1331
Conditional Survival in Head and Neck Squamous
Cell Carcinoma
Results From the SEER Dataset 1973 —1998
Clifton D. Fuller, MD1
Samuel J. Wang, MD2
2
Charles R. Thomas, Jr., MD
Henry T. Hoffman, MD3
Randal S. Weber, MD4
David I. Rosenthal, MD5
1
Department of Radiation Oncology and Gradu-
ate Division of Radiological Sciences, University
of Texas Health Science Center at San Antonio,
San Antonio, Texas.
2
Department of Radiation Medicine, Oregon
Health and Science University Cancer Institute,
Portland, Oregon.
3
Department of Otolaryngology

Head and
Neck Surgery, University of Iowa, Iowa City,
Iowa.
4
Department of Head and Neck Surgery, Univer-
sity of Texas M. D. Anderson Cancer Center,
Houston, Texas.
5
Department of Radiation Oncology, University of
Texas M. D. Anderson Cancer Center, Houston,
Texas.
Portions of this data were presented at the
American Society of Clinical Oncology Annual
Meeting, June 2
6, 2006, Atlanta, GA.
Address for reprints: David I. Rosenthal, MD,
Department of Radiation Oncology, University of
Texas M. D. Anderson Cancer Center, 1515 Hol-
combe Blvd., Unit 97, Houston TX 77030; Fax:
(713) 563-2331; E-mail: dirosenthal@mdanderson.
org
Received November 13, 2006; revision received
December 13, 2006; accepted December 18,
2006.
ª 2007 American Cancer Society
DOI 10.1002/cncr.22563
Published online 26 February 2007 in Wiley InterScience (www.interscience.wiley.com).
BACKGROUND. Survival statistics for patients with head and neck squamous cell
carcinomas (HNSCC) are commonly calculated from the time of diagnosis. The
less commonly employed conditional survival (CS) analyzes survival for patients
who have survived a period of time after diagnosis. Useful prognostic information
for cancer survivors is provided by CS analysis. Estimated baseline CS parameters
for HNSCC were sought using large-scale cancer registry data.
METHODS. HNSCC cases identified from the Surveillance, Epidemiology, and End
Results (SEER) Program were accessed to identify those diagnosed between 1973
and 1998. Five-year observed, relative, and cumulative CS calculations were per-
formed, with secondary stratification by site, extent of disease, and age.
RESULTS. The overall 5-year observed survival for all sites increased from 47.8%
for 76,181 included patients from the time of diagnosis to 64.4% for those 43,985
patients alive at 3 years, and thereafter plateaus. The greatest increase in CS was
for oropharyngeal cancers, which more than doubled over the first decade of sur-
veillance (26.5%–60%). Distant disease showed a 10-year increase in CS (17.4%–
60.4%), whereas localized disease CS was essentially static, ranging from 66.1% to
68.5%; for those over 65 at diagnosis it ranged from 39.9–52.9%, whereas patients
<65 years at diagnosis ranged from 53.8–73.5%.
CONCLUSIONS. Benchmark CS estimates for domestic HNSCC cohorts were devel-
oped from the SEER database. CS is a useful tool to assist clinicians in predicting
the probability of demise from HNSCC for patients surviving 1 or more year after
diagnosis. Cancer 2007;109:1331–43.
2007 American Cancer Society.
KEYWORDS: conditional survival, head and neck cancer, squamous cell.
S urvival statistics from cancer are typically presented as the prob-
ability of demise within a fixed period from diagnosis. However,
the probability of death after initial management is not static. The
concept of ‘conditional probability’ identifies that hazard rates may
change in a measurable way over time. Conditional survival (CS)
applies this concept to determine the probability that a patient who
has survived for a designated period will be alive at another fixed
interval. CS denotes the changing likelihood of demise during a
span of intervals after diagnosis and can be represented graphically
as the slope of a survival curve at a point in time after diagnosis.
Thus, CS analysis offers meaningful prognostic information to
patients who have survived initial cancer management.1
Several previously reported cancer conditional survival data series
reveal distinct patterns of CS that vary substantially among diagnoses.2,3
These studies have explored CS in breast,2,4–6 lung,7,8 brain,9–11 pros-
tate,2,3,12 and gastrointestinal malignancies.13–16 To our knowledge no
1332 CANCER April 1, 2007 / Volume 109 / Number 7
formal literature has explored the CS patterns of head
and neck squamous cell carcinomas (HNSCC). We
determined CS for this patient population using data
from the Surveillance, Epidemiology, and End Results
(SEER) Program of the National Cancer Institute (NCI)
dataset.17
MATERIALS AND METHODS
