Received: 10 May 2017 | Accepted: 30 May 2017

DOI: 10.1002/jso.24742

RESEARCH ARTICLE

The prognostic utility of baseline alpha-fetoprotein for

hepatocellular carcinoma patients

Jack P. Silva BS | Richard A. Gorman BS | Nicholas G. Berger MD |

Susan Tsai MD, MHS | Kathleen K. Christians MD | Callisia N. Clarke MD |

Harveshp Mogal MD | T. Clark Gamblin MD, MS, MBA

Department of Surgery, Division of Surgical

Oncology, Medical College of Wisconsin,
Milwaukee, Wisconsin
Correspondence
T. Clark Gamblin, MD, MS, MBA, Department
of Surgery, Chief, Division of Surgical
Oncology, Medical College of Wisconsin,
9200 West Wisconsin Ave, Milwaukee, WI
53226.
Email: tcgamblin@mcw.edu
Introduction: Alpha-fetoprotein (AFP) has a valuable role in postoperative surveillance
for hepatocellular carcinoma (HCC) recurrence. The utility of pretreatment or baseline
AFP remains controversial. The present study hypothesized that elevated baseline AFP
levels are associated with worse overall survival in HCC patients.
Methods: Adult HCC patients were identified using the National Cancer Database
(2004-2013). Patients were stratified according to baseline AFP measurements into
the following groups: Negative (<20), Borderline (20-199), Elevated (200-1999), and
Highly Elevated (>2000). The primary outcome was overall survival (OS), which was
analyzed by log-rank test and graphed using Kaplan-Meier method. Multivariate
regression modeling was used to determine hazard ratios (HR) for OS.
Results: Of 41 107 patients identified, 15 809 (33.6%) were Negative. Median overall
survival was highest in the Negative group, followed by Borderline, Elevated, and
Highly Elevated (28.7 vs 18.9 vs 8.8 vs 3.2 months; P < 0.001). On multivariate analysis,
overall survival hazard ratios for the Borderline, Elevated, and Highly Elevated groups
were 1.18 (P = 0.267), 1.94 (P < 0.001), and 1.77 (P = 0.007), respectively (reference
Negative).
Conclusion: Baseline AFP independently predicted overall survival in HCC patients
regardless of treatment plan. A baseline AFP value is a simple and effective method to
assist in expected survival for HCC patients.

KEYWORDS
alpha-fetoproteins, biomarkers, carcinoma, hepatocellular, prognosis, survival analysis, tumor

1 | INTRODUCTION diagnosis, prognosis, and treatment of HCC is vital, and serum

Liver cancer was responsible for an estimated 745 000 deaths
worldwide in 2012, making it the second leading cause of cancer
death worldwide.1 The geographical patterns for incidence and
mortality are nearly identical, which suggests a poor prognosis.
Hepatocellular carcinoma (HCC) is the predominant histological
subtype, accounting for 90-95% of all primary liver cancers. Early
biomarkers may assist in management.
Alpha-fetoprotein (AFP) synthesis is repressed in healthy adults,
and elevated levels of AFP are often associated with HCC. AFP is the
most widely used tumor biomarker available for the early detection of
HCC, but its most established role is in the surveillance period.2 The
diagnostic value of serum AFP if often challenged due to its undefined
cut-off point and questionable sensitivity (41-65%).3 Many early-stage

J Surg Oncol. 2017;1–10. wileyonlinelibrary.com/journal/jso © 2017 Wiley Periodicals, Inc. | 1

2 | SILVA ET AL.

