Levitt's CO breath test in the differential diagnosis of chronic isolated hyperbilirubinemia

Isolated hyperbilirubinemia (HB) is a common clinical laboratory finding of elevated serum bilirubin levels in the context of apparently normal global liver function and normal levels of biochemical markers of hepatocellular injury and cholestasis [1]. Three basic mechanisms have been found to result in isolated HB. The first is increased bilirubin production, which is seen in various hemolytic diseases and in individuals with hematopoiesis (hemolysis in situ). The second mechanism is decreased hepatocellular uptake and conjugation of unconjugated bilirubin, which is seen mainly in Gilbert syndrome and Crigler–Najjar syndrome. Both of these syndromes are caused by dysfunctional hepatic uridine diphosphate glucuronosyltransferase family 1 member A1 (UGT1A1), due to mutations affecting the gene UGT1A1. The third mechanism is decreased hepatocellular secretion of conjugated bilirubin, which occurs in Dubin–Johnson syndrome, a rare condition, as well as in the even more rare Rotor syndrome.

Clinically, the evaluation of a patient with isolated HB is quite different from that of a patient who presents with elevated bilirubin associated with elevated liver enzyme levels, with the latter suggesting either a cholestatic or hepatocellular process [2]. Traditionally, the first step in the evaluation of a patient presenting with isolated HB is to fractionate a sample of bilirubin from the patient to determine if the patient's bilirubin is conjugated or unconjugated. In the latter case, the second step is to differentiate between a hemolytic and non-hemolytic origin. Based on these findings, the type of isolated HB can then be classified (i.e. (un)conjugated, (non-)hemolytic) [2].

Direct bilirubin levels measured by the commonly employed diazo reaction method often overestimates the actual conjugated bilirubin level, especially in hemolytic samples, resulting in potential misclassification of isolated HB cases [2–4]. However, isolated conjugated HB is extremely rare (i.e. principally consisting of patients with Dubin–Johnson syndrome and Rotor syndrome). Thus, in the vast majority of isolated HB cases, unconjugated HB is observed [5], making reliable differentiation between HB of hemolytic versus non-hemolytic origin a key diagnostic objective. Current hemolytic marker-based clinical tests—such as a complete blood count (CBC), reticulocyte count, peripheral blood smear, and lactate dehydrogenase (LDH) assay, as well as assessments of serum haptoglobin and plasma free hemoglobin—are far from satisfactory with respect to sensitivity and specificity. Many patients do not have a clear HD classification diagnosis even after the numerous etiological tests.

Erythrocyte (RBC) lifespan refers to the average period that RBCs survive after being released into circulation from bone marrow. Typically, healthy adults have an average RBC lifespan of 120 d [6, 7]. An abnormally short RBC lifespan is the fundamental characteristic of hemolysis. Therefore, RBC lifespan measurement should give a more accurate diagnosis of hemolysis than indirect hemolytic marker assays [8]. However, RBC lifespan measurement is seldomly conducted due to the cumbersome and protracted nature of classical labelling methods (e.g.⁵¹Cr and ¹⁵N-glycine), which can take several weeks, or even months.

Based on the knowledge that exhaled endogenous CO comes primarily from degraded hemoglobin following RBC destruction, established mainly by Rurnal et al [9–12], and thus the notion that mean RBC lifespan can be estimated based on endogenous CO production rate, Levitt's team [13, 14] developed a simple, rapid CO breath test to determinate RBC lifespan. Our previous study confirmed that Levitt's CO breath test gives similar results to those obtained with classical labelling methods [15]. The aim of this pilot prospective diagnostic study was to evaluate Levitt's CO breath test to discriminate hemolytic from non-hemolytic HB in adults.

2.1. Study design and participants

2.2. Levitt's CO breath test

The principle of Levitt's CO breath test is that endogenous CO in the breath originates mainly (∼70%) from heme oxidation during hemoglobin degradation following RBC rupture, such that the total capacity of CO from hemoglobin divided by the CO quantity released per day equates to mean RBC lifespan [10]. Given that the lungs mediate the only mechanism available to the body for CO excretion and the fact that alveolar ventilation volume (mL min⁻¹) is roughly equal to total blood volume (mL), Levitt's team [13, 14] deduced that RBC lifespan can be calculated from exhaled endogenous alveolar CO concentration (ppm) and peripheral blood hemoglobin concentration (g mL⁻¹) according to the following formula:

$$\begin{align}& {\text{RBC}}\,{\text{survival}}\,\left( {\text{d}} \right) \nonumber \\ &\quad = \left( {1380} \right) \times \left( {\left[ {{\text{hemoglobin}}} \right]} \right) /\nonumber \\ &\qquad{\text{endogenous}}\ {\text{alveolar}}\ {\text{CO}}\ {\text{concentration}}.\end{align}$$