Data Selection Criteria
The SEER Program17 is a population-based cancer
database that geographically encompasses more than
a quarter of the US population within its catchment
areas. SEER program registries collect data on patient
demographics, cancer type and site, stage, first
course of treatment, and vital status at last contact,
among other variables.
Using the publicly available download of the
most recent release of the SEER17 as well as SEERStat
6.2.3,18 we analyzed survival data from all patients
diagnosed with HNSCC between 1973 and 1998.
Cases diagnosed until 1998 were included to ensure
at least 5 years of follow-up data (through December,
2003). Cases were not included in the analysis if they
were only identified by death certificate or autopsy.
Cases with multiple primaries were excluded.
To characterize the relation between extent of
disease and CS, SEER Historic Staging was used,
excluding in situ cases. SEER Historic Staging is
designed to allow comparison between different eras,
despite alterations in collaborative staging measures
(such as AJCC staging). Tumors are coded at diagno-
sis as in situ, localized, regional, distant, or unstaged.
The SEER Glossary of Statistical Terms defines loca-
lized cancer as ‘‘limited to the organ in which it
began, without evidence of spread,’’ regional disease
as ‘‘beyond the original (primary) site to nearby
lymph nodes or organs and tissues,’’ distant disease
as ‘‘[having] spread from the primary site to distant
organs or distant lymph nodes,’’ and unstaged as
‘‘cancer for which there is not enough information to
indicate a stage.’’ Specific definitions of extent of dis-
ease/anatomical boundaries by Head and Neck sub-
site for SEER Historic Staging may be found online
as part of the SEER Summary Staging Manual 2000.19
It is important to note that comparisons between
SEER Historic Stage and the more clinically oriented
AJCC staging must be addressed on a site-by-site ba-
sis; for instance, distant disease does not necessarily
imply metastatic disease as defined by AJCC stage
designation.20
To examine the effect of subsite on CS, all
patients were categorized into the following subsite
designations, as defined by the SEER data dictionary:
Lip, Tongue, Salivary Gland, Floor of Mouth, Gum
and Other Mouth, Nasopharynx, Tonsil, Oropharynx,
Hypopharynx, Other Oral Cavity and Pharynx, Nose,
Nasal Cavity and Middle Ear, and Larynx (ICD-O-3
Site designates C000-C009, C019-C029,C079-C089,
C040-C049, C030-C039, C050-C059, C060-C069,
C110-C119, C090-C099, C100-C109,C129, C130-C139,
C300-C301, C310-C319, and C320-C329, respectively).
Historic SEER staging was unavailable for Nose,
Nasal Cavity, and Middle Ear cases between 1973–
1982, and rates were not included for small groups
where the standard error (SE) was greater than 10%.
Microscopically confirmed squamous cell carcinoma
by ICD-O-3 morphology codes (8052, 8070-8079)
were required for inclusion.
Survival Analysis
The SEER dataset affords calculation of survival
probabilities of both observed survival and relative
survival. Observed survival is raw survival, and does
not differentiate between causes of death. Observed
survival represents the probability of surviving all
causes of death for a given time interval. Expected
survival is the estimated percentage surviving in a
comparable set of individuals without cancer. Rela-
tive survival is calculated as a ratio between the per-
centage of observed survival in HNSCCC patients
and a similar population without any cancer diagno-
sis. Relative survival is designed to account for com-
peting causes of death and adjusts for the general
survival rate of the US population by race, sex, age,
and temporal span. The SEER data calculates relative
survival with expected survival rate tables delivered
with the SEER Public-Use databases (specifically, the
database ‘‘U.S. 1970, 1980, 1990 (White, Black,
Other’’)).
Definition of Conditional Survival
CS is predicated on the conditional probability calcu-
lation, commonly utilized in biostatistics of survival
modeling. CS can be readily calculated from product
limit (Kaplan-Meier) survival data. The mathematical
definition of CS can be expressed as follows:
Let S(t) be product limit survival at time t. Con-
ditional survival, CS(y|x), is the probability of surviv-
ing y years, given that the patient cohort has already
survived x years. Conditional survival thus becomes:
Sðx þ yÞ
CSðy j xÞ ¼ :
SðxÞ
Consequently, to compute the 5-year CS for a patient
that has already survived 1 year, the ratio of the
product limit survival at 5 þ 1 years is divided by the