TABLE 1 A comparison of clinicopathologic characteristics for hepatocellular carcinoma patients stratified by baseline alpha-fetoprotein (AFP)
values
Highly
Borderline Elevated AFP elevated AFP
Negative AFP AFP (20-199) (200-1 999) (>2 000)
All patients (<20) n = 15 809 n = 13 005 n = 8 183 n = 10 110
Patient characteristics, n (%) n = 47 107 (33.6%) (27.6%) (17.4%) (21.5%) P-value
Age, median (IQR) 62 (55-71) 62 (56-70) 61 (56-69) 61 (56-70) 61 (55-70) <0.001
Gender 0.012
Male 36 049 (76.5) 12 116 (76.6) 9830 (75.6) 6277 (76.7) 7826 (77.4)
Female 11 058 (23.5) 3693 (23.4) 3175 (24.4) 1906 (23.3) 2284 (22.6)
Race <0.001
Caucasian 34 264 (73.7) 12 334 (79.1) 9395 (73.2) 5738 (71.1) 6797 (68.1)
African American 7695 (16.6) 1826 (11.7) 2242 (17.5) 1509 (18.7) 2118 (21.2)
Asian/Pacific Islander 3514 (7.6) 1085 (7.0) 912 (7.1) 651 (8.1) 866 (8.7)
Other 998 (2.2) 344 (2.2) 278 (2.2) 173 (2.1) 203 (2.0)
Hispanic ethnicity 6095 (13.3) 1982 (12.9) 1715 (13.6) 1101 (13.9) 1297 (13.3) 0.155
Charlson-Deyo score <0.001
0 20 964 (44.5) 6764 (42.8) 5689 (43.7) 3744 (45.8) 4767 (47.2)
1 12 822 (27.2) 4347 (27.5) 3685 (28.3) 2226 (27.2) 2564 (25.4)
≥2 13 321 (28.3) 4698 (29.7) 3631 (27.9) 2213 (27.0) 2779 (27.5)
Year of diagnosis <0.001
2004-2007 171 (0.4) 66 (0.4) 40 (0.3) 41 (0.5) 24 (0.2)
2008-2010 11 341 (24.1) 3624 (22.9) 3226 (24.8) 2006 (24.5) 2485 (24.6)
2011-2013 35 595 (75.6) 12 119 (76.7) 9739 (74.9) 6136 (75.0) 7601 (75.2)
Tumor grade <0.001
Well differentiated 5467 (32.5) 2799 (42.8) 1701 (34.6) 609 (23.9) 358 (12.7)
Moderately differentiated 7593 (45.2) 2972 (45.4) 2350 (47.8) 1161 (45.6) 1110 (39.5)
Poorly differentiated 3559 (21.2) 735 (11.2) 821 (16.7) 739 (29.0) 1264 (45.0)
Undifferentiated 196 (1.2) 37 (0.6) 43 (0.9) 39 (1.5) 77 (2.7)
Tumor size <0.001
< 3 cm 12 624 (30.8) 6025 (41.1) 4219 (36.1) 1620 (23.5) 760 (9.9)
3-4.9 cm 10 460 (25.5) 4015 (27.4) 3308 (28.3) 1814 (26.3) 1323 (17.2)
≥5 cm 17 872 (43.6) 4621 (31.5) 4170 (35.7) 3472 (50.3) 5609 (72.9)
Cirrhosis 8935 (80.3) 3498 (78.0) 2684 (82.2) 1380 (81.5) 1373 (81.4) <0.001
Surgical/procedural <0.001
intervention
None 35 577 (75.7) 10 159 (64.4) 9356 (72.2) 6724 (82.5) 9338 (92.6)
Surgery/local destruction 11 395 (24.3 5608 (35.6) 3606 (27.8) 1429 (17.5) 752 (7.5)
Type of surgery <0.001
Wedge/segmentectomy 2293 (32.5) 1099 (31.1) 630 (30.2) 333 (37.4) 231 (42.5)
Lobectomy 1099 (15.6) 520 (14.7) 266 (12.8) 144 (16.2) 169 (31.1)
Total hepatectomy/ 3186 (45.2) 1714 (48.6) 1075 (51.5) 330 (37.1) 67 (12.3)
transplant
Other 471 (6.7) 197 (5.6) 115 (5.5) 83 (9.3) 76 (14.0)
Margin status <0.001
R0 6851 (90.9) 3484 (93.2) 2080 (90.8) 835 (87.1) 452 (82.8)
R1/R2 683 (9.1) 253 (6.8) 212 (9.3) 124 (12.9) 94 (17.2)
(Continues)
SILVA ET AL.
| 3

TABLE 1 (Continued)
Highly
Borderline Elevated AFP elevated AFP
Negative AFP AFP (20-199) (200-1 999) (>2 000)
All patients (<20) n = 15 809 n = 13 005 n = 8 183 n = 10 110
Patient characteristics, n (%) n = 47 107 (33.6%) (27.6%) (17.4%) (21.5%) P-value
Beam radiation 1845 (4.2) 579 (3.9) 475 (4.0) 333 (4.4) 458 (4.8) 0.002
Systemic chemotherapy 21 017 (46.0) 7081 (46.1) 6295 (49.8) 3785 (47.7) 3856 (39.5) <0.001
Palliation 3338 (7.1) 673 (4.3) 718 (5.5) 675 (8.3) 1272 (12.6) <0.001
Analytic stage group <0.001
I 15 234 (36.0) 7094 (49.0) 4737 (40.1) 1995 (27.5) 1408 (16.1)
II 9422 (22.3) 3748 (25.9) 3119 (26.4) 1602 (22.0) 953 (10.9)
III 9560 (22.6) 2109 (14.6) 2346 (19.9) 1998 (27.5) 3107 (35.5)
IV 8078 (19.1) 1525 (10.5) 1602 (13.6) 1674 (23.0) 3277 (37.5)
Facility <0.001
Community cancer program 18 088 (38.9) 4966 (31.8) 4656 (36.1) 3406 (42.1) 5060 (51.1)
Academic cancer program 28 416 (61.1) 10 641 (68.2) 8237 (63.9) 4693 (58.0) 4845 (48.9)

Bold signifies statistically significant P-values (< 0.05).

HCC patients do not express elevated AFP levels, and patients with
chronic diseases like hepatitis and cirrhosis may have elevated serum
AFP in the absence of malignancy.2 However, following definitive
diagnosis of HCC, elevated pretreatment AFP has been described as an
independent predictor of survival and recurrence in several institu-
tional studies and reviews. 3–6
Previous studies analyzing the utility of baseline AFP for prognosis
of HCC show inconsistent results, and are often limited to by small
sample size. No universal agreement on the value of an elevated
baseline AFP has been established. This study sought to examine the
prognostic utility of baseline serum AFP values in a nationally
representative population of HCC patients, with the hypothesis that
increased AFP independently predicts poor overall survival.
ethnicity, treatment center, and year of diagnosis. Disease character-
istics were described by Charlson-Deyo score, tumor grade, tumor size,
stage, cirrhosis, and treatment (local destruction, surgery with margin
status, radiotherapy, chemotherapy, and/or palliation).
Further analysis was performed on a subset of patients who
underwent “curative surgery,” defined as those who underwent
transplantation (code 61 or 75), wedge/segmental resection (code
20-25), lobectomy (30-37), or extended lobectomy (50-52), and were
documented with an R0 margin status. Identical variables were
collected for the subgroup analysis and patients were again stratified
by baseline AFP level as above.
This study was performed following approval from the Institu-
tional Review Board at the Medical College of Wisconsin.