We conducted the breath test with an automated instrument, namely the ELS Tester (Seekya Biotec. Co., Ltd, Shenzhen, China). The test involved three steps. (1) For breath sample collection, in the morning (8:00–11:30) without a fasting requirement, each subject inhaled deeply, held their breath for 10 s, and then exhaled into a tri-channel collection system that discards the first 300 ml, presumed to be dead space air, and then directs subsequent alveolar air into a foil bag (700 ml). If needed, the procedure was repeated until the collected air sample reached the collection bag capacity. The filled bag was detached and sealed. An atmospheric sample was collected just after breath sampling. Alveolar air and atmospheric samples were stored at room temperature and analyzed within 2 d of collection. (2) To measure hemoglobin concentration, a CBC was carried out on the breath sampling day. (3) Finally, for RBC lifespan determination, an automated ELS tester was used to measure endogenous alveolar CO concentration by nondispersive infrared spectroscopy of the paired alveolar and environmental air samples. The ELS tester used these data, together with an inputted hemoglobin concentration value, to calculate RBC lifespan according to Levitt's formula (above). Briefly, for step 3, the alveolar and environmental air sample bags were connected to inlet ports and each patient's hemoglobin concentration datum was inputted; then, automatic measurement by the ELS TESTER was initiated by pushing the start button on the machine, which reported each participant's RBC lifespan within 15 min.

Previously, we showed that the mean normal RBC lifespan determined Levitt's CO breath test was 126 d (range, 75–177 d), a value similar to that obtained by the standard biotin-labelling technique (mean, 115 d; range, 70–140 d) [7, 15].

2.3. Outcomes

2.4. Statistics

3.1. Patient characteristics

3.2. RBC lifespans

The RBC lifespans obtained for each of the 130 subjects are shown in figure 1. The mean RBC lifespan of the 77 patients with non-hemolytic HB was 93 ± 26 d, which was significantly shorter than the normal mean reference value (126 d; t = −11.254, p = 0.001), but much longer than that of the 53 patients with hemolytic HB (36 ± 17 d; t = 14.2, p = 0.001). RBC lifespan did not differ significantly between the 26 patients with simple hemolytic HB and the 27 patients with comorbid hemolysis and Gilbert syndrome (32 ± 14 d vs. 40 ± 18 d; t = − 1.983, p = 0.053).

Zoom In Zoom Out Reset image size

3.3. Diagnostic performance

All 53 (100%) patients with hemolytic HB and 21/77 (27.3%) patients with non-hemolytic HB had RBC lifespans below normal range (lower limit of normal by Levitt's breath test = 75 d). ROC analysis indicated that a cutoff RBC lifespan of 60 d was optimal for group separation (figure 2). With this, we reached an accuracy of 95.4% (95%CI 90.0–98.1) with 94.3% sensitivity (95%CI 84.0–98.7) and 96.1% specificity (95%CI 88.7–99.1). The area under for the ROC curve (AUC) was 0.982 (95%CI 0.965–0.999).

Zoom In Zoom Out Reset image size

Of 130 eligible adult patients with chronic isolated HB enrolled over a 2 year period, only one patient (0.08%), who was found by a medical record search, had conjugated HB (table 1), confirming that the vast majority of cases of isolated HB are unconjugated, and thus that the key objective of differential diagnosis for isolated HB is to determine whether there is hemolysis. Levitt's CO breath test was confirmed to have very high sensitivity for detection of hemolysis, even mild hemolysis, in adult patients with chronic isolated HB. Using an RBC lifespan cutoffof 60 d, the diagnostic performance of the breath test for hemolytic HB was excellent with 94.3% sensitivity and 96.1% specificity (figure 2). However, among patients with hemolytic HB, the test could not distinguish between simple hemolytic HB and hemolysis coexisting with Gilbert syndrome (figure 1).

Gilbert syndrome is a well-known phenotypically heterogenous hereditary condition characterized by mild, chronic unconjugated HB in the absence of liver disease or overt hemolysis. Reported incidence rates for Gilbert syndrome range from 6% to 12% [2]. The pathogenesis of Gilbert syndrome has been linked to congenital mutations in UGT1A1 and consequent disruption of UGT1A1, the enzyme that conjugates glucuronic acid to bilirubin in hepatocytes and then converts it into an excretable molecule. Traditionally, unconjugated HB resulting from Gilbert syndrome has been attributed completely to this non-hemolytic mechanism because commonly used clinical biomarkers of hemolysis (e.g. serum LDH, haptoglobin, and reticulocyte count) are not found in patients with Gilbert syndrome. However, several early studies employing classical ⁵¹Cr-labeled RBC re-transfusion measurement techniques suggested that some 30% ∼ 80% of people with Gilbert syndrome have mild hemolysis indicated by a slightly shortened RBC lifespan [16–19]. Therefore, in addition to being a bilirubin conjugating disorder, bilirubin overproduction with mild hemolysis can be considered a diagnostic marker of Gilbert syndrome [17, 19, 20]. Consistent with these previous reports, our non-hemolytic group, which consisted mostly of patients with Gilbert syndrome, had a slightly reduced mean RBC lifespan (92 d vs. normal mean of 126 d). Indeed 38% of the patients in the non-hemolytic group had an RBC lifespan shorter than 75 d, which is the lower limit of the normal range. With respect to diagnosis, the present results suggest strongly that simple rapid Levitt's CO breath testing would be as sensitive as classical standard labelling methods for detecting hemolysis in the context of isolated HB.

