that as cholestasis reverses and biliary flow improves, the limited absorptive capacity in these patients leads to increased fecal vitamin K loss and any trend in INR, if present, would mirror the trend in fecal excretion of vitamin K.

Method:

Patient Enrollment:

This was a prospective observational study conducted at Boston Children’s Hospital and approved by its Institutional Review Board. Eligible patients included all children younger than 2 years who developed IFALD, a diagnosis that is made when a PN-dependent patient has two consecutive measures of direct bilirubin (DB) ≥ 2 mg/dL at least one week apart. At this point, treatment with the FOLE Omegaven (Fresenius Kabi, Germany) was initiated under a compassionate use protocol. Additionally, presence of an ostomy was required in order to ensure the accuracy of the stool output. Exclusion criteria include anticoagulation treatment with warfarin at the time of enrollment, cystic fibrosis or other causes of cholestasis including, but not limited to, Alagille syndrome, biliary atresia, biliary cirrhosis, primary sclerosing cholangitis, alpha-1 antitrypsin deficiency, and hepatitis, and underlying disorders of coagulation including, but not limited to, von Willebrand disease, hemophilia, and factor V Leyden deficiency. Once a patient met the criteria to be entered into the study, parental consent was obtained. PN provided to all patients contained a standard multivitamin regimen that met the adequate intake recommendation for vitamin K according to the National Institutes of Health (2–75 μg/day from birth to 18 years of age). The decision to provide additional vitamin K supplementation was driven by the treating physician and independent of the biomarkers being collected as a part of the study.

Data Collection:

For each patient, baseline clinical information was obtained from the electronic medical record system and included demographic data, birth weight, gestational age, residual bowel length, and etiology of intestinal failure. During the follow-up period, comorbidities, medical treatments, and laboratory measurements were recorded on a weekly basis. Bacteremia was diagnosed with a blood culture, intra-abdominal infection required radiographic evidence, and urinary tract infection (UTI), pneumonia, and wound infection were defined based on antibiotic treatment requirement. Two biochemical measurements of interest were INR and DB. At each follow-up time point, a sample of ostomy output was collected and preserved at −80 ^oC until the time of vitamin K analysis.

Quantification of Fecal Vitamin K and Plasma PIVKA-II:

An internal standard solution of PK-d₇ (Sigma-Aldrich, St. Louis, MO) was prepared in 100% ethanol at a concentration of 100 pg/μL. In a separate tube, 100 μL of stool sample, 100 μL of internal standard, and 900 μL of 100% ethanol were mixed. Extraction of vitamin K was achieved with a Bullet Blender homogenizer (Next Advance, Averill Park, NY) for 5 minutes. The mixture was then centrifuged at 4 ^oC at maximal speed for 10 minutes. The supernatant was carefully collected, dried, and reconstituted in 100 μL of 100% ethanol. Quantification of fecal concentrations of PK and 2 isoforms of MK, MK-4 and MK-7, was performed with mass spectroscopy at the Harvard Small Molecule Mass Spectrometry facility according to published protocols^13,14. Daily excretion of fecal vitamin K was determined from the product of fecal concentration and 24-hour stool volume.

Plasma indicator of vitamin K status was assessed by measurement of the protein induced by vitamin K absence-II (PIVKA-II) from discarded patient plasma samples obtained on a weekly basis. PIVKA-II refers to under-carboxylated factors II, whose levels increase in the plasma in the setting of vitamin K deficiency. PIVKA-II measurements were performed with an enzyme-linked immunosorbent assay according to the manufacturer’s protocol (MyBioSource, San Diego, CA).

Termination of Data Collection:

Data collection was terminated when one of the following criteria was met: death or liver/small bowel transplantation, reversal of cholestasis (< 2 mg/dL), reversal of ostomy, wean from PN, administration of warfarin for anticoagulation, elective withdrawal from the study, or loss to follow-up.

Statistical Analysis:

Baseline characteristics are expressed as frequency and percentage or median and interquartile range (IQR). Due to right skewness and the presence of outliers, fecal vitamin K concentrations were transformed using the base 10 logarithmic function. Biochemical measurements and fecal vitamin K concentrations were plotted against time during the course of follow-up. In order to perform aggregate analysis, the length of cholestasis reversal was scaled so that it was uniform across all patients. On this scale, t = 0 indicates the time of FOLE initiation and t = 1 represents complete cholestasis reversal, which was defined as the first DB that fell below 2 mg/dL. Two patients who exited the study before cholestasis reversal were excluded from this analysis. For every biochemical measurement (i.e. DB, INR, fecal vitamin K concentrations, and plasma PIVKA-II levels), the trend over time was estimated with locally estimated scatterplot smoothing (LOESS), a form of non-parametric local regression that uses weighted least-squares to fit a curve through a scatterplot of the data. The percent change from FOLE initiation (%△) was calculated as [(Y_i+1) – (Y_o+1)]/(Y_o+1)×100, where Y_o and Y_i denote the biomarker at t = 0 and t = i, respectfully, and 1 was added to each to prevent division by zero. For fecal PK and PIVKA-II, %△ was calculated on the logarithmic scale. The estimated mean (geometric mean for log-transformed variables) and 95% confidence interval (95% CI) at six equally-spaced time points during cholestasis reversal are reported. The within-patient correlation between INR and fecal PK was assessed using the method introduced by Bland and Altman¹⁵, with P<0.05 indicating statistical significance. All statistical analyses were performed with SAS version 9.4 (Cary, NC).