2 | M E TH O D S
2.1 | Patient selection
A retrospective review of the National Cancer Database (NCDB) was
performed (2004-2013). The NCDB is a joint project of the
Commission on Cancer of the American College of Surgeons and
the American Cancer Society. The liver participant user file (primary
site code C220) was used to identify adult patients diagnosed with
hepatocellular carcinoma (ICD-0-3 histological codes 8170-8175)
regardless of treatment or lack thereof. Patients were excluded if they
had no reported AFP lab value (CS Site-Specific Factor 3). Stratification
of the cohort was defined by the highest recorded pretreatment or
baseline AFP level. The four groups compared were defined as
Negative (AFP <20 ng/mL), Borderline (20-199 ng/mL), Elevated (200-
1999 ng/mL), or Highly Elevated (>2000 ng/mL). Stratification of AFP
groups was chosen based on previously published literature describing
cut-off points for baseline AFP predictive of overall survival or disease
recurrence. Patient demographic variables included age, gender, race,
2.2 | Statistical analysis
The primary outcome was overall survival (OS), which was
determined by the number of months from diagnosis until death
or last contact. Secondary outcomes analyzed were 30-day
readmission, and survival rates at 1 month, 1 year, 3 years, and
5 years. Continuous variables were described by medians with
interquartile ranges (IQR), and categorical variables were described
by total numbers with percent frequencies. Data were compared by
Pierson’s Chi-squared test or Kruskal-Wallis test where appropriate.
Overall survival analysis was performed by Kaplan-Meier method
and compared with log-rank test. Univariate Cox regression analysis
was used to identify factors associated with overall survival.
Variables significant in the univariate model (P-value <0.05) were
included in the multivariate analysis. Hazard ratios (HR) and 95%
confidence intervals (CI) were calculated for each potential risk
factor to identify independent predictors of overall survival (P-value
<0.05). To confirm results from prior studies, multivariate analysis
4 | SILVA ET AL.

TABLE 2 Outcomes for hepatocellular carcinoma patients stratified by baseline alpha-fetoprotein (AFP) level
Negative Borderline (AFP Elevated (AFP Highly elevated
Treatment All patients (AFP <20) 20-199) 200-1999) (AFP >2000) P-value
Entire study population
(n = 47 107)
30d readmission, n (%) 809 (1.73) 336 (2.1) 219 (1.7) 99 (1.2) 155 (1.5) <0.001
Median overall survival, 13.4 (3.3-46.1) 28.7 (8.8-79.5) 18.9 (5.8-57.0) 8.8 (2.6-27.5) 3.2 (1.1-9.7) <0.001
months (IQR)
1-year Survival, % (95%CI) 48.1 (47.7-48.4) 69.7 (68.8-70.6) 60.7 (59.7-61.7) 42.6 (41.3-43.9) 21.4 (20.5-22.4) <0.001
5-year Survival, % (95%CI) 19.6 (19.3-19.9) 31.0 (29.0-33.0) 23.5 (21.3-25.8) 13.9 (12.1-15.7) 5.4 (3.6-7.8) <0.001
All curative surgery patients
(n = 5883)
30d readmission, n (%) 368 (6.3) 204 (6.8) 103 (6.0) 39 (5.6) 22 (5.7) 0.487
1-year survival, % (95%CI) 88.0 (87.4-88.5) 93.8 (92.7-94.7) 91.7 (90.1-93.0) 85.3 (82.0-88.1) 79.7 (74.7-83.9) 0.010
5-year survival, % (95%CI) 60.2 (59.3-61.1) 67.4 (62.0-72.2) 64.0 (57.5-69.8) 57.8 (50.4-64.5) 41.1 (31.1-50.9) <0.001
R0 wedge/segmentectomy
(n = 1805)
30d readmission, n (%) 85 (4.8) 43 (4.8) 24 (4.9) 10 (4.0) 8 (5.2) 0.941
1-year survival, % (95%CI) 85.0 (83.9-86.1) 93.5 (91.2-95.2) 88.2 (84.4-91.1) 78.9 (72.2-84.2) 85.4 (77.6-90.7) 0.023
5-year survival, % (95%CI) 47.6 (45.8-49.5) 52.7 (42.1-62.3) 52.3 (39.2-63.9) 46.0 (33.8-57.3) - <0.001
R0 lobectomy/extended
(n = 1224)
30d readmission, n (%) 54 (4.5) 23 (4.1) 12 (4.2) 8 (4.7) 11 (5.8) 0.794
1-year survival, % (95%CI) 79.5 (78.0-80.9) 90.0 (86.7-92.6) 83.2 (77.6-87.6) 79.2 (70.6-85.6) 71.3 (63.1-78.0) 0.003
5-year survival, % (95%CI) 41.2 (39.2-43.3) 54.9 (47.4-61.7) - 43.7 (32.4-54.5) - <0.001
R0 transplantation (n = 2854)
30d readmission, n (%) 229 (8.1) 138 (8.9) 67 (7.0) 21 (7.4) 3 (7.5) 0.387
1-year survival, % (95%CI) 93.0 (92.4-93.5) 95.3 (93.9-96.4) 95.6 (93.9-96.9) 93.5 (89.4-96.0) 94.5 (79.8-98.6) 0.767
5-year survival, % (95%CI) 73.7 (72.6-74.8) 78.5 (71.6-83.8) 76.2 (69.4-81.7) 72.9 (60.8-81.8) 59.6 (30.1-80.0) 0.005

