Leon A. Adams, Richard K. Sterlinga
DOI: http://dx.doi.org/10.1016/j.jhep.2017.02.011
Publication stage: In Press Accepted Manuscript
Published online: February 17, 2017
Accurate identification of liver disease severity and fibrosis stage is paramount in the management of those with chronic liver disease. In the past, this was often done by liver biopsy. However, due to it’s invasiveness and risks of bleeding, pain, and sampling error, non-invasive assessment of liver disease has gained increasing attention over the last decade. Non-invasive tests can be divided into serum tests and imaging tests. Although standard “liver function” tests, such as alanine aminotransferase (ALT) and aspartate aminotransferase (AST) are inaccurate when used alone, several models have been developed that use them in combination with other markers of advanced liver disease, such as platelet count. Of those models that utilize routine, readily available tests, the AST-Platelet Ratio Index (APRI) and FIB-4 have gained the most attention[[1], [2]]. Both APRI and FIB-4 have high specificity and negative predictive value’s (NPV) for advanced fibrosis or cirrhosis[[1], [2]]. However, both have only moderate positive predictive values (PPV) and many patients fall in between the upper and lower cut-offs giving an indeterminate result. More complex serum panels have been developed including Fibrosure/Fibrotest[3][3] and Fibrometer[4][4], which may offer additional accuracy compared to APRI or FIB-4 but have extra cost. In addition, the alternative non-invasive imaging tests, such as vibration controlled transient elastrography (VCTE)[5][5] or magnetic resonance elastrography (MRE)[6][6] are generally only available in specialized centers, leaving many with chronic liver disease inadequately assessed. Because one serum or imaging test alone does not provide 100% sensitivity and specificity, there has been an attempt to combine tests, either performed together or sequentially[[4], [7], [8], [9], [10], [11], [12], [13], [14], [15]]. Thus, the rationale for the study by Boursier and colleagues published in the current issue of the Journal of Hepatology was to develop such a test as part of an algorithm to more accurately and easily identify advanced liver disease[16][16].
To that goal, the study aimed to develop and validate a stepwise algorithm that could be easily used by all providers to facilitate detection of advanced fibrosis in those with chronic liver disease. The cohort studied consisted of 3754 subjects with chronic liver disease who had undergone liver biopsy and VCTE, who were divided 2:1 into a derivation and validation set. They initially evaluated the utility of both APRI and FIB-4 as an initial screening test. Because FIB-4 outperformed APRI (higher sensitivity), they used it to compare to their new “first line test” which included age, gender, gamma-GT (GGT), AST, platelet count, and prothrombin time (PT). Each of these variables was assigned a number (0-4 depending on the variable and cut off) in the “easy LIver Fibrosis Test (eLIFT), with a score of at least 8 providing 80% sensitivity for advanced fibrosis. In a core group of 1946 subjects, they found that while FIB-4 and eLIFT had similar sensitivity (77-78%), eLIFT had higher specificity (91%), NPV (79%), and PPV (91%) with fewer false positive results than FIB-4, especially in those over age 60. They concluded that eLIFT was better than FIB-4 as a general screening test.
Subsequently, they determined what should be the “second line test” by comparing APRI, FIB-4, liver stiffness by VCTE, and Fibrometer (with or without VCTE) and found that FibrometerVCTE had the highest number of biopsies avoided (81%). A new algorithm was proposed with eLIFT as the initial screening test followed by FibrometerVCTE in those with an eLIFT score ≥ 8 for confirmation. Using this two-step strategy, 46% of patients (33% with eLIFT score <8 and 14% with eLIFT ≥ 8 but FibrometerVCTE <0.384) would not need to see a specialist thus avoiding liver biopsy or additional testing. This two-step approach found that 34% had advanced fibrosis (eLIFT >8 and FibrometerVCTE ≥ 0.715) leaving only 19% with an indeterminate result that might go on to liver biopsy. This combined strategy worked better than FIB-4 or FibrometerVCTE alone in identifying those with advanced fibrosis. Finally, they followed 1275 patients longitudinally (median follow-up 2.9 years) to determine if the eLIFT-FibrometerVCTE strategy could predict liver-related and all-cause mortality. Although eLIFT and FIB-4 had similar performance for all-cause mortality, FibrometerVCTE had the best performance for predicting liver-related mortality.
