Alpha-fetoprotein is a predictor of outcome in acetaminophen-induced liver injury


  • See Editorial on Page 19

  • Conflict of interest: Nothing to report.


An increase in alpha-fetoprotein (AFP) following hepatic necrosis is considered indicative of hepatic regeneration. This study evaluated the prognostic value of serial AFP measurements in patients with severe acetaminophen-induced liver injury. Prospectively, serial measurements of AFP were performed in 239 patients with acetaminophen intoxication and a peak alanine aminotransferase (ALT) level above 1,000 U/L. AFP was measured using an enzyme-linked immunoassay (EIA) with a detection limit below 0.4 μg/L. The optimum threshold of AFP to discriminate nonsurvivors was identified. An increase in AFP above 4 μg/L occurred in 158 (79%) of 201 survivors compared with 11 of 33 nonsurvivors (33%; P < .00001). The increase in AFP occurred a mean of 1.0 days (range, −2 to +6 days) after peak ALT in survivors compared with 4.1 days (range, +2 to +7 days) in nonsurvivors (P < .00001). Starting on the day of peak ALT, AFP values were significantly higher in survivors than in nonsurvivors. A threshold AFP of 3.9 μg/L on day +1 after peak ALT to identify nonsurvivors had a sensitivity of 100%, a specificity of 74%, a positive predictive value of 45%, and a negative predictive value of 100%. In conclusion, an increase in AFP was strongly associated with a favorable outcome in patients with acetaminophen-induced liver injury. AFP may be useful as a supplement to existing prognostic criteria. We suggest that the introduction of highly sensitive EIAs for the detection of AFP will require a reevaluation of AFP as a prognostic marker in acute nonneoplastic liver disease. (HEPATOLOGY 2005;41:26–31.)

Acetaminophen-induced hepatotoxicity is the most common cause of fulminant hepatic failure (FHF) in Denmark, the United Kingdom, and the United States.1–3 The only treatment option that radically improves the outcome of FHF is emergency liver transplantation.4 However, because some patients with FHF—particularly those with acetaminophen-induced FHF—have the potential to survive without transplantation, it is important for the clinician to identify patients with the capacity for spontaneous recovery early.

The prognosis of acetaminophen-induced FHF depends on the extent of liver damage, the ability to avoid or control for potentially lethal extrahepatic complications, and the patient's capacity for hepatic regeneration. Of the several markers proposed for prognostic prediction in acetaminophen-induced FHF, the majority assess only the extent of hepatic dysfunction or the degree of multiorgan dysfunction associated with FHF.4–10 Recently, hypophosphatemia has been related to hepatic regeneration; otherwise, regenerative markers have not been extensively studied in acetaminophen-induced liver injury.11

Serum alpha-fetoprotein (AFP) has been used primarily as a marker of hepatocellular carcinoma, although it is now recognized that elevation of AFP regularly occurs in a range of nonneoplastic liver disease.12–16 Following acute liver injury with extensive necrosis, an increase in AFP is interpreted as a sign of dedifferentiated hepatic regeneration.12 In FHF, AFP has been studied predominantly in patients with non–acetaminophen-induced liver failure, and the majority of studies find that an elevation of AFP is associated with a favorable prognosis.15, 17–25 Only a few, older studies have included patients with acetaminophen poisoning.15, 19 Thus, the prognostic value of AFP in acetaminophen-induced hepatotoxicity is widely undetermined.

In this study, we prospectively evaluated the prognostic value of serial measurements of AFP in 239 patients with severe acetaminophen-induced liver injury.


FHF, fulminant hepatic failure; AFP, alpha-fetoprotein; ALT, alanine aminotransferase; INR, international normalized ratio; KCH, King's College Hospital; EIA, enzyme-linked immunoassay.

Patients and Methods


The study was performed on 239 patients admitted with acetaminophen-induced hepatotoxicity (peak alanine aminotransferase [ALT] level above 1,000 U/L) at the Department of Hepatology, Copenhagen University Hospital, from 1999 to 2002. There were 148 women (62%) and 91 men. The median age was 41 years (interquartile range, 28-53 years; range, 12-80 years). In 219 cases, the acetaminophen poisoning resulted from ingestion of a well-defined single overdose. In the remaining 20 cases, the time of ingestion was unknown or the poisoning was the result of multiple or chronic overdosing. For the single overdose cases, the median acetaminophen overdose was 35 g (interquartile range, 23-50 g) and the median time from overdose to initiation of N-acetylcysteine treatment was 30 hours (interquartile range, 18-49 hours). Of the 239 patients, 70 progressed into FHF defined by the development of hepatic encephalopathy.26 Of those 70 patients, 32 survived spontaneously and 33 died without liver transplantation. The remaining 5 patients, who were transplanted, were excluded from the following analyses.


