ORIGINAL ARTICLE – HEPATOBILIARY TUMORS

Cure Model Survival Analysis After Hepatic Resectionfor Colorectal Liver Metastases

Alessando Cucchetti, MD1, Alessandro Ferrero, MD, PhD2, Matteo Cescon, MD, PhD1, Matteo Donadon, MD3,

Nadia Russolillo, MD2, Giorgio Ercolani, MD1, Giacomo Stacchini, MD1, Federico Mazzotti, MD1, Guido Torzilli,

MD, PhD3, and Antonio Daniele Pinna, MD, PhD1

1Department of Medical and Surgical Sciences, S. Orsola-Malpighi Hospital, Alma Mater Studiorum - University of

Bologna, Bologna, Italy; 2Department of General and Oncological Surgery, Ospedale Mauriziano Umberto I, Turin, Italy;3Department of Hepatobiliary Surgery, Humanitas Research Hospital, School of Medicine - University of Milan, Milan,

Italy

ABSTRACT

Background. Statistical cure is achieved when a patient

population has the same mortality as cancer-free individ-

uals; however, data regarding the probability of cure after

hepatectomy of colorectal liver metastases (CLM) have

never been provided. We aimed to assess the probability of

being statistically cured from CLM by hepatic resection.

Methods. Data from 1,012 consecutive patients undergo-

ing curative resection for CLM (2001–2012) were used to

fit a nonmixture cure model to compare mortality after

surgery to that expected for the general population matched

by sex and age.

Results. The 5- and 10-year disease-free survival was 18.9

and 15.8 %; the corresponding overall survival was 44.3

and 32.7 %. In the entire study population, the probability

of being cured from CLM was 20 % (95 % confidence

interval 16.5–23.5). After the first year, the mortality

excess of resected patients, in comparison to the general

population, starts to decline until it approaches zero 6 years

after surgery. After 6.48 years, patients alive without tumor

recurrence can be considered cured with 99 % certainty.

Multivariate analysis showed that cure probabilities range

from 40.9 % in patients with node-negative primary tumors

and metachronous presentation of a single lesion \3 cm, to

1.5 % in patients with node positivity, and synchronous

presentation of multiple, large CLMs. A model for the

calculation of a cure fraction for each possible clinical

scenario is provided.

Conclusions. Using a cure model, the present results

indicate that statistical cure of CLM is possible after hep-

atectomy; providing this information can help clinicians

give more precise answer to patients’ questions.

Colorectal cancer is a major cause of mortality

throughout the world.1, 2 At the time of the initial diag-

nosis, 20–25 % of patients will have clinically detectable

colorectal liver metastases (CLM), and an additional

40–50 % of patients will develop liver metastases after

resection of the primary tumor; approximately 20–30 % of

these patients will have the liver as the exclusive site of

disease.3,4 These latter patients could be considered as

having an organ-specific confined disease, so that hepatic

resection can be viewed as a potentially curative treat-

ment.4 Recently published literature suggests that long-

term survival can be achieved, with a 10-year survival rate

ranging from 17 to 28 %, but whether resected patients can

be considered cured is currently unknown.4,5 In epidemi-

ology, and as endorsed by the United States National

Cancer Institute, cure is said to occur when the mortality of

patients, treated for a specific disease, returns to the same

level as that of the general population; these patients can be

considered statistically cured from the specific disease

because they are just as likely to die as a member of the

general population.6–8 Cure rate models represent survival

models incorporating a cure fraction. They assume that a

proportion of subjects exists who will never die as a

Electronic supplementary material The online version of thisarticle (doi:10.1245/s10434-014-4234-0) contains supplementarymaterial, which is available to authorized users.

� Society of Surgical Oncology 2014

First Received: 30 July 2014

M. Cescon, MD, PhD

e-mail: matteo.cescon@unibo.it

Ann Surg Oncol

DOI 10.1245/s10434-014-4234-0

consequence of the specific disease that has been treated.

Survival after specific treatment can be considered as the

sum of a fraction of patients who have mortality similar to

that expected in the general population and who can be

considered cured of their disease, and that of patients who

are bound to die after being treated.6–8 The cure fraction

can provide a useful measure of the probability of success

of a therapeutic approach; however, this aspect, that is

survival incorporating a cure fraction, has never been

properly investigated in the setting of CLM. The main aim

of the present study was to assess the probability of being

cured from CLM by hepatic resection using a cure-rate

model.