FIGURE 1. Annual 20-year cumulative observed, relative, and expected survival data for all head and neck squamous cell carcinoma (HNSCC) patients. Error
bars were not included, as standard error was less than 0.4% for each data point.
Kaplan-Meier survival at 1 year. With longer follow-
up after diagnosis, changes in cohort CS serves as an
indicator of the variation in hazard rate over time.
CS probabilities were calculated using the
actuarial life table method included in the SEER*Stat
software.18 The calculations were then confirmed by
product-limit calculation using JMP 6 (SAS Institute,
Cary, NC) and graphics were generated using SPSS
11 (SPSS, Chicago, IL) and StatView 5.0.1 (SAS). We
computed observed 5-year CS and stratified results
by: age (>65 vs <65), historic SEER stage A, and spe-
cified subsites. Historic SEER staging was unavailable
for Nose, Nasal Cavity, and Middle Ear cases between
1973–1982, and rates were not included for small
groups where the SE was greater than 10%.
The SEER dataset allows calculation of both
observed survival and relative conditional survival.
Observed survival, also known as crude or raw sur-
vival, is available in SEER, as patients are character-
ized as either alive or dead at a certain timepoint
after diagnosis. Relative survival is a calculated esti-
mate of the probability of dying from the primary
cancer alone at a point in time. Relative survival is
calculated in SEER as a ratio, using observed survival
Conditional Survival in HNSCC/Fuller et al. 1333
and expected survival. For clarity, we will refer to CS
calculated using observed survival as ‘‘observed 5-
year conditional survival’’ (OCS), defined as the
probability of surviving all causes of mortality for 5
years from a specified timepoint. ‘‘Relative 5-year
conditional survival’’ (or RCS) denotes the calculated
probability of not dying of cancer for 5 years from a
given point after survival. Expected survival (the esti-
mated survival probability of the matched non-can-
cer population) was recorded from SEERStat and
utilized to calculate ‘‘expected conditional survival’’
(ECS), which reflects the 5-year expected survival of
a demographically analogous noncancer population.
Expected survival (and thus ECS) may be utilized to
estimate what survival for patients would be had
they not been diagnosed with neoplastic disease.
RESULTS
A total of 76,181 HNSCC patients were identified in
the SEER dataset as meeting inclusion criteria. The
annual observed, relative, and expected survival for
all 76,181 patients are shown in Figure 1. Relative
survival tends to more closely approximate observed
1334 CANCER April 1, 2007 / Volume 109 / Number 7
FIGURE 2. Graphic plot of 5-year probability of survival from all causes (observed conditional survival, or OCS), 5-year estimated probability of survival from
cancer alone (relative conditional survival, or RCS). Expected conditional survival (ECS), the estimated noncancer survival for an analogous demographic cohort
is shown.
survival for the initial 2 years after diagnosis and
then more closely approximates the expected survival
in the second decade after diagnosis (Fig. 1). Five-
year CS follows a similar pattern, with RCS closely
approximating ECS in the first 2 years after diagnosis
and then closely approximating 5-year expected sur-
vival in the later years after diagnosis (Fig. 2). Error
bars were not included, as the SE was less than 0.4%
for each data point (Table 1).
Subsite CS probabilities are recorded in Table 1.
Figures 3 and 4 indicate OCS and RCS, respectively,
grouped by extent of disease as defined by SEER his-
toric staging in both graphical and tabular formats.
Figure 5 similarly displays CS parameters for those
age 65 or greater as compared with those <65 years
at diagnosis. Tables 2 and 3 show the number at risk
annually, as well as annual CS calculations and SE
for each extent of disease and age stratifications.
DISCUSSION
The dynamic risk calculation of CS augments re-
ported static survival statistics, such as 5-year overall
survival from diagnosis. Conceptually, CS provides a
useful mechanism to prevent reification of static
time-of-diagnosis risk assessment as a fixed predic-
tive moment; that is, it forces the physician to
remember that risk changes as a patient survives
more time from diagnosis. More practically, CS is a
boon both to patients and clinicians.1,2 Calculated
CS may be used as a ‘‘look-up table’’ to stratify
patients over time, allowing patients realize that, as
time progresses, their chance of demise changes.
Clinicians may, for example, use CS data in an effort
to institute data-driven optimization of posttherapy
surveillance. For instance, many physicians minimize
follow-up surveillance after 3–5 years, often with lit-
tle justification based on survival data. Additionally,
CS has utility as another mechanism to detect differ-
entials in population survival patterns.2,9 In this se-
ries, the differential survival between lip neoplasms
and gum and other mouth tumors at 5 years post-
diagnosis might lead to increased comparative vigi-
lance of subsite gum and other mouth patients. This
assumes that surveillance would detect a recurrence
 Conditional Survival in HNSCC/Fuller et al. 1335