For each case, all measurements of AFP, international normalized ratio (INR), and ALT were recorded prospectively by the principal author (L.E.S.). The measurements were related to the time since overdose (when known), the day of peak ALT, and the day of onset of hepatic encephalopathy (in patients with FHF). AFP was measured using a microparticle enzyme immunoassay (AxSYM AFP; Abbott Laboratories, Abbott Park, IL), which has sensitivity better than 0.4 μg/L and a linear range of 0.5 to 350 μg/L (normal range, 0.0-11.0 μg/L [≈ng/mL]). Arterial lactate concentrations at the admission and after fluid resuscitation were recorded retrospectively by review of the hospital charts and were available in 165 and 164 patients, respectively.

The Department of Hepatology at Rigshospitalet, Copenhagen University Hospital, is the tertiary care center of liver disease in Denmark, and the majority of patients are transferred from other hospitals. All patients were treated and monitored in accordance with the department standard. Since 1999, daily measurement of AFP has been routine in severe cases of acetaminophen-induced hepatotoxicity. In accordance with Danish recommendations, all patients with suspected acetaminophen overdose regardless of severity were treated with N-acetylcysteine without delay upon primary admission.27 The standard regimen consists of infusion of 150 mg/kg N-acetylcysteine intravenously as a bolus, then 50 mg/kg over 4 hours, followed by repeated infusions of 100 mg/kg over 16 hours until three consecutive recovering values of the INR have been demonstrated. Mechanical ventilation, hemodialysis, and plasmapheresis were instituted when indicated by the clinical condition of the patients. Prognosis was estimated by the King's College Hospital (KCH) selection criteria for urgent liver transplantation.5


The Mann-Whitney test was used for the comparison of a variable between two subgroups, Fisher's test was used for comparison of frequencies, and Spearman's test was used for correlations. For multivariate analysis a logistic regression analysis was applied. Receiver-operating characteristic analysis was used to identify the optimum threshold values to discriminate nonsurvivors. Data were analyzed using the Statistica 6.0 statistical software (StatSoft Inc., Tulsa, OK). A P value of less than .05 was considered statistically significant.


Starting on the third day (i.e., 72-96 hours) after the acetaminophen overdose, AFP values were significantly higher in survivors than in nonsurvivors (Fig. 1). A peak ALT value was identifiable in all patients and appeared at a mean of 2.9 days (range, 1-6 days) after the overdose. When AFP values were related to the day of peak ALT, AFP values were significantly higher in survivors than in nonsurvivors starting on the day of peak ALT (Fig. 2). Whereas the most striking increase in AFP occurred in the survivor group, a smaller increase in AFP was noticed in the nonsurvivor group (Table 1). Thus, an increase in AFP above 4 μg/L (arbitrarily selected level) was seen in 11 (33%) of 33 nonsurvivors compared with 158 of 201 survivors (79%; Fisher exact test, P < .00001). This increase occurred significantly later in the 11 nonsurvivors (mean, 4.1 days [range, +2 to +7 days] after peak ALT) than in the 158 survivors (mean, 1.0 days [range, −2 to +6 days] after peak ALT; Mann-Whitney test, P < .00001).

Figure 1.

Sequential AFP values (mean and SE) in survivors and nonsurvivors relative to the time of overdose. *P < .05. OD, overdose.

Figure 2.

Sequential AFP values (mean and SE) in survivors and nonsurvivors relative to the day of maximum ALT. ALT, alanine aminotransferase.

Table 1. AFP and INR Values in the Days Following the Day of Maximum ALT in Survivors and Nonsurvivors
DayAFP (μg/L)INR
  • NOTE. All values are given as mean ± SD.

  • *

    P < .01,

  • P < .00001 between survivors and nonsurvivors (Mann-Whitney test).