METHODS

Study Population

Between January 2001 and December 2012, 1130 con-

secutive patients undergoing first curative resection (R0)

for CLM without extrahepatic disease were identified at

three Italian tertiary referral hospitals; 118 patients were

excluded because of incomplete clinical data. The study

population finally consisted of 1,012 patients: 507 were

from the Ospedale Mauriziano Umberto I, Turin; 348 were

from the S. Orsola-Malpighi Hospital, Bologna; and 157

were from the Istituto Clinico Humanitas, Rozzano, Milan.

Patients were deemed to have resectable disease in the

presence of an adequate performance status if the metas-

tases could be completely resected and if the future

remnant liver volume would be adequate. Standard

demographic and clinical data were collected for each

patient. The number, size, and distribution of the hepatic

metastases were recorded as well as the extension of the

hepatectomy, according to the International Hepato-Biliary

Association nomenclature.9 Primary tumor stage, presen-

tation of CLM (synchronous vs. metachronous), time

elapsed from the resection of the primary tumor to clinical

evidence of metastases [disease-free survival (DFS) from

colon surgery], and the adoption of adjuvant therapy, with

respect to the hepatic surgery, were also detailed. The

chemotherapy strategy was prescribed by the oncologist in

charge. The follow-up included clinical examination, esti-

mation of serum carcinoembryonic antigen level, and

computed tomography or magnetic resonance imaging at 3,

6, and/or every 12 months, as indicated by the oncologist

in charge and on the basis of the clinical course. The fol-

low-up was carried out by assessing the records of the

respective institutions regarding laboratory, interventional,

and pathologic data, and by telephone contact directly with

the patient, or alternatively with the referring clinician for

patients residing in regions distant from hospitals.

Cure Fraction Model

Mathematical details of the cure model are provided in

the Supplementary Materials. The first requirement for the

application of the cure model is the statistical plausibility

of cure.10 This means that cure time should occur within a

reasonable time frame so as to ensure that a cure is plau-

sible within the patient’s lifetime. If a proportion of

patients who are not expected to die of disease exists in the

study population, the Kaplan–Meier survival curve will

tend to a plateau on the y-axis (Fig. 1).7 In the present

analysis, we used DFS after hepatic resection as the pri-

mary survival measure for the cure model. We decided to

use DFS instead of disease-specific survival (which is

calculated from the time of surgery until cancer-related

death) because this latter measure does not take into

account, as events, patients who are alive with recurrence,

and it is not correct to define these patients as cured of their

disease.4,5,11 For the same reason, overall survival was used

as secondary survival measure for completeness of results.

DFS was calculated from the time of the hepatectomy until

clinical evidence of recurrence or patient death. The

mathematical model applied here is the nonmixture cure

fraction model that was chosen because it has a background

in modeling of tumor recurrence.12,13,17 The estimations of

expected survival of the general population (the compara-

tor used for cure fraction calculation) at the time of patient

event (death or recurrence) were derived from population

survival tables obtained from the Italian National Institute

of Statistics, matched by age and sex.14 Details regarding

statistical analysis are reported in the Supplementary

Materials.

0

0.2

.4.6

.81

2 4

Disease-free Overall

6

Years

Surv

ival

8 10 12

FIG. 1 Overall and disease-free survival (DFS) of 1,012 consecutive

patients who underwent resection for colorectal liver metastases

(CLM) (gray areas represent 95 % confidence intervals). As can be

noted, DFS curve tends to flatten at end of fourth year after resection,

confirming that the cure hypothesis can be accepted 7

N. A. Cucchetti et al.

RESULTS

Baseline characteristics of the study population (1,012

patients) are detailed in Table 1. Within a median follow-

up of 47.8 months (range, 1 day to 12 years), 618 patients

(61.1 %) experienced tumor recurrence, and 374 patients

(37.0 %) died. The 30- and 90-day postoperative death

rates were 0.3 and 1.0 %, respectively. The 3-, 5-, and

10-year DFS rates were 26.0, 18.9, and 15.8 %, respec-

tively, and the corresponding overall survival rates were

62.1, 44.3, and 32.7 % (Fig. 1). Tumor recurrence was

only intrahepatic in 284 patients, both intra- and extrahe-

patic in 121, and extrahepatic only in 195 patients (data not

available for 18 patients). Tumor progression accounted for

77.2 % of the causes of death. Relationships between

clinical variables are detailed in the Supplementary

Materials.