TABLE 1
Number of Patients at Risk and Derived CS Parameters Stratified by SEER-designated Subsites

Years survived
SEER subsite from diagnosis No. at risk OCS  SE(%) RCS  SE(%) ECS(%)

All 0 76,181 47.8  0.2 56.3  0.2 84.9

1 61,071 54.7  0.2 64.3  0.2 85
2 50,065 61.4  0.2 72.7  0.2 84.4
3 43,985 64.4  0.2 76.8  0.2 83.8
4 39,632 65.5  0.3 78.7  0.3 83.2
5 36,215 65.8  0.3 79.6  0.3 82.6
6 31,646 65.9  0.3 80.4  0.3 82
7 27,658 65.8  0.3 80.8  0.3 81.4
8 24,074 65.6  0.3 81.1  0.3 80.9
9 20,845 65.6  0.4 81.8  0.4 80.2
10 17,994 64.9  0.4 81.6  0.4 79.6
Lip 0 7876 75.2  0.5 94.2  0.6 79.8
1 7503 74  0.5 93.3  0.6 79.4
2 7056 73.9  0.5 93.5  0.7 79
3 6657 73.2  0.6 93.2  0.7 78.6
4 6272 72.5  0.6 92.9  0.7 78.1
5 5883 72.2  0.6 92.8  0.8 77.9
6 5325 71.9  0.6 93.1  0.8 77.3
7 4810 71.7  0.7 93.1  0.9 77
8 4330 71.7  0.7 93.6  1 76.6
9 3884 71.5  0.8 93.8  1 76.2
10 3474 70.5  0.8 93.1  1.1 75.7
Tongue 0 11,621 41.4  0.5 47.8  0.5 86.4
1 8797 50.5  0.5 58.1  0.6 87
2 6695 61.2  0.6 70.5  0.7 86.7
3 5796 65.3  0.6 75.7  0.8 86.3
4 5218 66.6  0.7 77.6  0.8 85.8
5 4770 67.3  0.8 78.9  0.9 85.3
6 4128 67  0.8 79.1  1 84.6
7 3525 66.4  0.9 79  1.1 84.1
8 3010 66.1  1 79.1  1.2 83.5
9 2544 66.6  1.1 80.2  1.3 83
10 2153 66.5  1.1 80.8  1.4 82.3
Salivary Gland 0 848 33.5  1.6 45.3  2.2 74.1
1 614 41.7  2 55  2.6 75.9
2 457 49.9  2.4 65.8  3.1 75.9
3 368 53.5  2.7 70.1  3.6 76.3
4 311 60  3 78.8  3.9 75.9
5 284 59.9  3.2 78.8  4.2 76
6 240 61.2  3.4 80.9  4.6 75.7
7 193 63.9  3.7 84.4  4.9 75.7
8 158 64.8  4.2 84  5.4 77.2
9 141 65.6  4.4 84.3  5.7 76.8
10 120 68.6  4.6 87.9  6 76.4
Floor of Mouth 0 5999 46.3  0.6 53  0.7 87.3
1 4890 51.2  0.7 58.5  0.8 87.5
2 3952 57.8  0.8 66.3  0.9 87.1
3 3447 60.2  0.9 69.5  1 86.7
4 3061 61.2  0.9 71  1.1 86.2
5 2766 60.4  1 70.5  1.2 85.6
6 2386 60.4  1.1 71.1  1.3 85
7 2096 60.6  1.1 71.6  1.3 84.6
8 1815 59  1.2 70.3  1.5 84
9 1570 58.3  1.3 70.1  1.6 83.2
10 1345 57.6  1.4 69.7  1.8 82.7

(continued)
1336 CANCER April 1, 2007 / Volume 109 / Number 7

TABLE 1
(continued)

Years survived
SEER subsite from diagnosis No. at risk OCS  SE(%) RCS  SE(%) ECS(%)