This study has several strengths; the large patient population enabled comparison of several methods of fibrosis assessment and detected relatively small differences in test performance. Furthermore, the proposed algorithm was accurate across a wide spectrum liver disease increasing applicability and attractiveness to implement in the community. In addition, eLIFT and FibrometerVCTE were demonstrated to be prognostic of liver related outcomes, re-enforcing the appropriateness to use them as tests to guide patient management. Nevertheless, the eLIFT test utilizes some serum tests that may not be routinely ordered by community physicians, such as prothrombin time, which in itself may suffer inter-laboratory variability. In addition, the use of AST and GGT in the eLIFT algorithm, resulted in a reduction in specificity in the setting of alcoholic liver disease. Nevertheless, this was resolved when combined with the FibrometerVCTE. The FibrometerVCTE was developed using the M probe, which provides higher values than the XL probe and has a lower success rate in obesity, limiting applicability in this population. For these reasons, independent validation will be important to confirm the accuracy of the eLIFT-FMVCTE algorithm. Lastly, the study population was not identified from the general population, and so caution should be exercised in trying to extrapolate this algorithm to screen for fibrosis in the general population.
Combining non-invasive algorithms is an attractive way of increasing diagnostic accuracy for liver fibrosis, however uncertainty exists on the best way to combine tests and which tests to use. For example, tests may be used concurrently with discordant results leading to a diagnostic biopsy. Alternatively two modalities may be combined into one diagnostic formula, or thirdly, tests may be used sequentially, with the first being a screening test and the second used following an indeterminate screen or as a confirmatory test for a positive screening test. The approach by Boursier utilizes a combination of firstly, a two-step approach using eLIFT as a screening test and FibrometerVCTE as the confirmatory test, which in itself combines VCTE and a serum test[16][16]. Thus, this approach requires three non-invasive tests to be performed, adding to the complexity and cost.
In addition to the current study, a range of alternative combination strategies have been examined (outlined in Table 1Table 1), primarily among patients with chronic hepatitis C (CHC)[[7], [8], [9], [14], [15], [16], [17], [18], [19]]. These have included the SAFE algorithm which sequentially combines APRI and Fibrotest[8][8], sequential APRI and Hepascore[13][13], concurrent Fibrotest and Fibroscan[14][14] and combining Fibrometer and Fibroscan into one diagnostic model[15][15]. In non-alcoholic fatty liver disease, concurrent Fibroscan and NAFLD Fibrosis Score and sequential FIB4 and BARD have been proposed[[7], [12]], whilst in chronic hepatitis B, a combination of concurrent ALT with Fibroscan followed by the Enhanced Liver Fibrosis Score or the Forns index has been examined[[9], [10]] Overall, these studies demonstrate an increase in diagnostic accuracy and higher predictive values with multiple non-invasive tests compared with single tests, particularly for the determination of moderate degrees of fibrosis (e.g. METAVIR F2+).
Fibroscan available in 729 patients. Mixed liver disease included patients with liver disease realted to alcohol, hepatitis c virus (HCV), hepatitis B virus (HBV), non-alcoholic fatty liver disease (NAFLD) and other causes. “→” indicates sequentially performed tests, “+” indicates concurrent tests, “/” indicates combined tests into one formula, ELF=Enhanced Liver Fibrosis score, APRI= AST to platelet ratio index.