The cure model converged for the entire study popula-

tion and for each subgroup analyzed (p \ 0.001 in all

cases). In the entire study population, the probability of

being cured from CLM by hepatic resection was 20 %

(95 % confidence interval 16.5–23.5), and the median

survival of uncured patients was 0.9 years (Fig. 2a). The

excess of hazard after surgery is plotted in Fig. 2b and

starts from a 25 % increased risk of death early after

resection with respect to the general population. In the first

postoperative year, the excess of hazard increased to

approximately 60 % in the entire group and up to 90 % in

uncured patients. At the end of the first year after surgery,

the hazard of uncured patients progressively increased over

TABLE 1 Baseline characteristics of the study population used for

cure fraction estimation

Variable All patients

(n = 1,012)

Age, year, median (IQR) 63 (56–70)

Age C 60 year 650 (64.2 %)

Male sex 619 (61.2 %)

Primary tumor T stage

T1 12 (1.2 %)

T2 91 (9.0 %)

T3 763 (75.4 %)

T4 146 (14.4 %)

Primary tumor nodal status

N0 312 (30.8 %)

N1 444 (43.9 %)

N2 256 (25.3 %)

Synchronous presentation 537 (53.1 %)

Disease-free interval from colon surgery C 24 month 226 (22.3 %)

Largest tumor size, median (IQR) 3.2 (2.0–5.0)

\3 cm 413 (40.8 %)

3–5 cm 386 (38.1 %)

[5 cm 213 (21.1 %)

No. of liver metastases

1 408 (40.3 %)

2 195 (19.3 %)

3 125 (12.4 %)

C4 284 (28.1 %)

Adjuvant chemotherapy (n = 830) 641 (77.2 %)

Extension of hepatectomy

Single or multiple wedges 443 (43.8 %)

Segmentectomy with or without wedges 103 (10.2 %)

Bisegmentectomy with or without wedges 168 (16.6 %)

Major hepatectomy with or without wedges 298 (29.4 %)

Any postoperative complications (Dindo [ 1) 330 (32.6 %)

IQR interquartile range (25th–75th percentiles)

0

0.5

11.

50

.2.4

.6.8

1

2 4

Whole group Uncured patients

6

Years

Exc

ess

haza

rd r

ate

Rel

ativ

e su

rviv

al

8 10 12

0 2 4 6

Years

A

B

8 10 12

FIG. 2 Cure model results. a Relative survival of entire group of

patients and uncured patients. In entire group, from sixth year after

surgery onward, survival curve reaches plateau at 20 %, which

represents cure fraction. b Excess hazard rate of entire study group

and of uncured patients. At end of the first year after surgery, excess

of hazard starts to decrease until it approaches zero at end of 6 years

after surgery. Conversely, in uncured patients, excess of hazard

progressively increases over time

Statistical Cure for Colorectal Liver Metastases

time, whereas the entire group showed a progressive

reduction until it approached zero. The excess of hazard in

the entire group decreased until a 99 % level of confidence

in the general population at 6.48 years after hepatic

resection, indicating that after this time point, a patient

alive without tumor recurrence could be considered cured

with 99 % certainty. Considering overall survival, instead

of DFS, as a survival measure, the death rate after surgery

for CLM equals the death rate of the general population in

41.8 % of cases (95 % confidence interval 35.4–48.2) with

a median survival of remaining patients of 2.5 years.

Cure fractions, time to cure, and median survival of

uncured patients stratified by clinical and tumoral features

are detailed in Table 2. The significant determinants of

cure probabilities (p \ 0.05) were age, primary tumor node

status, metastases presentation, DFS from colon surgery,

largest tumor size, number of metastases, the adoption of

adjuvant chemotherapy after hepatectomy, and the occur-

rence of postoperative complications. These variables were

entered in the subsequent multivariate cure model; results

are reported in Table 3. After backward removal of non-

significant variables, the final model included primary node

TABLE 2 Univariate cure-fraction calculation stratified by baseline characteristics

Variable Cure fraction,

% (95 % CI)

