In 1991, we reported an increase in the permeability surface area product (PS) of the blood-brain barrier (BBB) to ammonia in patients with severe liver disease and minimal hepatic encephalopathy (mHE).1 We thought this explained the development of “toxin hypersensitivity” as disease progresses, a phenomenon we documented in animals2 and why patients with mHE might have normal blood ammonia levels.
This finding is challenged by Keiding et al., who report a reduction in the PS product for ammonia in patients with cirrhosis with and without HE3. They attribute the difference to superior imaging and physiological modeling. Data from these two studies are shown in Table 1.
|Source||PS Control||PS Minimal or no Overt HE||PS Overt HE||CBF Control||CBF Minimal or No Overt HE||CBF Overt HE|
|Lockwood et al.1 mid-thalamic slice values ± SD||0.13 ± 0.03 mL/g tissue/min||0.22 ± 0.07* mL/g tissue/min||—||0.58 ± 0.12 mL/g tissue/min||0.46 ± 0.16 mL/g tissue/min||—|
|Keiding et al.3 cortex values ± SEM||0.34 ± 0.03 mL/mL tissue/min||0.31 ± 0.03* mL/mL tissue/min||0.21 ± 0.02* mL/mL tissue/min||0.47 ± 0.03 mL/mL tissue/min||0.44 ± 0.03 mL/mL tissue/min||0.41 ± 0.04 mL/mL tissue/min|
|Calculated from Keiding et al., Table 2 CBF and K1||0.45 (cortex) mL/mL tissue/min||0.57 (cortex) mL/mL tissue/min||0.50 (cortex) mL/mL tissue/min||—||—||—|
To evaluate these data, it is necessary to consider the validity to the assumptions of the models and whether the results agree with other published data. Our model was extremely straightforward and made few assumptions aside from irreversible trapping of ammonia by the brain. Keiding et al. used a graphical approach that requires serial PET images and arterial blood samples4. This model assumes bidirectional movement of ammonia across the BBB with irreversible trapping. Keiding et al. complicate the model by including 13N-urea and 13N-glutamine in the intravascular compartment in addition to 13N-ammonia. They assumed that urea crosses the BBB, without metabolic trapping, with kinetics that are identical to water, an assumption at odds with other data.5 These deviations require an explanation beyond their comment that urea was added to improve the goodness of the fit to the model. Their assumption that the intravascular compartment occupies 1% of the brain volume is about one fourth the value reported by others6. It is also puzzling that mean PS values computed using their equation 1 and mean values of CBF and K1 in their Table 2, bear little resemblance to PS values they report, as shown in the table.
The CBF measurements made by Keiding et al. don't show the expected reductions in patients with overt HE.7 The small standard errors they report imply little difference between patients with grade IV HE and those less severely impaired. The failure to observe a reduction in CBF requires explanation, particularly since this variable appears twice in the formula for computing PS (their equation 1).
Finally, they report that the rate constant for conversion of ammonia to glutamine is 0.1 per minute. This is in contrast to the data of Cooper et al., who reported that the t½ for the trapping reaction was 1-3 seconds, or less.8 What effect would the insertion of this value into model equations have on the PS product?
Progress is based on the development of better methods to address important problems. When disparities in results arise, it is essential to employ multiple methods and careful tests of validity to avoid error. The PS product issue is important and a great deal of our understanding of the role of ammonia and the cerebral dysfunction associated with HE depends on resolving this dilemma.