Determining the optimal combination of tests is difficult as studies have been performed in different diseases, used different cut-offs and aimed to predict different stages of fibrosis. Notably, the only independent prospective multi-center evaluation of a range of non-invasive fibrosis tests including Fibrotest, Fibrometer, Hepascore, APRI and Fibroscan, found no difference in percentage of well-classified patients or biopsies avoided between synchronous combinations of the above tests, for the prediction of cirrhosis in patients with CHC[11][11]. However, other studies have found the combination of Fibroscan and a serum test (Fibrotest) is more accurate than the combination of two sequential serum tests incorporated into the SAFE algorithm (APRI and Fibrotest)[[14], [15]]. Using a serum based test in combination with elastography is an appealing strategy as they assess different pathophysiological properties associated with fibrosis, and thus this is currently suggested by EASL Guidelines[20][20]. However, the requirement for a concurrent Fibroscan limits the applicability of this strategy into the community. The ideal algorithm to screen the general population should be a combination of an inexpensive, accessible and highly sensitive test to minimize missed diagnoses, followed by a highly specific test to confirm the diagnosis. To this end, the eLIFT cut-off of eight was chosen to aim for 80% sensitivity, which translated to up to one third of patients with advanced fibrosis being missed but only 3-4% of cirrhotics. One option to minimize false negative cases would be to lower the eLIFT threshold, however this would be at the cost of a greater false positive rate and needless referral for further assessment. Other potential screening tests include FIB4 and APRI in combination given their wide-spread availability and low cost followed by elastrography in those with discordant or increased results above the lower threshold.
Because one test is often inadequate to answer both questions: which patient has minimal liver disease (F0-1) and secondly, which patient has advanced fibrosis (F3-4), it is clear that non-invasive fibrosis markers should not be used in isolation but incorporated into clinical acumen, imaging and other biochemical tests. Overall accuracy and predictive values are improved if two non-invasive fibrosis assessments are used. A sensitive, easy to perform screening test is required for community practitioners (e.g. serum test). More expensive and accurate confirmatory tests such as complex serum models or elastography, can then be done in the hepatology clinic. Currently, further independent comparisons are required to determine the optimal algorithms, however this will also be influenced by factors such including expense and availability. Of equal importance however, is the need to change the paradigm of liver disease assessment in the community to include fibrosis algorithms rather than relying on standard liver enzymes. Until then the current study is a nudge in the right direction.
http://www.journal-of-hepatology.eu/article/S0168-8278(17)30106-X/fulltext
To that goal, the study aimed to develop and validate a stepwise algorithm that could be easily used by all providers to facilitate detection of advanced fibrosis in those with chronic liver disease. The cohort studied consisted of 3754 subjects with chronic liver disease who had undergone liver biopsy and VCTE, who were divided 2:1 into a derivation and validation set. They initially evaluated the utility of both APRI and FIB-4 as an initial screening test. Because FIB-4 outperformed APRI (higher sensitivity), they used it to compare to their new “first line test” which included age, gender, gamma-GT (GGT), AST, platelet count, and prothrombin time (PT). Each of these variables was assigned a number (0-4 depending on the variable and cut off) in the “easy LIver Fibrosis Test (eLIFT), with a score of at least 8 providing 80% sensitivity for advanced fibrosis. In a core group of 1946 subjects, they found that while FIB-4 and eLIFT had similar sensitivity (77-78%), eLIFT had higher specificity (91%), NPV (79%), and PPV (91%) with fewer false positive results than FIB-4, especially in those over age 60. They concluded that eLIFT was better than FIB-4 as a general screening test.
Subsequently, they determined what should be the “second line test” by comparing APRI, FIB-4, liver stiffness by VCTE, and Fibrometer (with or without VCTE) and found that FibrometerVCTE had the highest number of biopsies avoided (81%). A new algorithm was proposed with eLIFT as the initial screening test followed by FibrometerVCTE in those with an eLIFT score ≥ 8 for confirmation. Using this two-step strategy, 46% of patients (33% with eLIFT score <8 and 14% with eLIFT ≥ 8 but FibrometerVCTE <0.384) would not need to see a specialist thus avoiding liver biopsy or additional testing. This two-step approach found that 34% had advanced fibrosis (eLIFT >8 and FibrometerVCTE ≥ 0.715) leaving only 19% with an indeterminate result that might go on to liver biopsy. This combined strategy worked better than FIB-4 or FibrometerVCTE alone in identifying those with advanced fibrosis. Finally, they followed 1275 patients longitudinally (median follow-up 2.9 years) to determine if the eLIFT-FibrometerVCTE strategy could predict liver-related and all-cause mortality. Although eLIFT and FIB-4 had similar performance for all-cause mortality, FibrometerVCTE had the best performance for predicting liver-related mortality.
