Chronic liver disease impairs bacterial clearance in a human model of induced bacteremia

Abstract

Sepsis often causes impaired hepatic function. Patients with liver disease have an increased risk of bacteremia. This is thought to be secondary to impaired reticuloendothelial system function. However, this has not been demonstrated clinically. Since transient bacteremia occurs following toothbrushing, we hypothesized that subjects with cirrhosis would have impaired bacterial clearance following toothbrushing compared with subjects with pulmonary disease and healthy controls. After baseline blood was drawn, the subjects underwent a dental examination to determine plaque index and gingival index. Following toothbrushing, blood was drawn at 30 seconds, 5 minutes, and 15 minutes. Bacteremia was measured using quantitative real-time PCR with primers that amplify all known bacteria. We found greater than 75% incidence of bacteremia following toothbrushing. While control and pulmonary subjects were able to clear this bacteremia, subjects with cirrhosis had prolonged bacteremia. Baseline and peak bacterial load correlated with plaque index, suggesting that dental hygiene predicts the degree of bacteremia. However, only the severity of cirrhosis was predictive of bacterial clearance at 15 minutes, suggesting that liver function is important in clearing bacteremia. In this study, we demonstrate clinically that cirrhosis results in impaired bacterial clearance. This suggests that cirrhotic patients may be more susceptible to sepsis because of ineffective bacterial clearance.

Figure 1

Subjects with cirrhosis have delayed bacterial clearance following toothbrushing. (A) Nonbacteremic blood was spiked ex vivo with serial dilutions of stock culture of E. coli. The stock culture had approximately 107 colony forming units (CFU)/mL by optical density. Bacterial load was measured by standard plating and by DNA isolation followed by quantitative real‐time PCR with primers specific for the bacterial 16S ribosome gene. A log transformation was performed to correct for unequal variances. ANOVA followed by Bonferroni's test for multiple comparisons demonstrates no difference in bacterial quantification at higher bacterial loads. However, PCR was able to detect lower levels of bacteria to a greater degree of accuracy (*p > 0.05). (B) Blood was drawn from the subjects at baseline and 30 seconds, 5 minutes, and 15 minutes following toothbrushing. Bacterial load was measured in blood samples by quantitative real‐time PCR. A log transformation was performed to correct for unequal variances. ANOVA followed by Bonferroni's test for multiple comparisons demonstrates a significant difference between bacterial load in cirrhotic subjects and controls at all time points (*p < 0.01). At 15 minutes, there was a difference in bacterial load in cirrhotic subjects compared with those with COPD and controls (**p < 0.01). There was no difference in bacterial load between subjects with COPD and controls.

Figure 2

Baseline and peak bacterial load correlate with plaque index. (A) Linear regression correlating log transformation of bacterial load in all subjects with bacteremia at baseline with plaque index (PI) and gingival index (Gl) demonstrates a significant correlation: r²= 0.48 and p < 0.01 for PI and r²= 0.46 with p= 0.01 for Gl. (B) Using linear regression, we compared the log transformation of bacterial load at 30 seconds with the plaque index in all bacteremic subjects. There was a significant relationship based upon linear regression analysis: r²= 0.42 with p < 0.01. (C) We compared the log transformation of bacterial load at 30 seconds with the plaque index in all bacteremic subjects. There was a significant relationship based upon linear regression analysis: r²= 0.41 with p < 0.01.

Figure 3

In cirrhotic subjects, there is a strong correlation between plaque index and baseline and peak bacterial load. A subgroup analysis was performed com paring a log transformation of the bacterial load at baseline, 30 seconds, and 5 minutes in cirrhotic subjects with bacteremia and plaque index. (A) Using linear regression, there was a significant relationship between baseline bacterial load and plaque index in cirrhotic patients: r²= 0.87 with p < 0.001. (B) Using linear regression, there was a significant relationship between bacterial load at 30 seconds and plaque index in cirrhotic subjects: r²= 0.48 with p = 0.0138. (C) Using linear regression, there was a significant relationship between bacterial load at 5 minutes and plaque index in cirrhotic subjects: r²= 0.47 with p= 0.0155.

Figure 4

Severity of cirrhosis is an important determinant of prolonged bacteremia following toothbrushing. (A) Bacterial load was compared at 15 minutes after toothbrushing between subjects with Child's A cirrhosis and those with Child's B cirrhosis who had detectable enteric bacteremia. A log transformation of bacterial load was performed to correct for unequal variances. A Mann‐Whitney test demonstrates a significant increase in bacterial load in subjects with Child's B cirrhosis compared with those with Child's A cirrhosis (*p < 0.01). (B) The same subjects were classified based upon the MELD score. The Mann‐Whitney test demonstrates a significant increase in bacterial load at 15 minutes in subjects with a MELD score >10 compared with those with a MELD score <10 (*p <0.05). (C) In cirrhotic subjects, we evaluated the relationship between bacterial load at 15 minutes and the MELD score. Linear regression analysis revealed a significant correlation between bacterial load at 15 minutes and the MELD score (r²= 0.33, p= 0.02).