Time to cure,

year (95 % CI)

Median survival of uncured

cases, year (95 % CI)

Whole study population 20.0 (16.5–23.5) 6.48 (5.39–7.80) 0.91 (0.83–0.99)

Age (p = 0.037)

\60 year 16.6 (12.2–21.0) 6.56 (5.45–7.89) 0.87 (0.78–0.95)

C60 year 22.0 (17.9–26.2) 6.40 (5.32–7.69) 0.92 (0.85–1.02)

Gender (p = 0.406)

Male 20.9 (16.7–25.0) 6.45 (5.36–7.76) 0.92 (0.84–1.01)

Female 18.7 (14.1–23.3) 6.52 (5.41–7.84) 0.89 (0.81–0.98)

Primary tumor T stage (p = 0.116)

T1–T2 27.0 (17.3–36.6) 6.29 (5.21–7.59) 0.99 (0.87–1.12)

T3–T4 19.2 (15.7–22.8) 6.50 (5.41–7.84) 0.90 (0.83–0.98)

Primary tumor nodal status (p = 0.001)

Negative 29.8 (23.5–36.0) 6.21 (5.16–7.48) 1.02 (0.93–1.13)

Positive 16.1 (12.5–19.6) 6.61 (5.49–7.97) 0.84 (0.79–0.94)

Presentation of metastases (p = 0.001)

Metachronous 27.5 (22.4–32.7) 6.28 (5.21–7.57) 1.01 (0.92–1.11)

Synchronous 13.6 (10.1–17.2) 6.70 (5.54–8.09) 0.83 (0.76–0.91)

DFS from colon surgery (p = 0.002)

\24 month 28.8 (21.9–35.7) 6.19 (5.15–7.44) 1.01 (0.91–1.12)

C24 months 17.6 (14.1–21.1) 6.51 (5.41–7.82) 0.88 (0.81–0.96)

Largest tumor size (p = 0.001)

\3 cm 25.8 (21.1–30.6) 6.40 (5.31–7.71) 0.99 (0.90–1.09)

3–5 cm 18.6 (15.2–22.0) 6.61 (5.48–7.97) 0.91 (0.83–0.99)

C5 cm 11.4 (7.3–15.5) 6.82 (5.65–8.28) 0.79 (0.72–0.88)

No. of liver metastases (p = 0.001)

1 30.6 (25.4–35.8) 6.31 (5.23–7.61) 1.07 (0.97–1.18)

2 or 3 18.7 (15.4–21.9) 6.65 (5.51–8.03) 0.94 (0.86–1.02)

C4 6.8 (3.8–9.7) 7.11 (5.86–8.62) 0.73 (0.66–0.80)

Adjuvant chemotherapy (p = 0.014)

Yes 16.4 (12.7–20.0) 6.56 (5.42–7.93) 0.88 (0.80–0.96)

No 25.5 (18.4–32.5) 6.29 (5.21–7.60) 0.98 (0.88–1.11)

Postoperative complications (p = 0.001)

No 23.2 (19.0–23.4) 6.36 (5.30–7.64) 0.95 (0.87–1.04)

Yes 13.9 (9.7–18.1) 6.65 (5.53–8.01) 0.83 (0.75–0.91)

Time to cure was calculated assuming a 99 % level of confidence (a = 0.01)

CI confidence interval

N. A. Cucchetti et al.

status (p = 0.001), metastases presentation (p = 0.047),

largest tumor size (p = 0.001), and number of metastases

(p = 0.001). The constant term was 40.9 %, representing

the estimated cure fraction for patients with an N0 primary

tumor and the metachronous presentation of a single lesion

not larger than 3 cm. In such a favorable case (present in

3.8 % of the study population), the time to cure was

5.84 years. Covariate effects are expressed in differences

in the cure fraction. For example, in similar patients (N0,

metachronous presentation of a single lesion) but with a

lesion 3–5 cm in size, the cure fraction was 4.1 lower than

that of the constant term (resulting in 36.8 %); if the size of

the tumor was larger than 5 cm, the cure fraction was 8.2

lower than that of the constant term (resulting in 32.7 %).

In the presence of all unfavorable prognostic factors (3.7 %

of patients), the cure fraction dropped to 1.5 % and the

time to cure increased to 7.31 years. Corresponding figures

considering overall survival, instead of DFS, as a survival

measure are reported in the Supplementary Materials and

ranged from 78.9 % in the best-case scenario to 8.2 % in

the worst-case scenario.