The entire study population is presented in addition to subsets of patients who underwent curative surgery, defined as resection or transplantation with R0
margin status.

was repeated using different AFP cut-off values for the variable of
interest, as described in previous publications.
Statistical analysis was repeated for the curative surgery cohort
using identical methodology. Outcomes were stratified and compared
by surgical procedure and baseline AFP level. However, median
survival was not analyzed in this subset of patients due to survival
times being significantly longer than the follow-up period.
3 | RE SULTS
3.1 | Population characteristics
A total of 118 800 HCC patients were identified in the dataset, of
which 47 107 had numerical AFP lab values and met the inclusion
criteria. Within the study population, 15 809 patients were AFP
Negative (33.6%), 13 005 were Borderline (27.5%), 8183 were
Elevated (17.4%), and 10 110 were Highly Elevated (21.5%). Only
25.3% of patients received a surgical or procedural intervention
(n = 11 395). Of those 11 395 surgical patients, radiofrequency
ablation (RFA) was the most commonly performed procedure
(n = 3618; 31.8%). The 30-day readmission rate was 1.7%, and the
median overall survival (IQR) was 13.4 (3.3-46.1) months. The survival
rates at 1 month, 1 year, 3 years, and 5 years were 87.9%, 48.1%,
27.0%, and 19.6%, respectively.
3.2 | Cohort comparison
The study groups showed significant differences in all variables
describing demographic characteristics, disease burden, and treat-
ment, with the sole exception of Hispanic ethnicity. Notably, Negative
baseline AFP was associated with smaller tumor size, absence of
cirrhosis, surgical intervention, R0 margin status, early stage disease,
and treatment at an academic cancer program (Table 1). Unadjusted
comparison of the groups in the whole study population revealed
variation in all measured outcomes. AFP Negative patients had the
highest 1-year and 5-year survival rate (Table 2). Figure 1 shows the
Kaplan-Meier curve for overall survival, demonstrating shorter median
survival time for patients with higher baseline AFP levels.
SILVA ET AL.
| 5

3.3 | Cox regression

In the Cox regression, each variable in the univariate analysis was
associated with significant differences in overall survival. In the
multivariate model including all variables, factors identified as
independent predictors of worse survival were age, undifferenti-
ated tumor grade, type of surgical treatment, R1/R2 margin
status, palliative therapy, and tumor stage (Table 3). Asian/Pacific
Islander race was associated with improved survival. Lastly,
baseline AFP level was also identified as an independent
predictor of overall survival. With Negative AFP as a reference,
Elevated AFP was a negative predictor of overall survival (HR
1.94; 95%CI 1.39-2.71), as was Highly Elevated AFP (HR 1.77;

FIGURE 1 Kaplan-Meier curve of overall survival in 95%CI 1.17-2.68).

hepatocellular carcinoma patients stratified by baseline alpha- The multivariate model was then retested with the same variables,
fetoprotein (AFP) level but with different methods of classifying baseline AFP (Table 4). Each
of the categorical AFP variables tested showed significant hazard

TABLE 3 Cox regression analysis showing hazard ratios (HR) with 95% confidence intervals (CI) for overall survival in hepatocellular carcinoma
patients
Univariate Multivariate

Variable HR 95%CI P-value HR 95%CI P-value

Age, continuous 1.01 1.01-1.02 <0.001 1.01 1.00-1.03 0.040
Female gender 0.89 0.88-0.91 <0.001 0.92 0.71-1.19 0.516
Race
Caucasian Ref - - Ref - -
African American 1.14 1.12-1.17 <0.001 0.87 0.61-1.23 0.424
Asian/Pacific Islander 0.78 0.76-0.80 <0.001 0.57 0.39-0.85 0.006
Other 0.90 0.85-0.95 <0.001 0.99 0.46-2.15 0.983
Hispanic Ethnicity 0.92 0.90-0.94 <0.001 0.86 0.55-1.34 0.502
Charlson-Deyo Score
0 Ref - - Ref - -
1 0.98 0.96-0.99 0.007 1.10 0.83-1.47 0.499
≥2 1.13 1.12-1.15 <0.001 1.13 0.84-1.53 0.418
Year of diagnosis
2004-2007 Ref - - Ref - -
2008-2010 0.91 0.90-0.93 <0.001 1.06 0.25-4.50 0.936
2011-2013 0.80 0.79-0.82 <0.001 1.03 0.24-4.39 0.967
Tumor Grade
Well differentiated Ref - - Ref - -
Moderately differentiated 1.03 1.00-1.06 0.022 1.18 0.86-1.62 0.299
Poorly differentiated 1.73 1.68-1.79 <0.001 1.43 0.98-2.10 0.067
Undifferentiated 1.54 1.43-1.66 <0.001 2.92 1.19-7.19 0.020
Tumor size
< 3 cm Ref - - Ref - -
3-4.9 cm 1.53 1.50-1.57 <0.001 1.14 0.84-1.56 0.398
(Continues)
 6 | SILVA ET AL.