Results:

Ten patients were enrolled in the study from May 2016 to March 2018. Eight of them were male and eight were Caucasian (Table 1). The median gestational age at birth was 32.5 (29.0–36.0) weeks and the median birth weight was 1210 (760–2080) g. The median length of bowel from the ligament of Treitz to stoma was 32.5 (18.5–47.5) cm. The shortest recorded bowel length was 2 cm in a patient with duodenostomy, while the longest was 67 cm. There was a wide range of etiologies for SBS. The most common one was necrotizing enterocolitis, followed by midgut volvulus. Other diagnoses included gastroschisis, omphalocele, intestinal atresia, meconium ileus, Meckel’s diverticulum, and cloacal exstrophy.

Table 1:

Baseline characteristics of the study cohort.

Characteristic	Number (%)/ Median (IQR)
Male	8 (80%)
Race
White	8 (80%)
Black	2 (20%)
Gestational age (wk)	32.5 (29.0, 36.0)
Birth weight (g)	1210 (760, 2080)
Bowel length (cm)	32.5 (18.5, 47.5)
Diagnosis of SBS*
NEC/SIP	4 (40%)
Midgut volvulus	2 (20%)
Gastroschisis	1 (10%)
Omphalocele	1 (10%)
Intestinal atresia	1 (10%)
Meconium ileus	1 (10%)
Meckel’s diverticulum	1 (10%)
Cloacal exstrophy	1 (10%)

Abbreviations: IQR, interquartile range; NEC, necrotizing enterocolitis; SBS, short bowel syndrome; SIP, spontaneous intestinal perforation.

^*Patients may have had >1 diagnosis.

Most patients in the study were diagnosed with some form of infection during the follow-up period, with bacteremia being the most common one, followed by wound and intra-abdominal infections (Table 2). Four out of 10 patients received either intravenous or intramuscular vitamin K during the follow-up period. One of them received 2 courses of vitamin K. All patients were administered some form of antibiotic during their hospitalization course. Specifically, all patients received penicillin antibiotics, of which the most common agents included piperacillin/tazobactam and ampicillin. Other commonly utilized antibiotics included vancomycin (8 patients), cephalosporin (6 patients), and aminoglycoside and carbapenem (4 patients each).

Table 2:

Comorbidities and treatments of patients during the study period.

Comorbidities/Medications	Number (%)
Comorbidities
Bacteremia	7 (70%)
Wound infection	3 (30%)
Intra-abdominal infection	2 (20%)
UTI	2 (20%)
Pneumonia	1 (10%)
Medications
IV/IM vitamin K	4 (40%)
LMWH	2 (20%)
Antibiotic
Penicillin	10 (100%)
Vancomycin	8 (80%)
Cephalosporin	6 (60%)
Aminoglycoside	4 (40%)
Carbapenem	4 (40%)

Abbreviations: IV, intravenous; IM, intramuscular; LMWH, low molecular weight heparin; UTI, urinary tract infection.

Eight out of 10 patients were followed at least until reversal of hepatic cholestasis (DB ≤ 2 mg/dL). Two patients completed the study while their DB was still above 2. One of them was weaned from PN while the other one had early stoma reversal. The median follow-up period was 75 (43–101) days. The median time from enrollment to cholestasis reversal was 59 (19–78) days.

LOESS regression was performed on the 8 patients who were followed beyond the time of complete cholestasis reversal. As represented by the LOESS curves, DB steadily decreased during the course of cholestasis reversal (Figure 1). However, INR experienced minimal changes during this period. In contrast, the fecal concentrations of PK steadily increased from FOLE initiation to complete cholestasis reversal (Figure 1). At the end of this period, the geometric mean (95% CI) for daily excretion of fecal PK had increased from 25.1 (5.0, 158.5) ng to 158.5 (31.6, 1000.0) ng (Table 3). This represents an increase of 104.9 (15.7, 194.1) % on the logarithmic scale. Examination of INR and fecal PK in individual patients revealed no clear relation between the temporal trends of the two measurements (Figure 2), as confirmed by the within-patient correlation between the two measurements (r = −0.10; P=0.50).