DISCUSSION

In oncology, cure models represent an interesting

method of studying patient outcome because they make it

possible to know if and when patients alive without tumor

recurrence can be considered cured from their cancer. The

clinical utility of the cure fraction lies in the possibility of

informing patients regarding the probability of success of a

specific treatment, providing a single measure that can be

easily understood, rather than survival rates.6,7 Considering

the epidemiologic burden of colorectal cancer and the

relatively high probability of developing liver metastases,

it can be quickly understood how this measure can be

useful in daily clinical practice.1,2,6,7 The majority of

cancer survivors are concerned about their cancer recurring

and have learned to live with the uncertainty of being able

to return to a normal life. The possibility of giving a precise

estimation of the probabilities of being alive without tumor

recurrence can help clinicians to give a more precise

answer to patients requesting this information.

In the present study, the probability of being cured,

without tumor recurrence, from CLM by hepatectomy was

estimated to be 20 % in the entire study population and was

modified by tumoral features which are known to affect

prognosis.3–5 In the best-case scenario, which unfortunately

can be observed only in a small proportion of cases

(3.8 %), the probability of being cured by resection

reached 40.9 %; the presence of one or more adverse

prognostic factors progressively lowered the probability of

being cured, reaching a minimum of 1.5 % (Table 3).

These data can well fit the need for individual patients to

know what the chances are of returning to a normal life,

namely to have the same cancer-free life expectancy as the

general population, on the basis of the characteristics of

their disease. It should be noted that in the worst-case

scenario, a cure probability of 1.5 % can be considered a

TABLE 3 Multivariate cure

model and final cure prediction

in relation to clinical and

tumoral features

The final cure model was

obtained after the backward

removal of nonsignificant

variables resulting from the

initial model (order of

exclusion: age, disease-free

survival, postoperative

complications, adjuvant

chemotherapy). The constant

term of the final model

represents the estimated cure

fraction for patients with N0

primary tumor node status and

metachronous presentation of a

single lesion not larger than

3 cm. The covariate effects for

the cure fraction are assumed to

be additional to the constant

term

CI confidence interval

Characteristic Status Cure proportion, % (95 % CI) p

Initial model

Constant 41.9 (32.4 to 51.4) 0.001

Age C60 year ?0.2 (-1.0 to ?1.4) 0.641

Primary tumor nodal status N? -6.5 (-12.9 to -1.1) 0.046

Presentation of metastases Synchronous -5.7 (-12.3 to ?0.8) 0.086

Disease-free interval from colon surgery \24 months -2.4 (-10.4 to ?5.6) 0.555

No. of liver metastases 2 or 3 -7.2 (-10.3 to -4.0) 0.001

C4 -14.4 (-20.6 to -8.0) –

Largest tumor size 3–5 cm -3.5 (-5.8 to -1.2) 0.003

C5 cm -7.0 (-11.6 to -2.4) –

Adjuvant chemotherapy No -4.4 (-10.6 to ?1.9) 0.172

Postoperative complications Yes -1.4 (-3.9 to ?1.2) 0.289

Final model

Constant 40.9 (33.7 to 48.1) 0.001

Primary tumor node status N? -8.5 (-14.8 to -2.4) 0.007

Presentation of metastases Synchronous -5.1 (-10.5 to -0.1) 0.047

No. of liver metastases 2 or 3 -8.8 (-11.8 to -5.8) 0.001

C4 -17.6 (-23.6 to -11.6) –

Largest tumor size 3–5 cm -4.1 (-6.5 to -1.8) 0.001

C5 cm -8.2 (-13.0 to -3.6) –

Statistical Cure for Colorectal Liver Metastases

poor outcome; nevertheless, even in this infrequent unfa-

vorable scenario (3.7 %), surgery probably remains the

only chance for radical treatment. In addition, the consid-

erable difference observed between DFS and overall

survival (Fig. 1) and the observed equipoise of the death

rate after CLM surgery to the death rate of the general

population of 41.8 %, support the possibility of long-term

survival even in the case of tumor recurrence after hepa-

tectomy by both repeated hepatic surgery and the adoption

of efficient chemotherapeutic strategies, thus providing

further hope to patients in search of a return to a normal

life.15,16

The present results can give a clearer idea of the like-

lihood of the success of the surgery to oncologists and can

better define the appropriate therapeutic strategy. To our

knowledge, few reports have analyzed the cure rate of liver

cancers, and they do not distinguish among specific histo-

logic types or type of therapy adopted. From population-

based estimations, the cure fraction for liver cancer was

found to be about 10 %.17,18 The difference between pre-

vious results and the results presented here is probably the

consequence of the proportion of patients with other liver

cancers (such as hepatocellular carcinoma) that have an