TABLE 3 (Continued)
Univariate Multivariate

Variable HR 95%CI P-value HR 95%CI P-value

≥5 cm 2.78 2.72-2.84 <0.001 1.38 0.97-1.96 0.071
Cirrhosis 1.18 1.13-1.24 <0.001 1.30 0.98-1.70 0.064
Type of surgery
Total Hepatectomy/Transplant Ref - - Ref - -
Wedge/Segmentectomy 2.47 2.34-2.61 <0.001 1.73 1.21-2.46 0.002
Lobectomy 2.81 2.65-2.99 <0.001 1.88 1.26-2.81 0.002
Other 2.62 2.40-2.85 <0.001 1.79 1.04-3.11 0.037
R1/R2 margin 2.93 2.76-3.11 <0.001 2.49 1.68-3.70 <0.001
Beam radiation 1.20 1.16-1.24 <0.001 0.66 0.23-1.85 0.426
Systemic chemotherapy 0.72 0.71-0.73 <0.001 0.84 0.63-1.11 0.218
Palliation 2.34 2.28-2.41 <0.001 3.93 1.69-9.12 0.001
Analytic Stage Group
I Ref - - Ref - -
II 1.18 1.15-1.21 <0.001 1.41 1.07-1.86 0.014
III 2.82 2.76-2.88 <0.001 1.76 1.21-2.56 0.003
IV 4.62 4.52-4.73 <0.001 4.31 2.09-8.87 <0.001
Facility
Community cancer program Ref - - Ref - -
Academic cancer program 0.56 0.56-0.57 <0.001 0.78 0.60-1.01 0.062
AFP Level
Negative Ref - - Ref - -
Borderline 1.31 1.26-1.36 <0.001 1.18 0.88-1.58 0.267
Elevated 2.07 2.00-2.16 <0.001 1.94 1.39-2.71 <0.001
Highly elevated 3.78 3.64-3.92 <0.001 1.77 1.17-2.68 0.007

Bold signifies statistically significant P-values (< 0.05).

ratios for OS at higher AFP values with the exception the 10 000 ng/ for this subset of patients, but the Kaplan-Meier survival curve
mL cutoff point. (Fig. 2) demonstrated a clear advantage for patients with Negative or
Borderline baseline AFP (P < 0.001).

3.4 | Curative resection of transplantation cohort

A subset of 5883 patients underwent curative resection or
transplantation with R0 margins. Among those “curative surgery”
patients, the median age (IQR) was 60 (54-67), most were male
(73.0%), Caucasian (73.4%), cirrhotic (65.4%), and treated at an
academic cancer program (78.7%). Their tumors were most
commonly moderately differentiated (55.8%), and 43.2% were
under 3 cm. The majority of the cohort had Negative baseline AFP
levels (51.5%), while 29.8% were Borderline, 12.1% were Elevated,
and 6.6% were Highly Elevated. Outcomes for these patients,
stratified by baseline AFP level, are shown in Table 2. Cox univariate
and multivariate regression of curative surgery patients again
revealed that increased baseline AFP level independently predicted
worse overall survival (Table 5). Median survival was not calculated
4 | DISCUSSION
4.1 | Current practices
Screening for HCC in high-risk patients commonly involves serum
AFP measurements. This is often performed in conjunction with
ultrasonography to examine for tumor presence. A rise in AFP or a
questionable lesion on ultrasound indicates the need for cross-
sectional imaging. However, the utility of AFP as a screening
biomarker has been challenged, and some guidelines suggest
foregoing it to rely on ultrasound alone.7 Questionable sensitivity
and specificity limit the use of AFP as a singular screening tool,
but the ease of access and low cost have kept the measurement
relevant as an adjunct to ultrasound.8 Identifying the etiology and
SILVA ET AL.
| 7