Day 0 (maximum ALT)4.2 ± 3.12.6 ± 1.1*3.3 ± 1.44.9 ± 1.9
Day +19.2 ± 9.02.4 ± 0.82.4 ± 1.24.5 ± 1.5
Day +219.2 ± 19.72.9 ± 1.71.9 ± 0.74.3 ± 1.7
Day +331.5 ± 44.73.7 ± 2.91.8 ± 0.64.2 ± 1.8
Day +436.0 ± 44.74.9 ± 5.91.9 ± 0.63.8 ± 1.9
Day +563.2 ± 68.08.1 ± 13.61.9 ± 0.64.1 ± 1.7
Day +662.0 ± 68.05.2 ± 3.5*2.0 ± 0.64.4 ± 1.7
Day +768.6 ± 77.96.8 ± 6.0*1.9 ± 0.64.1 ± 1.3

INR values were significantly lower in survivors than in nonsurvivors starting on the day of peak ALT (Table 1). The typical time course of INR was an increase to a distinct peak followed by a rapid decrease. This pattern was observed in 199 (99%) of the 201 survivors compared with 1 of of the 33 nonsurvivors (3%; Fisher: P < .00001). In the remaining 2 survivors and 32 nonsurvivors, INR either showed no decrease or the time course was protracted. In the 158 survivors who had an increase in AFP, the increase in AFP preceded or coincided with the peak in INR in 47 cases (30%), whereas the peak in INR in the other 111 cases preceded the increase in AFP.

In Table 2, the prognostic value of AFP and INR measurements on days 0, +1, and +2 (after peak ALT) were assessed and compared with those of the KCH criteria. The discriminative values used in Table 2 were identified via receiver-operating characteristic analysis (Fig. 3). When corresponding values of AFP and INR on day +1 were examined (Fig. 4), all nonsurvivors were characterized by the combination of an AFP below 3.9 μg/L and an INR above 2.4. AFP on day +1 showed a positive correlation with ALT on day +1 (Spearman r = 0.18; P = .02) and negative correlations with INR on day +1 (Spearman's test, r = −0.50; P < .00001), bilirubin on day +1 (Spearman's test, r = −0.21; P = .004), arterial lactate at admission (Spearman, r = −0.51; P < .00001), arterial lactate after fluid resuscitation (Spearman's test, r = −0.36; P = .00002), and patient age (Spearman's test, r = −0.17; P = .02). In a logistic regression analysis including the above variables as independent variables, a high AFP value was independently associated with survival.

Table 2. Assessment of AFP and INR as Prognostic Indicators in Comparison With the KCH Criteria in 234 Patients With Acetaminophen-Induced Hepatic Injury
  1. Abbreviations: x, number of patients fulfilling the criterion; n, number of patients eligible for analysis; PPV, positive predictive value; NPV, negative predictive value.

AFP <3.3 μg/L at day 0 (maximum ALT)101/1902480522493
AFP <3.9 μg/L at day +174/188331007445100
AFP <4.8 μg/L at day +229/1272384925998
INR >3.4 at day 0 (maximum ALT)92/2252374642594
INR >2.6 at day +174/2303194784299
INR >2.7 at day +236/1992288926198
AFP <3.9 and INR >2.4 at day +154/188331008661100
KCH criteria30/2342370977795
Figure 3.

Receiver-operating characteristic curves plotting sensitivity vs. 1-specificity for different cutoff values of AFP and INR on days 0, +1, and +2 after peak ALT. The identity line corresponds with a complete lack of discriminative power. AFP, alpha-fetoprotein; INR, international normalized ratio.

Figure 4.

Corresponding values of AFP and INR on day +1 after the day of maximum ALT for nonsurvivors (▴) and survivors with (▵) or without (○) hepatic encephalopathy. INR, international normalized ratio.

Among the 65 patients with FHF, AFP values were significantly higher in survivors than in nonsurvivors starting on the day of onset of hepatic encephalopathy (Fig. 5). Thirty patients with FHF fulfilled the KCH criteria. Twenty-three (77%) of the 30 patients who fulfilled the KCH criteria died compared with 10 (29%) of the 35 patients who did not (Fisher exact test, P = .0002). The individual time courses of AFP in patients with FHF who fulfilled or did not fulfill the KCH criteria are shown in Figs. 6 and 7, respectively.

Figure 5.

Sequential AFP values (mean and SE) relative to the day of onset of hepatic encephalopathy in patients who survived or died from FHF. *P < .05. HE, hepatic encephalopathy.

Figure 6.

Individual time courses of AFP (on a logarithmic scale) in survivors (○) and nonsurvivors (●) who fulfilled the KCH criteria. HE, hepatic encephalopathy.

Figure 7.

Individual time courses of AFP (on a logarithmic scale) in survivors (○) and nonsurvivors (●) who had fulminant hepatic failure without fulfilling the KCH criteria. AFP, alpha-fetoprotein.