unfavorable outcome with respect to resectable CLMs in

the population-based estimations. The present report is thus

the first to give an estimation of the cure fraction of CLM.

On the basis of a large multicenter experience involving

three high-volume hospitals with low postoperative mor-

tality, an up-to-date chemotherapy approach, and good

long-term survival, the present results seem reliable.

Together with the probability of being cured, patients

may also want to know when they will be considered cured.

In the past, some attempts have been made to answer this

question. Surgical series have suggested that patients who

survive 10 years after surgery appear to be cured of their

disease on the basis of a disease-specific probability of

death of approximately 1 %, but these results only indi-

rectly support the possibility of achieving a cure.4,5 To be

fair, the cure fraction should be determined by comparing

the DFS of treated people against a matched control group

who have never had the disease.11 Considering disease-

specific survival as an outcome measure, as in previous

reports, patients who are alive with recurrence are con-

sidered to be cured just the same. This observation is also

pertinent if overall survival is considered as a temporal end

point. From a formal statistical point of view, the term of

comparison can be survival of the general population that is

inevitably affected by age-specific causes of death

(including, for example, other cancer types). However, we

chose to use DFS to provide unequivocal information for

patients. In fact, it can be difficult for patients to understand

that they can be considered cured, even with tumor

recurrence after surgery, only because there are other

persons in the general population that are currently dying

from other diseases. In the present study, these aspects

were considered and accomplished by considering DFS as

an end point and by analyzing relative survival using the

cure model. Results from the present analysis suggest that

if there is a cure, it will be achieved earlier than 10 years

after hepatic surgery. In fact, the excess hazard rate

increases quickly in the first year after surgery (Fig. 2b) as

a consequence of both postoperative and early recurrence,

which leads to a very unfavorable outcome, but after the

first year, it starts to decline until it approaches zero from

the sixth year after surgery onward.19 More precisely, in

the entire study group, a patient alive without tumor

recurrence will be considered cured, with 99 % certainty,

between 5.84 and 7.31 years, depending on tumoral fea-

tures. That is, the more advanced the tumor stage, the

longer the time needed to achieve the cure. This is an

important aspect to consider when communicating with

patients who have been free of recurrence for a long time

and who would like to know when they will be able to

breathe a sigh of relief. Regarding this temporal aspect, it

can be pointed out that a not negligible proportion of

patients, in the present analysis, would be censored before

reaching this temporal end point. This aspect has already

been evaluated in specific modeling studies, suggesting that

a certain threshold is required before the statistical cure

rate can be estimated.20 For liver cancer, this threshold was

estimated to be 3.7 years. As can be noted from Fig. 1, at

this temporal end point, confidence intervals of curves are

quite narrow, supporting the fact that most patient survivals

were still not right censored. Thus, assumptions at the basis

of the cure model application can be reasonably considered

to be fulfilled.

Some other results from the present study deserve more

discussion. In particular, no benefit from adjuvant chemo-

therapy after hepatectomy was observed; instead, it seems

that adjuvant chemotherapy provided a lower cure fraction

(Table 2). This paradoxical aspect is explained by the more

advanced tumor sizes at the time of the colonic and hepatic

surgery of the patients who received adjuvant chemother-

apy (Supplementary Materials). This observation is

reinforced by the multivariate cure model, which excluded

adjuvant chemotherapy as an independent prognostic factor

in reaching a cure. The pivotal role of adjuvant chemo-

therapy is not under examination in the present analysis,

especially in view of the rapid progress observed in the

field over the past decade; however, to correctly estimate

its benefit in achieving a statistical cure, dedicated, well-

matched studies are needed.3,16

In conclusion, the present study provides a statistical

measure useful in the assessment of the success that can be

expected from the hepatic resection of CLM. The cure

fraction can be of interest for clinicians and, in particular,

N. A. Cucchetti et al.

for patients who would like to be fully informed regarding

what awaits them after hepatic surgery.