TABLE 4 Results from several Cox multivariate analyses depicting 4.2 | Utility of baseline AFP
the hazard ratio for various baseline alpha-fetoprotein values on
overall survival in hepatocellular carcinoma patients Previous studies have reported on the prognostic utility of baseline or
pretreatment AFP levels. AFP has been described as an independent
Multivariate
predictor of overall survival for patients with varying disease severities
Variable Hazard ratio 95%CI P-value
and management plans.12–18 However, results are not consistent;
AFP stratified
Zhang et al and Schraiber et al separately determined that AFP level
Negative Ref
may predict recurrence free survival (RFS) following treatment, but not
Borderline 1.18 0.88-1.58 0.267 overall survival.6,19 Despite the increasing evidence for the predictive
Elevated 1.94 1.39-2.71 <0.001 power of AFP for overall survival in HCC patients, a lack of consistent
Highly elevated 1.77 1.17-2.68 0.007 methodology has prevented consensus. Many studies are limited to
NCDB reported single institutions or small sample sizes, or focused on only one
WNL Ref treatment scheme. The current study examines the prognostic ability
of baseline AFP for a large, nationally representative population of
Borderline 1.33 0.42-4.19 0.631
HCC patients who underwent a variety of treatments. The additional
Elevated 1.17 1.01-1.35 0.039
analysis of patients who underwent R0 resection or transplantation
AFP>20 1.43 1.12-1.84 0.005
allowed further insight into the utility of a preoperative serum AFP
AFP>30 1.47 1.15-1.89 0.002
level. Surgical patients with R0 margins are expected to have an
AFP>100 1.75 1.35-2.26 <0.001 improved prognosis, and a return to baseline AFP is associated with
AFP>200 1.74 1.33-2.28 <0.001 survival. Although the current study could not analyze trends in AFP
AFP>1000 1.45 1.03-2.05 0.032 following surgery, the data showed that pretreatment AFP alone
AFP>10 000 1.32 0.78-2.24 0.305 independently predicted long-term survival following curative resec-

Borderline removed tion or transplantation. Pretreatment AFP may be useful for

comparison to postoperative levels, but also provides independent
AFP<20 Ref
prognostic value.
AFP>200 1.84 1.34-2.52 <0.001

Other variables included in the multivariate model were identical to those

shown in Table 3.
Bold signifies statistically significant P-values (< 0.05).
extent of HCC is important during the diagnostic phase of
management. Cross-sectional imaging, hepatitis serology, assess-
ment of comorbidities, identification of metastases, and liver
function tests are all common in this stage, but serum AFP
measurement is not uniformly performed. For the staging of HCC, a
breadth of classification systems have been independently
validated as diagnostic or prognostic. Some systems like the
Cancer of the Liver Italian Program (CLIP) and Groupe d’Etude et
de Traitement du Carcinome Hepatocellulaire (GETCH) incorporate
AFP into their staging.9,10 Other common methods of classification
like the American Joint Committee on Cancer (AJCC), Okuda,
Barcelona Clinic Liver Cancer (BCLC), and Japanese Integrated
Staging (JIS) do not consider AFP. HCC etiology and outcomes
differ between geographic locations and disease subtypes, so
some staging systems may be better suited for specific patients.
An area of established utility for serum AFP measurements is in
surveillance following treatment. Surgical resection or transplan-
tation have the potential to be curative, but surveillance AFP
measurements are utilized to monitor for recurrence. If pretreat-
ment AFP was documented, a return to baseline is indicative of a
positive treatment response and may be associated with improved
survival.11 Furthermore, the current data showed that baseline AFP
retains its prognostic potential in patients who underwent R0
resection or transplantation.
4.3 | Affecting potential change
Increased serum AFP has been previously associated with poorly
differentiated tumors and increased tumor size, which may influence
survival, and the current data supports this finding.4 After adjusting
for these tumor characteristics in the multivariate analysis, AFP level
remained an independent predictor of overall survival. The mecha-
nism linking serum AFP to decreased survival is not clearly defined,
and further research is still required. The scope of this study pertains
more to the implications of a serum biomarker predicting overall
survival. Although the current data confirms that treatment choice
influences survival, the prognostic value of the baseline AFP level
remained intact regardless of selected treatment. The Cox regression
model accounted for surgical procedure, chemotherapy, radiation,
and palliative care. In the subset of patients who underwent curative
surgery, AFP remained an independent predictor of survival on
multivariate analysis (Table 5). The current data provides expected
outcomes for such surgical patients stratified by procedure and
baseline AFP (Table 2).
AFP is an easily accessible and relatively inexpensive laboratory
value that could be used to discuss expected outcomes for HCC
patients undergoing various management plans.8 Given the survival
variation between patients with different AFP levels (Fig. 1), a baseline
serum AFP measurement is valuable when staging HCC, although
some popular staging systems do not consider AFP. As an independent
predictor of overall survival, this laboratory value has the potential to
influence management irrespective of other disease characteristics.
8 | SILVA ET AL.

TABLE 5 Cox regression analysis showing hazard ratios (HR) with 95% confidence intervals (CI) for overall survival in hepatocellular carcinoma
patients who underwent curative surgery, defined as resection or transplantation with R0 margin status
Univariate Multivariate