This study has several strengths; the large patient population enabled comparison of several methods of fibrosis assessment and detected relatively small differences in test performance. Furthermore, the proposed algorithm was accurate across a wide spectrum liver disease increasing applicability and attractiveness to implement in the community. In addition, eLIFT and FibrometerVCTE were demonstrated to be prognostic of liver related outcomes, re-enforcing the appropriateness to use them as tests to guide patient management. Nevertheless, the eLIFT test utilizes some serum tests that may not be routinely ordered by community physicians, such as prothrombin time, which in itself may suffer inter-laboratory variability. In addition, the use of AST and GGT in the eLIFT algorithm, resulted in a reduction in specificity in the setting of alcoholic liver disease. Nevertheless, this was resolved when combined with the FibrometerVCTE. The FibrometerVCTE was developed using the M probe, which provides higher values than the XL probe and has a lower success rate in obesity, limiting applicability in this population. For these reasons, independent validation will be important to confirm the accuracy of the eLIFT-FMVCTE algorithm. Lastly, the study population was not identified from the general population, and so caution should be exercised in trying to extrapolate this algorithm to screen for fibrosis in the general population.
Combining non-invasive algorithms is an attractive way of increasing diagnostic accuracy for liver fibrosis, however uncertainty exists on the best way to combine tests and which tests to use. For example, tests may be used concurrently with discordant results leading to a diagnostic biopsy. Alternatively two modalities may be combined into one diagnostic formula, or thirdly, tests may be used sequentially, with the first being a screening test and the second used following an indeterminate screen or as a confirmatory test for a positive screening test. The approach by Boursier utilizes a combination of firstly, a two-step approach using eLIFT as a screening test and FibrometerVCTE as the confirmatory test, which in itself combines VCTE and a serum test[16][16]. Thus, this approach requires three non-invasive tests to be performed, adding to the complexity and cost.
In addition to the current study, a range of alternative combination strategies have been examined (outlined in Table 1Table 1), primarily among patients with chronic hepatitis C (CHC)[[7], [8], [9], [14], [15], [16], [17], [18], [19]]. These have included the SAFE algorithm which sequentially combines APRI and Fibrotest[8][8], sequential APRI and Hepascore[13][13], concurrent Fibrotest and Fibroscan[14][14] and combining Fibrometer and Fibroscan into one diagnostic model[15][15]. In non-alcoholic fatty liver disease, concurrent Fibroscan and NAFLD Fibrosis Score and sequential FIB4 and BARD have been proposed[[7], [12]], whilst in chronic hepatitis B, a combination of concurrent ALT with Fibroscan followed by the Enhanced Liver Fibrosis Score or the Forns index has been examined[[9], [10]] Overall, these studies demonstrate an increase in diagnostic accuracy and higher predictive values with multiple non-invasive tests compared with single tests, particularly for the determination of moderate degrees of fibrosis (e.g. METAVIR F2+).
| Author | n | Liver Disease | Outcome | Algorithms | Diagnostic Accuracy | Sensitivity | Specificity | NPV | PPV |
|---|---|---|---|---|---|---|---|---|---|
| Boursier 2009 | 332 | Mixed | Cirrhosis | APRI → Fibrotest | 80% | 44% | 93% | 81% | 71% |
| Fibrotest + Fibroscan | 94% | 89% | 96% | 96% | 90% | ||||
| Fibrometer / Fibroscan | 91% | 75% | 97% | 91% | 91% | ||||
| Sebastiani 2009 | 2035 | HCV | Cirrhosis | APRI → Fibrotest | 92% | 90% | 93% | 56% | 99% |
| Castera 2010 | 302 | HCV | Cirrhosis | APRI → Fibrotest | 89% | 86% | 90% | 94% | 78% |
| Fibrotest + Fibroscan | 96% | 89% | 98% | 96% | 95% | ||||
| Sebastiani 2012 | 1013 | HCV | Cirrhosis | APRI → Fibrotest | 91% | 82% | 92% | 98% | 57% |
| APRI + Fibrotest. | 94% | 73% | 97% | 96% | 73% | ||||
| Boursier 2012 | 1785⁎ | HCV | Cirrhosis | APRI → Fibrotest | 89% | 61% | 93% | 95% | 56% |
| Fibrotest + Fibroscan | 94% | 86% | 95% | 98% | 76% | ||||