ACKNOWLEDGMENT This work is dedicated to the memory of

Prof. Lorenzo Capussotti.

DISCLOSURE The authors declare no conflict of interest.

REFERENCES

1. Jemal A, Bray F, Center MM, et al. Global cancer statistics. CA

Cancer J Clin. 2011;61:69–90.

2. US Cancer Statistics Working Group. United States cancer sta-

tistics: 1999–2010 incidence and mortality Web-based report.

Atlanta, GA: US Department of Health and Human Services,

Centers for Disease Control and Prevention and National Cancer

Institute; 2013. Available at: http://www.cdc.gov/uscs. Accessed

1 Feb 2014.

3. Garden OJ, Rees M, Poston GJ, et al. Guidelines for resection of

colorectal cancer liver metastases. Gut. 2006;55(Suppl 3):iii1–8.

4. Tomlinson JS, Jarnagin WR, DeMatteo RP, et al. Actual 10-year

survival after resection of colorectal liver metastases defines cure.

J Clin Oncol. 2007;25:4575–80.

5. Pulitano C, Castillo F, Aldrighetti L, et al. What defines ‘‘cure’’

after liver resection for colorectal metastases? Results after

10 years of follow-up. HPB (Oxford). 2010;12:244–9.

6. Statistical Methodology and Applications Branch, Data Modeling

Branch, National Cancer Institute. Cansurv, version 1.1. Available at:

http://surveillance.cancer.gov/cansurv/index.html. Accessed 1 Feb

2014.

7. Othus M, Barlogie B, Leblanc ML, et al. Cure models as a useful

statistical tool for analyzing survival. Clin Cancer Res. 2012;18:

3731–6.

8. Lambert PC, Thompson JR, Weston CL, et al. Estimating and

modeling the cure fraction in population-based cancer survival

analysis. Biostatistics. 2007;8:576–94.

9. Terminology Committee of the IHPBA. The Brisbane 2000 ter-

minology of liver anatomy and resections. Abingdon: Taylor &

Francis; 2000. p. 333–9.

10. Smoll NR, Schaller K, Gautschi OP. The cure fraction of glio-

blastoma multiforme. Neuroepidemiology. 2012;39:63–9.

11. Fuller AF, Griffiths CT. Gynecologic oncology. Amsterdam:

Wolters Kluwer; 1983.

12. Lambert PC. Modeling of the cure fraction in survival studies.

Stata J. 2007;3:1–25.

13. Tsodikov AD, Ibrahim JG, Yakovlev AY. Estimating cure rates

from survival data: an alternative to two-component mixture

models. J Am Stat Assoc. 2003;98:1063–78.

14. Italian population life tables. Available at: http://demo.istat.it/.

Accessed 1 Dec 2013.

15. Adam R, Bismuth H, Castaing D, et al. Repeat hepatectomy for

colorectal liver metastases. Ann Surg. 1997;225:51–60.

16. Adam R, De Gramont A, Figueras J, et al. The oncosurgery

approach to managing liver metastases from colorectal cancer: a

multidisciplinary international consensus. Oncologist. 2012;

17:1225–39.

17. Cvancarova M, Aagnes B, Fossa SD, et al. Proportion cured

models applied to 23 cancer sites in Norway. Int J Cancer.

2013;132:1700–10.

18. Dal Maso L, Guzzinati S, Buzzoni C, et al. Long-term survival,

prevalence, and cure of cancer: a population-based estimation for

818,902 Italian patients and 26 cancer types. Ann Oncol. 2014;

25:2251-60.

19. Vigano L, Capussotti L, Lapointe R, et al. Early recurrence after

liver resection for colorectal metastases: risk factors, prognosis,

and treatment. A LiverMetSurvey-based study of 6,025 patients.

Ann Surg Oncol. 2014;21(4):1276–1286.

20. Tai P, Yu E, Cserni G, et al. Minimum follow-up time required

for the estimation of statistical cure of cancer patients: verifica-

tion using data from 42 cancer sites in the SEER database. BMC

Cancer. 2005;5:48.

Statistical Cure for Colorectal Liver Metastases