Variable HR 95%CI P-value HR 95%CI P-value

Age, continuous 1.03 1.03-1.03 <0.001 1.01 1.00-1.03 0.113
Female gender 1.01 0.95-1.07 0.786 0.95 0.72-1.25 0.717
Race
Caucasian Ref - - Ref - -
African American 1.26 1.17-1.36 <0.001 0.85 0.58-1.24 0.391
Asian/Pacific Islander 0.78 0.71-0.86 <0.001 0.55 0.35-0.87 0.010
Other 0.79 0.64-0.99 0.037 1.07 0.52-2.22 0.849
Hispanic ethnicity 0.76 0.69-0.84 <0.001 0.76 0.47-1.25 0.286
Charlson-Deyo score
0 Ref - - Ref - -
1 0.98 0.92-1.05 0.605 1.12 0.82-1.54 0.476
≥2 0.94 0.88-1.00 0.061 1.26 0.91-1.75 0.163
Year of diagnosis
2004-2007 Ref - - Ref - -
2008-2010 0.91 0.86-0.96 0.002 0.98 0.23-4.22 0.979
2011-2013 0.78 0.72-0.85 <0.001 1.02 0.24-4.42 0.976
Tumor grade
Well differentiated Ref - - Ref - -
Moderately differentiated 1.35 1.26-1.45 <0.001 1.14 0.81-1.59 0.448
Poorly differentiated 1.87 1.72-2.04 <0.001 1.17 0.77-1.77 0.471
Undifferentiated 2.17 1.71-2.75 <0.001 7.32 2.72-19.70 <0.001
Tumor size
< 3 cm Ref - - Ref - -
3-4.9 cm 1.49 1.39-1.60 <0.001 1.14 0.82-1.57 0.431
≥5 cm 2.43 2.28-2.59 <0.001 1.38 0.95-2.00 0.095
Cirrhosis 0.78 0.70-0.87 <0.001 1.46 1.08-1.97 0.013
Procedure
Total Hepatectomy/transplant Ref - - Ref - -
Wedge/segmentectomy 2.28 2.14-2.43 <0.001 2.22 1.51-3.26 <0.001
Lobectomy 2.78 2.60-2.96 <0.001 2.06 1.34-3.17 0.001
Systemic chemotherapy 0.64 0.60-0.69 <0.001 0.88 0.64-1.21 0.432
Analytic stage group
I Ref - - Ref - -
II 1.24 1.16-1.32 <0.001 1.43 1.07-1.91 0.014
III 2.72 2.52-2.93 <0.001 1.77 1.16-2.69 0.008
IV 4.38 3.69-5.20 <0.001 3.16 1.28-7.82 0.013
Facility
Community cancer program Ref - - Ref - -
Academic cancer program 0.70 0.66-0.75 <0.001 0.86 0.65-1.13 0.291
AFP level
Negative Ref - - Ref - -
Borderline 1.15 0.99-1.33 <0.001 1.14 0.85-1.56 0.370
(Continues)
SILVA ET AL.
TABLE 5 (Continued)
Univariate
Variable HR 95%CI
Elevated 1.82 1.52-2.17
Highly elevated 2.66 2.18-3.26
Bold signifies statistically significant P-values (< 0.05).
Following treatment, patients with elevated baseline AFP may be
selected for closer surveillance given their nearly two-fold likelihood of
mortality compared to AFP-negative patients.
4.4 | Limitations
The current study has several limitations. As with any retrospective
database review, selection bias is inherent, although multivariate Cox
regression analysis was performed in an effort to account for the clear
differences between the patient groups on unadjusted analysis.
However, the variables included in the multivariate model are not
comprehensive, and the NCDB database may not be fully adequate in
describing HCC disease status, etiology, and treatment details. Other
known prognostic factors for HCC include vascular invasion, multi-
focality, hepatic function, and hepatitis status. These variables are
not adequately captured in the NCDB, although tumor invasion and
multifocality are considered in the AJCC classification system used to
determine the “Analytic Stage Group” included in the current study. The
database also does not include specific details pertaining to regional
therapy treatments or chosen chemotherapeutic agent. The use of
Sorafenib and hepatic arterial therapy is common in the treatment of
HCC, but their association with baseline AFP and overall survival was
not able to be evaluated. However, the current analysis included and
adjusted for the use of systemic chemotherapy and differences in
surgical or procedural treatments. The difference in median overall
FIGURE 2 Kaplan-Meier curve of overall survival in
hepatocellular carcinoma patients who underwent curative surgery
(resection or transplantation with R0 margin) stratified by baseline
alpha-fetoprotein (AFP) level
| 9
Multivariate
P-value HR 95%CI P-value
<0.001 1.62 1.12-2.36 0.011
<0.001 1.97 1.26-3.09 0.003
survival seen on the Kaplan-Meier curve remains subject to bias from
confounding variables. Serum AFP values collected from the database
were not strictly continuous, but rather reported by units of 10 up to
100 ng/mL, then by units of 100 up to 1000 ng/mL, then by units of
1000 up to 10 000+ ng/mL. Another important aspect of the provided
AFP values is the timing of collection. Only the highest pretreatment or
baseline AFP value is reported, and it is not possible to compare AFP
levels before and after treatment. The lack of follow-up laboratory
values or recurrence data prevented analysis of disease-free survival,
and the value of the biomarker’s trend could not be examined. Despite
these limitations, the findings provide evidence that baseline AFP levels
are valuable in the management of HCC.
5 | CONCL US IO N
In conclusion, the current data suggests that median overall survival for
HCC patients decreases as baseline AFP level increases in an
unadjusted population. Investigation of AFP as a continuous variable
rather than categorically may be useful to further understand the
significance of baseline AFP levels. In comparison to a negative