| Fibrometer / Fibroscan | 87% | - | - | - | - | ||||
| Crisan 2012 | 446 | HCV | Advanced fibrosis | APRI + FIB4 + Fibrometer | 89% | 84% | 91% | 94% | 76% |
| APRI + FIB4 + Fibrotest | 85% | 88% | 83% | 95% | 68% | ||||
| APRI + FIB4 + Fibroscan | 86% | 62% | 100% | 59% | 100% | ||||
| Wong 2014 | 323 | HBV | Advanced fibrosis | Fibroscan + ELF | - | 65-66% | 86-92% | 79-80% | 76-85% |
| Petta 2014 | 321 | NAFLD | Advanced fibrosis | Fibroscan + FIB4 | - | 42-85% | 97-100% | 91-98% | 71-100% |
| Fibroscan + NFS | - | 25-83% | 100% | 93-99% | 100% | ||||
| FIB4 + NFS | - | 14-35% | 100% | 88-91% | 100% | ||||
| Boursier 2017 | 1946 | Mixed | Advanced fibrosis | eLIFT → Fibrometer/Fibroscan | - | 78% | 91% | 79% | 91% |
Fibroscan available in 729 patients. Mixed liver disease included patients with liver disease realted to alcohol, hepatitis c virus (HCV), hepatitis B virus (HBV), non-alcoholic fatty liver disease (NAFLD) and other causes. “→” indicates sequentially performed tests, “+” indicates concurrent tests, “/” indicates combined tests into one formula, ELF=Enhanced Liver Fibrosis score, APRI= AST to platelet ratio index.
Determining the optimal combination of tests is difficult as studies have been performed in different diseases, used different cut-offs and aimed to predict different stages of fibrosis. Notably, the only independent prospective multi-center evaluation of a range of non-invasive fibrosis tests including Fibrotest, Fibrometer, Hepascore, APRI and Fibroscan, found no difference in percentage of well-classified patients or biopsies avoided between synchronous combinations of the above tests, for the prediction of cirrhosis in patients with CHC[11][11]. However, other studies have found the combination of Fibroscan and a serum test (Fibrotest) is more accurate than the combination of two sequential serum tests incorporated into the SAFE algorithm (APRI and Fibrotest)[[14], [15]]. Using a serum based test in combination with elastography is an appealing strategy as they assess different pathophysiological properties associated with fibrosis, and thus this is currently suggested by EASL Guidelines[20][20]. However, the requirement for a concurrent Fibroscan limits the applicability of this strategy into the community. The ideal algorithm to screen the general population should be a combination of an inexpensive, accessible and highly sensitive test to minimize missed diagnoses, followed by a highly specific test to confirm the diagnosis. To this end, the eLIFT cut-off of eight was chosen to aim for 80% sensitivity, which translated to up to one third of patients with advanced fibrosis being missed but only 3-4% of cirrhotics. One option to minimize false negative cases would be to lower the eLIFT threshold, however this would be at the cost of a greater false positive rate and needless referral for further assessment. Other potential screening tests include FIB4 and APRI in combination given their wide-spread availability and low cost followed by elastrography in those with discordant or increased results above the lower threshold.
Because one test is often inadequate to answer both questions: which patient has minimal liver disease (F0-1) and secondly, which patient has advanced fibrosis (F3-4), it is clear that non-invasive fibrosis markers should not be used in isolation but incorporated into clinical acumen, imaging and other biochemical tests. Overall accuracy and predictive values are improved if two non-invasive fibrosis assessments are used. A sensitive, easy to perform screening test is required for community practitioners (e.g. serum test). More expensive and accurate confirmatory tests such as complex serum models or elastography, can then be done in the hepatology clinic. Currently, further independent comparisons are required to determine the optimal algorithms, however this will also be influenced by factors such including expense and availability. Of equal importance however, is the need to change the paradigm of liver disease assessment in the community to include fibrosis algorithms rather than relying on standard liver enzymes. Until then the current study is a nudge in the right direction.
http://www.journal-of-hepatology.eu/article/S0168-8278(17)30106-X/fulltext
No comments:
Post a Comment