Gum and Other Mouth 0 7556 40.8  0.6 49.1  0.7 83.1
1 5804 47.4  0.7 56.7  0.8 83.7
2 4542 54  0.7 64.9  0.9 83.3
3 3872 57.6  0.8 69.6  1 82.7
4 3426 58.3  0.9 71  1.1 82.1
5 3066 58.2  1 71.4  1.2 81.5
6 2627 59  1 72.6  1.3 81.2
7 2252 59.1  1.1 73.1  1.4 80.9
8 1931 58.6  1.2 73  1.5 80.2
9 1629 58.5  1.3 73.5  1.7 79.5
10 1366 57.9  1.4 73.3  1.8 79
Nasopharynx 0 2267 40.8  1 45  1.1 90.7
1 1738 49.4  1.2 53.9  1.3 91.7
2 1396 56.6  1.3 61.6  1.5 91.9
3 1161 63  1.5 68.8  1.6 91.6
4 1028 66.7  1.6 72.8  1.7 91.6
5 914 70.4  1.7 77  1.8 91.4
6 796 71.8  1.8 78.8  2 91.2
7 675 72.2  2 79.6  2.2 90.7
8 575 73.5  2.1 81.3  2.3 90.4
9 498 75.7  2.2 84.1  2.4 90
10 421 77.2  2.3 86.2  2.5 89.6
Tonsil 0 5864 40.3  0.6 45.7  0.7 88.2
1 4440 48.1  0.8 54.3  0.9 88.5
2 3458 56.6  0.9 64.1  1 88.4
3 2929 61.1  0.9 69.3  1.1 88.2
4 2612 61.3  1 69.8  1.2 87.9
5 2350 61.7  1.1 70.5  1.3 87.5
6 1957 62  1.2 71.4  1.4 86.9
7 1645 62.6  1.4 72.5  1.6 86.5
8 1385 62.9  1.5 73.3  1.7 85.8
9 1127 63.2  1.7 74.1  2 85.3
10 923 61.6  1.9 73.1  2.2 84.4
Oropharynx 0 1507 26.5  1.1 30.6  1.3 86.6
1 976 36.6  1.5 42  1.8 87.2
2 698 46.1  1.9 52.9  2.2 87.1
3 565 50.4  2.2 58.2  2.5 86.6
4 456 54.8  2.5 63.3  2.9 86.6
5 398 56.8  2.8 65.9  3.2 86.2
6 337 58.5  3 68.5  3.5 85.4
7 282 59.1  3.3 69.5  3.9 85
8 229 59.8  3.6 70.5  4.3 84.9
9 185 59.9  4.1 71.1  4.8 84.2
10 150 60  4.4 71.8  5.3 83.5
Hypopharynx 0 5301 23.4  0.6 27.4  0.7 85.3
1 3372 32  0.8 37.3  0.9 85.7
2 2243 42.2  1.1 49.5  1.2 85.3
3 1732 47.9  1.2 56.5  1.5 84.7
4 1427 50.7  1.4 60.5  1.7 83.8
5 1233 50.6  1.5 61.1  1.9 82.8
6 1024 50.3  1.7 61.4  2.1 81.9
7 848 51  1.9 62.7  2.3 81.3
8 696 50  2.1 62.1  2.6 80.5
9 580 49  2.3 61.6  2.9 79.6
10 470 48.9  2.5 62  3.2 79

(continued)
 Conditional Survival in HNSCC/Fuller et al. 1337

TABLE 1
(continued)

Years survived
SEER subsite from diagnosis No. at risk OCS  SE(%) RCS  SE(%) ECS(%)

Other Oral Cavity and Pharynx 0 1573 22.1  1 25.8  1.2 85.5
1 923 34.2  1.6 39.7  1.8 86.1
2 615 44.6  2 52.2  2.4 85.3
3 485 49.9  2.3 58.6  2.7 85
4 409 52.9  2.6 62.4  3.1 84.8
5 347 56.1  2.9 67  3.4 83.7
6 302 53  3.1 63.4  3.8 83.5
7 242 57.1  3.4 69  4.1 82.8
8 205 56.5  3.8 68.1  4.6 83
9 173 58.2  4.2 70.2  5 82.9
10 148 58.9  4.5 71.5  5.5 82.3
Nose, Nasal Cavity and Middle Ear 0 1947 42.7  1.1 51.2  1.4 83.2
1 1458 53.1  1.3 63.5  1.6 83.7
2 1142 63.2  1.4 75.7  1.7 83.5
3 999 66.4  1.5 79.9  1.9 83.1
4 909 68.3  1.6 82.5  2 82.8
5 824 70.8  1.7 86.1  2.1 82.3
6 725 71  1.8 86.8  2.2 81.8
7 640 70.9  1.9 87.5  2.4 81
8 556 69.6  2.1 86  2.6 81
9 497 69.1  2.3 86.1  2.8 80.3
10 440 66.7  2.5 83.7  3.1 79.7
Larynx 0 23,822 56.4  0.3 66.4  0.4 84.9
1 20,556 60.1  0.3 71  0.4 84.7
2 17,811 64.3  0.4 76.5  0.4 84.1
3 15,974 66.4  0.4 79.7  0.5 83.3
4 14,503 67.4  0.4 81.5  0.5 82.6
5 13,380 67.2  0.4 82  0.5 82
6 11,799 67.3  0.5 82.7  0.6 81.4
7 10,450 66.6  0.5 82.5  0.6 80.7
8 9184 66.5  0.5 82.9  0.7 80.2
9 8017 66.3  0.6 83.4  0.7 79.5
10 6984 65.3  0.6 82.8  0.8 78.9

SEER indicates Surveillance, Epidemiology, and End Results.