baseline AFP (<20 ng/mL), elevated (200-1999 ng/mL) and highly
elevated (>2000 ng/mL) AFP levels independently predict worse
overall survival in HCC patients regardless of treatment. The ability to
predict long-term survival persisted for patients who underwent
curative resection or transplantation. The ability to independently
predict patient outcomes validates serum AFP as a valuable
component of initial work-up and staging for HCC. Current HCC
guidelines and recommendations should consider the inclusion of
baseline AFP measurement in their staging methodology, as the result
may influence expected survival.
REFERENCES
1. Ferlay J, Soerjomataram I, Dikshit R, et al. Cancer incidence and
mortality worldwide: sources, methods and major patterns in
GLOBOCAN 2012. Int J Cancer. 2015;136:E359–E386.
2. Zhao YJ, Ju Q, Li GC. Tumor markers for hepatocellular carcinoma. Mol
Clin Oncol. 2013;1:593–598.
3. Debruyne EN, Delanghe JR. Diagnosing and monitoring hepatocellular
carcinoma with alpha-fetoprotein: new aspects and applications. Clin
Chim Acta. 2008;395:19–26.
4. Blank S, Wang Q, Fiel MI, Luan W, Kim KW, Kadri H, et al. Assessing
prognostic significance of preoperative alpha-fetoprotein in hepatitis
B-associated hepatocellular carcinoma: normal is not the new normal.
Ann Surg Oncol. 2014;21:986–994.
10 | SILVA
5. Pinato DJ, Karamanakos G, Arizumi T, Adjogatse D, Kim YW, Stebbing
J, et al. Dynamic changes of the inflammation-based index predict
mortality following chemoembolisation for hepatocellular carcinoma:
a prospective study. Aliment Pharmacol Ther. 2014;40:1270–1281.
6. Schraiber Ldos S, de Mattos AA, Zanotelli ML, et al. Alpha-fetoprotein
level predicts recurrence after transplantation in hepatocellular
carcinoma. Medicine (Baltimore). 2016;95:e2478.
7. Bruix J, Sherman M. Management of hepatocellular carcinoma: an
update. 2010 [Available from: http://www.aasld.org/practiceguide
lines/Documents/HCCUpdate2010.pdf.
8. Gounder PP, Bulkow LR, Meltzer MI, et al. Cost-effectiveness analysis
of hepatocellular carcinoma screening by combinations of ultrasound
and alpha-fetoprotein among Alaska Native people, 1983–2012. Int J
Circumpolar Health. 2016;75:31115.
9. Chevret S, Trinchet JC, Mathieu D, Rached AA, Beaugrand M, Chastang
C. A new prognostic classification for predicting survival in patients with
hepatocellular carcinoma. Groupe d’Etude et de Traitement du
Carcinome Hepatocellulaire. J Hepatol. 1999;31:133–141.
10. A new prognostic system for hepatocellular carcinoma: a retrospective
study of 435 patients: the Cancer of the Liver Italian Program (CLIP)
investigators. Hepatology. 1998;28:751–755.
11. Riaz A, Ryu RK, Kulik LM, et al. Alpha-fetoprotein response after
locoregional therapy for hepatocellular carcinoma: oncologic marker
of radiologic response, progression, and survival. J Clin Oncol. 2009;
27:5734–5742.
12. Llovet JM, Pena CE, Lathia CD, Shan M, Meinhardt G, Bruix J. Plasma
biomarkers as predictors of outcome in patients with advanced
hepatocellular carcinoma. Clin Cancer Res. 2012;18:2290–2300.
13. Gomaa AI, Hashim MS, Waked I. Comparing staging systems for
predicting prognosis and survival in patients with hepatocellular
carcinoma in Egypt. PLoS ONE. 2014;9:e90929.
SYNOPSIS
Elevated baseline AFP independently predicted worse overall survival for a nationally representative population of hepatocellular carcinoma
patients, regardless of treatment.
ET AL.
14. Park MS, Kim SU, Park JY, et al. Combination treatment of
localized concurrent chemoradiation therapy and transarterial
chemoembolization in locally advanced hepatocellular carcinoma
with intrahepatic metastasis. Cancer Chemother Pharmacol. 2013;
71:165–173.
15. Raoul JL, Bruix J, Greten TF, et al. Relationship between baseline
hepatic status and outcome, and effect of sorafenib on liver function:
SHARP trial subanalyses. J Hepatol. 2012;56:1080–1088.
16. Song MJ, Bae SH, Chun HJ, et al. A randomized study of cisplatin and
5-FU hepatic arterial infusion chemotherapy with or without
adriamycin for advanced hepatocellular carcinoma. Cancer Chemother
Pharmacol. 2015;75:739–746.
17. Wan P, Xia Q, Zhang JJ, et al. Liver transplantation for hepatocellular
carcinoma exceeding the Milan criteria: a single-center experience.
J Cancer Res Clin Oncol. 2014;140:341–348.
18. Santambrogio R, Opocher E, Costa M, et al. Hepatic resection for
“BCLC stage A” hepatocellular carcinoma. The prognostic role of
alpha-fetoprotein. Ann Surg Oncol. 2012;19:426–434.
19. Zhang L, Ge NL, Chen Y, et al. Long-term outcomes and prognostic
analysis of radiofrequency ablation for small hepatocellular carci-
noma: 10-year follow-up in Chinese patients. Med Oncol. 2015;
32:77.
How to cite this article: Silva JP, Gorman RA, Berger NG,
et al. The prognostic utility of baseline alpha-fetoprotein for
hepatocellular carcinoma patients. J Surg Oncol. 2017;1–10.
https://doi.org/10.1002/jso.24742