We found that if 75 d, the lower limit of the normal RBC lifespan range, were used as the cutofffor hemolytic HB diagnosis, we would reach 100% diagnostic sensitivity, but with a false positive rate as high as 27.3%. The high false positive rate was apparently the consequence of mild hemolysis in Gilbert syndrome being detected. Thus, the adjusted cutoffis critical for differential diagnosis of isolated HB. Based on our ROC analysis, we recommend an optimal RBC lifespan cutoff of 60 d to maximize both sensitivity and specificity.

Generally, Gilbert syndrome coexisting with hemolytic disease is considered to be uncommon, albeit with isolated single case reports [21–25]. We were surprised to find that more than half of our enrolled patients with hemolytic HB (37/53) had Gilbert syndrome, indicating that this phenomenon maybe much more common than is appreciated. Perhaps many such cases have been missed due to a lack of awareness and available techniques. The prevalence of Gilbert syndrome in the general population, at 6% ∼ 12%, indicates that a similar prevalence may be found among hemolytic patients by chance. Thus, the incidence of Gilbert syndrome among our sample of hemolytic HB patients far exceeding this range is noteworthy. Here, we were able to find undiagnosed Gilbert syndrome in a large number of patients because all of the participants in our study underwent Sanger sequencing of UGT1A1 and detailed hemolysis tests. Although RBC lifespan data were not useful for distinguishing between patients with simple hemolytic HB and those with comorbid hemolysis and Gilbert syndrome, it is evident from our baseline clinical characteristic data (table 1) that the latter patient subgroup tended to have higher hemoglobin and bilirubin levels than patients with simple hemolytic HB, consistent with previous single case reports [21–25]. Thus, discordance between significant unconjugated HB and less severe hemolysis may be a valuable clue that a patient may have hemolytic disease comorbid with Gilbert syndrome. Further studies are needed to clarify the clinical significance of this pattern of findings.

Our study had several limitations. First, we did not enroll patients with drug-induced non-hemolytic HB, a condition that needs to be considered in the differential diagnosis of isolated HB. Second, although non-hemolytic HB was diagnosed according to an established clinical rubric, it is possible that a few patients might have been misclassified. Third, because the study was initially designed as a proof-of-concept study, the number of patients was not large enough to be divided into a development set and a validation set, especially for patients with simple hemolytic HB. Therefore, further prospective testing of the recommended optimal cutoff value's performance should be done.

In summary, the present results show that Levitt's CO breath test is a highly sensitive method for detecting hemolysis in adult patients with chronic isolated HB. RBC lifespan measurements with this method can discriminate between patients with non-hemolytic HB and patients with hemolytic HB with high diagnostic accuracy. Thus, Levitt's CO breath test represents a new simple and rapid diagnostic approach to differentiating among different chronic isolated HB pathologies. The validity demonstrated here should be confirmed in a large-scale validation study. Finally, it should be noted that because smokers absorb large amounts of CO from tobacco smoke, this breath test may not be appropriate for smoking patients, at least not with the presently used formula for calculating RBC lifespan.

All authors had access to the study data and reviewed and approved the final manuscript.

The data that support the findings of this study are openly available at the following URL/DOI: https://figshare.com/(DOI:10.6084/m9.figshare.14130209).

Hou-De Zhang and Ling-Ling Kang designed the study, developed the study protocol, recruited subjects, collected breath samples, analyzed the data, and wrote the manuscript. Ze-Lin Liu edited the protocol, supervised all experimental steps, and contributed to hemolysis diagnoses. Quan-Sheng Han, Yuan-Wu Chen, Ling-Wen Liu and Xian-Hui Xie contributed to subject enrollment and data collection. Jun-Feng Luo was responsible to breath samples measurement. Yong-Qiang Ji and Guo-Liang Zhu contributed to the statistical analysis and mapping. Yong-Jian Ma and Kun-Mei Ji analyzed the data and revised the manuscript.

Hou-De Zhang, Yong-Qiang Ji, Guo-Liang Zhu, and Yong-Jian Ma were members of research and development team of the ELS TESTER. The other authors have no competing interests to declare.

This trial is registered with www.chictr.org.cn number, CHiCTRDDD17011592.

Written informed consent was obtained from all study participants.