for which a clinically meaningful intervention would
be available.
The advantage of using both OCS and RCS allows
even subtler gradations to be observed. OCS is more
useful for patient prognostication of raw risk of
death. However, RCS values are useful to help define
the risk of dying from cancer, which is of potentially
more import for clinical surveillance. Patients with
HNSCC may die of noncancer pathology, which must
be accounted for in risk stratification and patient
counseling. This is significant for patients with
HNSCC who have a significant mortality risk from
comorbidities stemming from frequent common risk
factors of tobacco and alcohol use.21,22 Because no
other articles have, to our knowledge, presented CS
statistics for HNSCC, both OCS and RCS calculations
were included for ease of reference.
The survival data for all HNSCC patients re-
corded in SEER during the 1973–1998 period identi-
fies 5-year OCS increasing from 47.8% at diagnosis to
plateau at 64.4% within 3 years. Additionally, the RCS
increased as a function of the decrement in expected
survival for an analogous noncancer population.
Broadly speaking, after 6 years there is an increasing
probability of HNSCC patients dying of other disease
processes (such as heart disease or stroke) vs their
primary cancer. Thus, a ‘generalized’ HNSCC patient
might be expected to be ‘cured’ after 6-year disease-
free survival insofar as their risk of mortality due to
cancer is equivalent to competing-cause mortality
risk of a ‘generalized’ noncancer patient. Nonetheless,
the crude overall survival probability of HNSCC
patients, as tracked with OCS, never achieves equiva-
lence with the noncancer population, as some risk of
1338 CANCER April 1, 2007 / Volume 109 / Number 7

FIGURE 3. Observed conditional survival (OCS) stratified by extent of disease.

FIGURE 4. Relative conditional survival (RCS) stratified by extent of disease.


FIGURE 5. Calculated observed conditional survival (OCS) and relative conditional survival (RCS) values for patients >65 vs <65 years of age at diagnosis.
dying from the index cancer, or associated sequelae,
is ‘overlaid’ in addition to other possible causes of
death.21–24
The observation that mortality probability (as
expressed by OCS) plateaus at 3 years may afford the
use of 3-year clinical trial survival data as a valid end-
point, as the patients’ probability of 5-year survival is
thereafter somewhat static. However, this observation
should be validated with established clinical data-
sets25,26 in order to verify that trends observed in his-
toric epidemiological datasets, such as SEER, are
verified in more recent prospective series.
Specific subsite analysis reveals that, whereas
HNSCC of the lip, which actually evinces a negative
OCS progression over time, and nasopharyngeal
tumors show a steady increase in OCS, most other
HNSCC subsites, generally, show substantial incre-
mental increases in OCS, plateauing rather rapidly at
3 years postdiagnosis. This reflects the trend for
HNSCC cumulatively. The greatest differential in OCS
is for oropharyngeal cancers; patient 5-year survival
probability more than doubles over the first decade
of surveillance postdiagnosis (26.5%–60%). RCS and
Conditional Survival in HNSCC/Fuller et al. 1339
ECS values are included to illustrate competing non-
cancer mortality risks as a reference.
As with comparable analyses of CS in other neo-
plastic disease, patients with advanced stage disease
and poorer initial prognoses see the greatest increases
in CS as they survive for longer periods of time from
diagnosis.13 The rapid increase in CS for patients with
distant disease may have useful clinical application.
For those admittedly few patients with distant disease
who survive >3 years from diagnosis (n ¼ 1203) sur-
vival probability improves rapidly as patients survive.
However, any enthusiasm for improved CS in patients
with SEER Historic Staging distant disease must be
tempered by the realization that an exceedingly small
proportion of patients survive long enough to see
substantial CS increases. Whereas OCS at 5 years
postdiagnosis is 60.4% for patients with distant dis-
ease, only 488/6645 patients (7%) achieve 5 years of
survival. Additionally, even this is likely to be an over-
estimate due to SEER staging vagaries.
For patients with localized disease the probabil-
ity of death from all causes within 5 years is rather
stable over the first 10 years of follow-up. Sequential
1340 CANCER April 1, 2007 / Volume 109 / Number 7

TABLE 2
Number of Patients at Risk and Derived CS Parameters Stratified by SEER-designated Extent of Disease (SEER Historic Stage A)

Years survived
SEER subsite from diagnosis No. at risk OCS  SE(%) RCS  SE(%) ECS(%)

All 0 76,181 47.8  0.2 56.3  0.2 84.9

1 61,071 54.7  0.2 64.3  0.2 85
2 50,065 61.4  0.2 72.7  0.2 84.4
3 43,985 64.4  0.2 76.8  0.2 83.8
4 39,632 65.5  0.3 78.7  0.3 83.2
5 36,215 65.8  0.3 79.6  0.3 82.6
6 31,646 65.9  0.3 80.4  0.3 82
7 27,658 65.8  0.3 80.8  0.3 81.4
8 24,074 65.6  0.3 81.1  0.3 80.9
9 20,845 65.6  0.4 81.8  0.4 80.2
10 17,994 64.9  0.4 81.6  0.4 79.6
Localized 0 30,865 66.1  0.3 84.2  0.3 78.6
1 28,511 66.5  0.3 83.8  0.3 79.4
2 25,766 68.4  0.3 83.2  0.4 82.3
3 23,679 69.1  0.3 82.6  0.4 83.7
4 21,849 69.3  0.3 82.0  0.4 84.5
5 20,302 68.8  0.3 81.4  0.4 84.6
6 18,088 68.6  0.4 80.7  0.5 85.0
7 16,037 68.5  0.4 80.2  0.5 85.3
8 14,127 68.1  0.4 79.7  0.5 85.5
9 12,356 68.2  0.5 79.0  0.6 86.2
10 10,784 67.4  0.5 78.4  0.6 86.0
Regional 0 33,008 37.8  0.3 85.8  0.3 44.1
1 24,843 45.4  0.3 86.2  0.4 52.6
2 18,838 54.4  0.4 86.0  0.4 63.2
3 15,826 59.0  0.4 85.5  0.5 69.0
4 13,873 60.7  0.4 85.0  0.5 71.4
5 12,427 61.4  0.5 84.5  0.6 72.7
6 10,536 61.9  0.5 84.0  0.6 73.7
7 8972 61.5  0.6 83.4  0.7 73.7
8 7607 61.0  0.6 82.9  0.7 73.7
9 6429 60.9  0.7 82.2  0.8 74.1
10 5413 60.3  0.7 81.6  0.9 73.9
Distant 0 6645 17.4  0.5 86.0  0.5 20.2
1 3448 29.3  0.8 86.9  0.9 33.7
2 2087 43.7  1.1 86.7  1.3 50.4
3 1573 51.6  1.3 86.2  1.5 59.8
4 1323 55.6  1.4 85.7  1.7 64.8
5 1152 57.4  1.6 85.5  1.8 67.2
6 955 58.5  1.7 85.4  2 68.4
7 814 59.4  1.9 84.6  2.2 70.2
8 685 62.7  2 84.2  2.4 74.5
9 579 61.3  2.2 83.6  2.6 73.3
10 488 60.4  2.4 83.1  2.9 72.6
Unstaged 0 4998 42.0  0.7 82.7  0.8 50.7
1 3789 50.0  0.8 83.6  1 59.8
2 3020 57.1  0.9 83.4  1.1 68.5
3 2597 61.2  1 82.8  1.2 73.9
4 2302 63.1  1 82.3  1.3 76.7
5 2080 64.3  1.1 82.0  1.3 78.5
6 1830 64.7  1.2 81.5  1.4 79.4
7 1616 64.3  1.3 80.9  1.6 79.5
8 1451 64.8  1.3 80.6  1.7 80.4
9 1287 65.1  1.4 80.0  1.8 81.4
10 1130 64.1  1.5 79.4  2.0 80.7