Figure 1:

Locally estimated scatterplot smoothing curves for percent change of INR, direct bilirubin, and fecal PK from the time of FOLE initiation until reversal of cholestasis. Abbreviations: FOLE, fish-oil lipid emulsion; INR, international normalized ratio; PK, phylloquinone.

Table 3:

Estimated values of DB, INR, daily fecal PK excretion, and PIVKA-II from locally estimated scatterplot smoothing regression. Shown are the mean (95% confidence interval; CI) for DB (mg/dL) and INR and the geometric mean (95% CI) for fecal PK (ng) and PIVKA-II (μg/mL). Percent change from FOLE initiation (%△) for fecal PK and PIVKA-II represents the estimated mean change on the logarithmic scale.

Scaled Time Points	DB (mg/dL)	DB (%Δ)	INR	INR (%Δ)	Fecal PK (ng)	Fecal PK (log 10)	Fecal PK (%Δ)	PIVKA-II (μg/mL)	PIVKA-II (log 10)	PIVKA-II (%Δ)
0.0	4.5 (3.6, 5.3)	--	1.25 (1.15, 1.35)	--	25.1 (5.0, 158.5)	1.4 (0.7, 2.2)	--	5.0 (1.6, 15.8)	0.7 (0.2, 1.2)	--
0.2	4.9 (4.1, 5.7)	3.6 (−18.6, 25.8)	1.32 (1.22, 1.41)	3.5 (−1.5, 8.6)	199.5 (39.8, 1000.0)	2.3 (1.6, 3.0)	93.6 (11.6, 175.5)	12.6 (5.0, 39.8)	1.1 (0.7, 1.6)	28.7 (5.4, 52.0)
0.4	4.2 (3.5, 4.9)	−5.2 (−24.8, 14.5)	1.38 (1.29, 1.47)	6.7 (2.0, 11.4)	63.1 (15.8, 316.2)	1.8 (1.2, 2.5)	52.6 (−22.4, 127.6)	6.3 (2.5, 20.0)	0.8 (0.4, 1.3)	8.2 (−14.6, 31.1)
0.6	3.2 (2.5, 4.0)	−23.8 (−44.5, −3.0)	1.29 (1.20, 1.38)	2.8 (−2.1, 7.7)	501.2 (125.9, 2511.9)	2.7 (2.1, 3.4)	103.3 (24.4, 182.2)	4.0 (1.6, 10.0)	0.6 (0.2, 1.0)	0.6 (−23.8, 24.9)
0.8	2.8 (2.1, 3.6)	−27.9 (−47.5, −8.2)	1.33 (1.24, 1.42)	0.7 (−4.1, 5.5)	251.2 (63.1, 1258.9)	2.4 (1.8, 3.1)	115.3 (40.7, 189.9)	6.3 (2.5, 15.8)	0.8 (0.4, 1.2)	−4.0 (−26.8, 18.8)
1.0	1.6 (0.8, 2.5)	−45.6 (−69.1, −22.2)	1.25 (1.15, 1.35)	0.2 (−5.2, 5.6)	158.5 (31.6, 1000.0)	2.2 (1.5, 3.0)	104.9 (15.7, 194.1)	5.0 (1.6, 15.8)	0.7 (0.2, 1.2)	−1.1 (−28.3, 26.2)

Abbreviations: DB, direct bilirubin; INR, international normalized ratio; PK, phylloquinone; PIVKA-II, protein induced by vitamin K absence – factor II.

Figure 2:

Scatterplots for percent change of INR, direct bilirubin and fecal PK from the time of FOLE initiation until reversal of cholestasis, by subject. Abbreviations: FOLE, fish-oil lipid emulsion; INR, international normalized ratio; PK, phylloquinone.

Of the 3 isoforms of vitamin K examined in this study, only PK demonstrated reliable detection. MK-4 was largely absent in feces while MK-7 concentration dropped below the detection limit in many samples, resulting in a widely fluctuating pattern of daily excretion.

Plasma concentrations of PIVKA-II also experienced little changes over the period of cholestasis reversal (Figure 3 and Table 3). In fact, the geometric mean (95% CI) plasma PIVKA-II levels at FOLE initiation and complete reversal were identical at 5.0 (1.6, 15.8) μg/mL, respectively. It should be noted that this concentration is well above the cut-off for vitamin K deficiency in adults, 0.1 μg/mL¹⁶.

Figure 3:

Locally estimated scatterplot smoothing curves for INR, direct bilirubin, and PIVKA-II from the time of FOLE initiation until reversal of cholestasis. Abbreviations: FOLE, fish-oil lipid emulsion; INR, international normalized ratio; PIVKA-II, protein induced in vitamin K absence – factor II.