The results of the present study show that an increase in AFP was strongly associated with a favorable outcome in patients with severe acetaminophen-induced liver injury. Even though an increase in AFP was not exclusively seen in survivors, it was both more frequent and detectable at an earlier stage in survivors compared with nonsurvivors.

The study illustrates that AFP is a dynamic parameter, which is why it is important in analyses to synchronize the time courses in different patients. In survivors, an increase in AFP was generally detectable 72 to 96 hours after the acetaminophen overdose. However, relating measurements to the time since overdose has its disadvantages, because information on time of overdose is often unreliable, unavailable, or even indefinable, as in cases of chronic or multiple overdosing. Standardizing biochemical measurements by using the day of peak ALT as a reference point is a practical approach, because a distinct peak in ALT is a common characteristic of FHF.24 In the present study, the timing of the AFP measurements proved very important, because an increase in survivors usually had appeared by day +1 (after peak ALT), whereas in nonsurvivors it did not appear earlier than day +2. Consequently, day +1 was the optimum time for AFP to discriminate survivors as illustrated by a high sensitivity and negative predictive value.

Coagulation factors are among the most important prognostic indicators in acetaminophen-induced FHF.5, 7, 28 In particular, continued improvement in serial determinations of prothrombin is strongly indicative of recovery.6, 29 This result was confirmed in the present study, in which a rapid decrease in INR was seen in virtually all (99%) of the survivors but virtually none (3%) of the nonsurvivors. Thus, an increase in AFP mainly provided additional prognostic information if it preceded the decrease in INR, which it did in 47 cases. Also, the information from corresponding values of AFP and INR appeared to be supplementary because only patients with the combination of a low AFP and a high INR value were at risk of dying. If INR is considered a marker of hepatic synthetic function and AFP a marker of hepatic regenerative activity, the interpretation would be that patients with severe synthetic dysfunction require extensive regeneration to survive, whereas patients with less severe synthetic dysfunction do not have the same requirements for regeneration.

Even though the KCH selection criteria have gained considerable acceptance, their discriminatory capability is not perfect.4, 5, 11, 30, 31 In the present study, 7 patients survived despite fulfilling the KCH criteria, and 10 patients died without fulfilling the criteria. Serial AFP measurements may provide additional information and serve as a guideline in marginal cases. Thus, an increase in AFP was preferentially seen in patients who eventually survived, even in the cases with a more protracted course. Alternatively, some patients with a protracted course apparently survived the initial insult of FHF, but subsequently showed no signs of regeneration and eventually died if they did not receive transplants. In those cases, a persistently low AFP might be interpreted as regenerative failure.

Because the original electrophoresis techniques used for determination of AFP were unable to detect levels below 200 to 500 μg/L, an elevation of AFP was almost exclusively seen in cases of hepatocellular carcinoma.12–15 The recognition that elevation of AFP regularly occurs in nonneoplastic liver disease was a direct result of the development in the 1970s of the radio immunoassay, which allowed detection of AFP in amounts as low as 5 to 20 μg/L.12, 13, 17 In the present study, the discriminative values identified for AFP lay within normal range and below even the optimal detection limit of the radio immunoassay of 5 μg/L. Thereby, our findings were dependent on the use of the even more sensitive enzyme-linked immunoassay (EIA), which has a detection limit for AFP below 0.5 μg/L. Whereas the improved sensitivity of the EIA is unlikely to be clinically relevant for the use of AFP as a tumor marker, it may significantly increase the usefulness of AFP in nonneoplastic liver disease. Thus, the introduction of a new technology (EIA) may once again require that AFP is re-evaluated as an early prognostic indicator in acute liver disease. Previously, AFP in acetaminophen poisoning had only been studied using the radio immunoassay.15, 19 However, more recent studies in non–acetaminophen-induced FHF using the EIA have produced results very similar to those in the present study.24, 25 AFP EIA may be performed as a routine analysis at a very limited cost (approximately $9 at our laboratory), which makes it practical as a prognostic marker for clinical use.

In conclusion, an elevation of AFP above 3.9 μg/L on the day after peak ALT was a highly discriminative indicator of survival from severe acetaminophen-induced liver injury. In many cases, the increase in AFP preceded the decrease in INR and thereby provided additional prognostic information. In cases of acetaminophen-induced FHF, AFP measurements may supplement the KCH criteria in marginal or protracted cases.