(continued)
 Conditional Survival in HNSCC/Fuller et al. 1341

TABLE 2
(continued)

Years survived
SEER subsite from diagnosis No. at risk OCS  SE(%) RCS  SE(%) ECS(%)

Stage undesignated (Blank) 0 665 38.4  1.9 83.8  2.3 45.8

1 480 49.5  2.3 84.5  2.7 58.6
2 354 62.1  2.6 84.8  3.0 73.2
3 310 66.0  2.7 84.5  3.2 78.2
4 285 68.1  2.8 84.2  3.3 80.8
5 254 70.5  2.9 84.2  3.4 83.7
6 237 73.0  2.9 83.7  3.4 86.8
7 219 73.1  3 82.9  3.6 88.2
8 204 70.1  3.2 82.5  3.9 85.0
9 194 69.6  3.3 81.7  4.0 85.2
10 179 67.6  3.5 81.5  4.3 82.9

TABLE 3
Number of Patients at Risk and Derived CS Parameters Stratified by Age Greater or Less Than 65 Years at Diagnosis

Years survived
Age at diagnosis from diagnosis No. at risk OCS  SE (%) RCS  SE(%) ECS(%)

All 0 76,181 47.8  0.2 56.3  0.2 84.9

1 61,071 54.7  0.2 64.3  0.2 85
2 50,065 61.4  0.2 72.7  0.2 84.4
3 43,985 64.4  0.2 76.8  0.2 83.8
4 39,632 65.5  0.3 78.7  0.3 83.2
5 36,215 65.8  0.3 79.6  0.3 82.6
6 31,646 65.9  0.3 80.4  0.3 82.0
7 27,658 65.8  0.3 80.8  0.3 81.4
8 24,074 65.6  0.3 81.1  0.3 80.9
9 20,845 65.6  0.4 81.8  0.4 80.2
10 17,994 64.9  0.4 81.6  0.4 79.6
<65 y 0 44,591 53.8  0.2 57.8  0.3 93.2
1 37,279 60.3  0.3 64.9  0.3 92.8
2 31,142 68.1  0.3 73.7  0.3 92.4
3 27,824 71.7  0.3 78.0  0.3 91.9
4 25,555 73.0  0.3 79.8  0.3 91.4
5 23,836 73.3  0.3 80.7  0.3 90.8
6 21,261 73.5  0.3 81.5  0.4 90.2
7 19,042 73.0  0.3 81.6  0.4 89.5
8 16,965 72.8  0.4 82.0  0.4 88.7
9 15,013 72.7  0.4 82.6  0.4 88.0
10 13,265 72.1  0.4 82.8  0.5 87.1
>65 y 0 34,175 39.9  0.3 53.7  0.4 74.4
1 25,889 46.5  0.3 63.1  0.4 73.7
2 20,666 51.2  0.4 70.7  0.5 72.5
3 17,658 52.8  0.4 74.4  0.5 71.1
4 15,424 52.9  0.4 75.9  0.6 69.6
5 13,613 52.3  0.5 76.6  0.7 68.3
6 11,471 51.5  0.5 77.1  0.8 66.8
7 9567 50.9  0.6 78.0  0.9 65.3
8 7937 49.5  0.6 77.6  1 63.8
9 6530 48.6  0.7 78.3  1.1 62.1
10 5339 45.8  0.8 76.1  1.3 60.2
1342 CANCER April 1, 2007 / Volume 109 / Number 7
RCS increases denote that, as time passes, fewer
patients die of their primary head and neck cancer
in comparison to competing noncancer causes;
nonetheless, it is sobering to realize that for localized
disease, at least, there is a rather small gain in overall
5-year survival probability.
In comparison to other sites, HNSCC presents a
distinct CS profile over time. For example, in an anal-
ysis of biliary tract cancer CS using SEER 11 it was
noted that gallbladder adenocarcinomas exhibit
markedly lower initial OCS (<15%); however, OCS
steadily increases to >60% after 5 years survived
postdiagnosis.13 Comparatively, lung cancer CS
reported by Merrill et al.7 demonstrated CS curves
that had not plateaued at 5 years postdiagnosis, in
comparison to approximately 3 years for most
HNSCC subsites. In breast cancer, Henson et al.5
demonstrated that, despite increases in CS for stage
IV breast cancer as survival from diagnosis pro-
gressed, minimal improvement was noted in the CS
of patient with stages I and II disease regardless of
the number of survived years. Although to some
degree this echoes the finding in our dataset,
whereby the greatest gains in CS were noted in
patients with the poorest initial survival, we were
able to observe an initial limited increase in OCS for
patients with local and regional only HNSCC.
This series represents the largest domestic head
and neck cancer case epidemiological data registry
addressing CS. Nonetheless, several caveats must be
noted. SEER offers temporal and geographic compre-
hensiveness and statistically robust cohort sizes, but
does little to differentiate between heterogeneous
treatment cohorts. SEER descriptions of chemothera-
peutic and radiation therapy techniques are somewhat
inexact, and do not differentiate by chemotherapy reg-
imen or radiotherapy fields implemented. As treat-
ments evolve over time, CS calculations using the
methodology described overweight those patients
with greater follow-up, necessarily reducing the
impact of newer treatment methods reflected in other
large composite datasets,25,27–29 as well as possible
shifts in the etiology of HNSCC.30 Additionally, several
categorical variables within the subsite designation
are poorly formulated (ie, ‘‘Other Oral Cavity and
Pharynx’’) and provide potential for inaccurate coding
of similar tumors. SEER is also geographically limited
to those regions reporting in a given time frame and,
thus, some differences in population composition
may be noted. Whereas SEERStat is a statistically vali-
dated and powerful program, recent authors have pur-
ported, using similar national cancer registry data,
that current methodological approaches used for
deriving relative survival, similar to those used in
SEERStat, may substantially overestimate SEs for
relative survival.31 SEERStat uses population-match-
ing algorithms which account for age and ethnicity
in calculation of expected and relative survival, but
does not include other risk or demographic factors
(such as Epstein-Barr virus and human papillomavirus
status, smoking, or alcohol utilization) which might
ideally be included comparing cancer and noncancer
patient risk. We did not include patients with multiple
primary tumors in this series in order to clarify mortal-
ity from primary HNSCC alone. Because the risk of
secondary neoplasms in HNSCC is estimated at
approximately 4% annually,32 these survival estimates
are potentially optimistic compared with raw survival
data without second primary tumor exclusion.33
Nonetheless, the SEER dataset represents the largest
extant sample of cancer data domestically, and affords
estimates of CS that are statistically robust and
broadly applicable for general US head and neck can-
cer populations. Consequently, although imperfect,
these data represent a useful benchmark for more ela-
borate survival modeling efforts.
Clinically, conditional survival is a useful adjunct
to standard survival reporting. However, CS should
be considered within the context of the full spectrum
of patient care, and should not be used as a single
reference for risk assessment. Additionally, we must
remind ourselves that risk stratification based on
large-scale population data such as SEER must
always be interpreted clinically for individual
patients, whose specific risk factors must be carefully
weighed. Other anatomical cancer sites have publicly
available risk-assessment tools such as the down-
loadable MSKCC Prediction Tools,34 Numeracy,36 or
Adjuvant!,17,18,35–38 whereas at present no such tool is
widely accessible for head and neck cancers. Simi-
larly, using datasets with greater treatment-related
and staging specificity,25,26 as well as accounting for
competing causes of mortality and comorbidities,
would only improve assessment of real-time risk. In
the interim, we hope that through addition of condi-
tional survival reports careful clinicians may have
another tool added to their armamentarium.
REFERENCES
1. Henson DE, Ries LA. On the estimation of survival. Semin
Surg Oncol. 1994;10:2–6.
2. Kato I, Severson RK, Schwartz AG. Conditional median sur-
vival of patients with advanced carcinoma: Surveillance,
Epidemiology, and End Results data. Cancer. 2001;92:2211–
2219.
3. Gloeckler Ries LA, Reichman ME, Lewis DR, Hankey BF,
Edwards BK. Cancer survival and incidence from the Sur-
veillance, Epidemiology, and End Results (SEER) program.
Oncologist. 2003;8:541–552.