Discussion:

The enterohepatic circulation plays a critical role in the metabolism and excretion of many forms of drugs, nutrients, and vitamins¹⁷. In patients with SBS, its disruption can result in certain physiologic alterations. In this study, we demonstrated an untoward side-effect of cholestasis reversal in children recovering from IFALD. As DB decreased and biliary excretion resumed, there was a progressive loss of PK via the stoma output. Interestingly, the INR remained relatively unchanged during this period and showed no correlation with fecal excretion of PK. This study again demonstrated that INR is a poor indicator of vitamin K status in this population.

Vitamin K is mostly absorbed in the jejunum and ileum¹⁸. In adults, up to 90% of total vitamin K intake exists in the form of PK¹⁹, which is also the predominant form of vitamin K in the plasma²⁰. On the other hand, MK is the major storage form of vitamin K in humans²⁰. Some studies have suggested that MKs produced by colonic bacteria can contribute up to 50% of total body vitamin K stores²¹. Given that only one patient in this study had a colostomy, the lack of colonic contents could have accounted for the failure to reliably detect MK-7 in the ostomy output. It should also be noted that there was no contribution of MK-4 to the total fecal vitamin K pool in this study. This isoform of vitamin K has been shown to have little impact on the overall vitamin K status due to its poor bioavailability²². However, MK-4 is unique among the vitamin K vitamers as it is the only one that can be endogenously synthesized by certain extra-hepatic tissues, contributing to its high concentration found in the brain, kidney, and pancreas²⁰.

Vitamin K deficiency is a well-known entity in patients with SBS or cholestasis. In these patients, poor absorption of vitamin K due to the lack of either biliary excretion, intestinal absorptive capacity, or both contribute to the prevalence of the condition. Adult patients with SBS and PN may present with significantly lower serum levels of vitamin K²³. In children, vitamin K deficiency was found in 10% of patients with SBS²⁴. In a study of cholestatic patients of all ages, 70% of them were found to have vitamin K deficiency despite the majority taking an oral PK supplement²⁵. In infants with cholestasis, up to 40% were found to have vitamin K deficiency on biochemical tests⁵. Additionally, intestinal absorption of mixed micellar PK in these patients was shown to be inefficient and erratic⁵. In this study, 40% of the children received some form of intravenous or intramuscular vitamin K supplementation during the course of follow-up, in addition to the standard vitamin supplementation delivered in PN. This could have explained the lack of a notable elevation in INR despite increased fecal loss of PK during cholestasis reversal.

For this study, we chose not to investigate the plasma levels of vitamin K as it is highly labile²⁶, has a short half-life²⁷, and varies widely in tissue distribution²⁸. Instead, we used an indirect measurement of vitamin K status, PIVKA-II²⁹. In adults, PIVKA-II has been used to detect vitamin K deficiency in many patient populations, including cancer patients and those in the intensive care unit^30,31. Unfortunately, PIVKA-II has been shown to be an unreliable indicator of vitamin K status, especially in premature infants³², and the lack of an age-adjusted referent range limits its utility as a biomarker for vitamin K status in infants. Regardless, the plasma PIVKA-II concentrations obtained from this study were well above the usual threshold for vitamin K deficiency in adults, 0.1 μg/mL. These numbers therefore should raise the concern for vitamin K deficiency, especially in the setting of increased fecal loss in a vulnerable population.

This study was limited by its small sample size, which precluded adjustments for possible confounding factors, such as gestational age, route of enteral intake, medications, comorbidities, and bowel length. Therefore, this study should be regarded as a hypothesis-generating investigation, and further prospective studies with a larger sample size are warranted. In addition, the lack of a reliable plasma biomarker of vitamin K also limited our conclusion on the impact of fecal PK excretion and overall vitamin K status. Despite these limitations, this study demonstrated several interesting findings. The increased fecal excretion of PK, combined with elevated levels of PIVKA-II, should raise concern for vitamin K deficiency in children with IFALD. Close monitoring and quantified supplementation of vitamin K may be necessary during the period of cholestasis reversal in this particularly susceptible population. Interestingly, the loss of fecal PK is not reflected in INR. This suggests that INR is a poor indicator of vitamin K status and practitioners should be cautioned when using it as a guide for vitamin K therapies, especially if the INR is within the normal range.

Clinical Relevancy Statement:

Children with short bowel syndrome and intestinal failure associated liver disease are especially at risk for vitamin deficiency. For those treated with fish oil lipid emulsion, the period of cholestasis reversal can accelerate intestinal loss of fat-soluble vitamins due to resumption of biliary flow and limited intestinal length. Among these fat-soluble vitamins, vitamin K is especially important due to its role in coagulative function. As a result, additional supplementation of vitamin K may be warranted during the period of cholestasis reversal.