4. Christensen LH, Engholm G, Ceberg J, et al. Can the sur-
vival difference between breast cancer patients in Denmark
and Sweden 1989 and 1994 be explained by patho-anatom-
ical variables? A population-based study. Eur J Cancer.
2004;40:1233–1243.
5. Henson DE, Ries LA, Carriaga MT. Conditional survival of
56,268 patients with breast cancer. Cancer. 1995;76:237–
242.
6. Meng L, Maskarinec G, Lee J. Ethnicity and conditional
breast cancer survival in Hawaii. J Clin Epidemiol. 1997;50:
1289–1296.
7. Merrill RM, Henson DE, Barnes M. Conditional survival
among patients with carcinoma of the lung. Chest. 1999;
116:697–703.
8. Skuladottir H, Olsen JH. Conditional survival of patients
with the four major histologic subgroups of lung cancer in
Denmark. J Clin Oncol. 2003;21:3035–3040.
9. Davis FG, McCarthy BJ, Freels S, Kupelian V, Bondy ML.
The conditional probability of survival of patients with pri-
mary malignant brain tumors: Surveillance, Epidemiology,
and End Results (SEER) data. Cancer. 1999;85:485–491.
10. Hwang SL, Yang YH, Lieu AS, et al. The conditional sur-
vival statistics for survivors with primary supratentorial
astrocytic tumors. J Neurooncol. 2000;50:257–264.
11. Lin CL, Lieu AS, Lee KS, et al. The conditional probabilities
of survival in patients with anaplastic astrocytoma or glio-
blastoma multiforme. Surg Neurol. 2003;60:402–406.
12. Luh JY, Wang SJ, Fuller CD, Thomas CR Jr. A SEER data-
base analysis of conditional survival for prostate cancer
patients. ASCO Meeting Abstracts. 2006;24(Suppl 18):14506.
13. Fuller C, Wang S, Kachnic L, Wong A, Thomas C. Condi-
tional survival of gallbladder adenocarcinoma treated with
radiotherapy: analysis from the SEER database. Int J Radiat
Oncol Biol Phys. 2005;63:S274.
14. Thomas CR, Fuller CD, Starling GD, Wang SJ. Conditional
survival in gallbladder carcinoma: results from the SEER 11
dataset. ASCO Meeting Abstracts. 2006;24(Suppl 18):4130.
15. Wang SJ, Zamboni BA, Wieand HS, et al. Conditional
survival for patients with colon cancer: an analysis of
National Surgical Adjuvant Breast and Bowel Project
(NSABP) trials C-03 through C-06. ASCO Meeting Abstracts.
2006;24:6005.
16. Merrill RM, Henson DE, Ries LA. Conditional survival esti-
mates in 34,963 patients with invasive carcinoma of the
colon. Dis Colon Rectum. 1998;41:1097–1106.
17. Siminoff LA, Ravdin P, Colabianchi N, Sturm CM. Doctor-
patient communication patterns in breast cancer adjuvant
therapy discussions. Health Expect. 2000;3:26–36.
18. Peele PB, Siminoff LA, Xu Y, Ravdin PM. Decreased use of
adjuvant breast cancer therapy in a randomized controlled
trial of a decision aid with individualized risk information.
Med Decis Making. 2005;25:301–307.
19. Surveillance, Epidemiology, and End Results (SEER) Pro-
gram Summary Staging Manual, 2000 (http://seer.cancer.
gov/tools/ssm), ed. 2000: Surveillance Research Program,
National Cancer Institute, DCCPS, Surveillance Research
Program, Cancer Statistics Branch.
20. SEER Summary Staging Manual, 2000: Codes and Coding
Instructions for Head and Neck (http://seer.cancer.gov/
tools/ssm/headneck.pdf), ed. 2000: Surveillance Research
Program, National Cancer Institute, DCCPS, Surveillance
Research Program, Cancer Statistics Branch.
Conditional Survival in HNSCC/Fuller et al. 1343
21. Deleyiannis FW, Thomas DB, Vaughan TL, Davis S. Alco-
holism: independent predictor of survival in patients
with head and neck cancer. J Natl Cancer Inst. 1996;88:
542–549.
22. Argiris A, Brockstein BE, Haraf DJ, et al. Competing causes
of death and second primary tumors in patients with
locoregionally advanced head and neck cancer treated with
chemoradiotherapy. Clin Cancer Res. 2004;10:1956–1962.
23. Piccirillo JF, Lacy PD, Basu A, Spitznagel EL. Development
of a new head and neck cancer-specific comorbidity index.
Arch Otolaryngol Head Neck Surg. 2002;128:1172–1179.
24. Spector JG, Sessions DG, Haughey BH, et al. Delayed re-
gional metastases, distant metastases, and second primary
malignancies in squamous cell carcinomas of the larynx
and hypopharynx. Laryngoscope. 2001;111:1079–1087.
25. Pignon JP, Bourhis J, Domenge C, Designe L. Chemother-
apy added to locoregional treatment for head and neck
squamous-cell carcinoma: three meta-analyses of updated
individual data. MACH-NC Collaborative Group. Meta-
Analysis of Chemotherapy on Head and Neck Cancer. Lan-
cet. 2000;355:949–955.
26. Hoffman HT, Karnell LH, Robinson RA, Pinkston JA, Menck
HR. National Cancer Data Base report on cancer of the
head and neck: acinic cell carcinoma. Head Neck. 1999;21:
297–309.
27. Zhen W, Karnell LH, Hoffman HT, Funk GF, Buatti JM,
Menck HR. The National Cancer Data Base report on squa-
mous cell carcinoma of the base of tongue. Head Neck.
2004;26:660–674.
28. Funk GF, Karnell LH, Robinson RA, Zhen WK, Trask DK,
Hoffman HT. Presentation, treatment, and outcome of oral
cavity cancer: a National Cancer Data Base report. Head
Neck. 2002;24:165–180.
29. Hoffman HT, Karnell LH, Funk GF, Robinson RA, Menck
HR. The National Cancer Data Base report on cancer of
the head and neck. Arch Otolaryngol Head Neck Surg. 1998;
124:951–962.
30. Gillison ML. Human papillomavirus-associated head and
neck cancer is a distinct epidemiologic, clinical, and mo-
lecular entity. Semin Oncol. 2004;31:744–754.
31. Brenner H, Hakulinen T. Substantial overestimation of
standard errors of relative survival rates of cancer patients.
Am J Epidemiol. 2005;161:781–786.
32. Leon X, Ferlito A, Myer CM 3rd, et al. Second primary
tumors in head and neck cancer patients. Acta Otolaryngol.
2002;122:765–778.
33. Fuller C, Wang SJ, Goldwein J, et al. Conditional survival
probability for head and neck squamous cell cancer:
results from the SEER 11 dataset. J Clin Oncol. 2006;
24(Suppl 18). Abstract 5519.
34. Memorial Sloan-Kettering Cancer Center Prediction Tools,
Vol. 2006. New York: Memorial Sloan-Kettering Cancer
Center; 2003.
35. Adjuvant Systemic Therapy Tools, Vol. 2006. Mayo Founda-
tion for Medical Education and Research; 2005.
36. Adjuvant! Online, Vol. 2006: Adjuvant Inc., 2003–2006.
37. Ravdin PM, Siminoff LA, Davis GJ, et al. Computer program to
assist in making decisions about adjuvant therapy for women
with early breast cancer. J Clin Oncol. 2001;19:980–991.
38. Siminoff LA, Gordon NH, Silverman P, Budd T, Ravdin PM.
A decision aid to assist in adjuvant therapy choices for
breast cancer. Psychooncology. 2006;15:1001–1013.