available to undergo immunohistochemical analysis. The
remaining patients in this arm of the study had pretreat-
ment biopsy specimens that were too small to evaluate by
this method. Of the 20 patients who were eligible for this
study, 10 were responders to chemotherapy and subse-
quently retained their larynges. The remaining 10 did not
respond to induction chemotherapy and required salvage
laryngectomy with postoperative radiation therapy for
disease treatment. The former group will be referred to as
the responders and the latter as the non-responders to
induction chemotherapy. All patients were male, and the
mean age at presentation was 62 years (Table I). Fifteen of
the 20 patients had stage III disease at the time of pre-
sentation, and 5 patients presented with stage IV tumors.
Overwhelmingly, T3 tumors predominated, with 17 T3
tumors, 1 T2, and 2 T4 tumors. Nodal status was deter-
mined to be predominantly N0 (15 patients), with only 2
patients presenting with N1 neck disease, 2 patients with
N2 disease, and 1 with N3 disease. The predominant site of
the primary tumor was supraglottic larynx (15 of 20). As
evident in Table I, there was no statistically significant
difference between the groups with regard to stage,
T-status, N-stage, race, age, sex, site of the primary, or
Karnovsky Performance Scale.
TABLE II.
Results.
Vessel Counts
(per HPF)
Range
P Value
Յ.008
Chemoresponsive
Chemoresistant
20.90 Ϯ 8.09
32.99 Ϯ 10.10
8.75–35.90
21.05–53.10
patients with the lower vessel counts (Fig. 7). The odds
ratio reveals that patients with vessel counts above the
mean are 3.5 times more likely to require a laryngectomy
than those below the mean. The sensitivity using this
value is 60% (95% CI 30%, 90%) with a specificity of 70%
(95% CI 42%, 98%), a positive predictive value of 67%
(95% CI 36%, 98%), and a negative predictive value of 64%
(95% CI 36%, 92%).
Finally, Kaplan-Meier’s survival curve analysis re-
vealed that patients in this study whose mean vessel
counts were above the mean, compared with those below
the mean, tended to display poorer long-term survival,
although these differences were not statistically signifi-
cant (P ϭ .54) (Figs. 4 and 6). However, when comparing
the most-vascular tumors in this group of patients (those
with a mean vessel count 1 SD above the mean) to the rest
of the group, a decrease in survival was noted (P ϭ .0345)
(Fig. 5).
Tumor Angiogenesis Counts
Mean vessel count for all 20 patients was 26.9 vessels
per high-power field (HPF), with a median of 25.575. The
group of patients who responded to chemotherapy and
went on to complete the organ preservation protocol had a
mean vessel count of 20.90 Ϯ 8.09 vessels per HPF. The
range was 8.75 to 35.90 vessels per HPF. In the group of
patients who did not respond to chemotherapy, the mean
vessel count was 32.99 Ϯ 10.10 vessels per HPF (range,
21.05–53.10) (Table II). This difference was found to be
highly statistically significant (P ϭ .0085). Patients with
vessel counts above the mean were eight times less likely
to respond to chemotherapy (relative risk ϭ 8.0) with a
screening sensitivity of 80% (95% confidence interval [CI]
45%, 100%), a specificity of 67% (95% CI 43%, 91%), a
positive predictive value of 44% (95% CI 12%, 76%), and a
negative predictive value of 91% (95% CI 74%, 100%).
Mean tumor vessel counts at the time of presentation did
not correlate with nodal stage at presentation (P ϭ .408)
or disease recurrence (P ϭ .2105).
DISCUSSION
The concept of angiogenesis has been firmly estab-
lished as a basic feature in tumor growth and metastasis.
Although it has long been recognized that solid tumors
contain large numbers of highly permeable blood vessels,
the critical importance of these vessels in allowing tumor
growth remained unappreciated until the early 1970s. At
that time, Dr. Judah Folkman in his landmark studies
defined the principles governing angiogenesis, and the
modern era of research in the field of tumor angiogenesis
began. Dr. Folkman hypothesized that new blood vessel
formation at the primary tumor site was absolutely re-
quired for tumor nodules to grow beyond a diameter of 2
mm.6,7 This is the size in which diffusion of nutrients and
waste products becomes the rate-limiting step to growth.
Inhibiting angiogenesis, therefore, would inhibit tumor
growth at this 2-mm size. On the other hand, tumors
which are able to recruit neovascularity are capable of
unlimited growth and metastasis.6 Folkman and others
have demonstrated a strong correlation between tumor
growth and vessel concentration in non-head and neck
tumors.8 A measurement of tumor angiogenesis or neovas-
cularity has been suggested as a prognostic marker in
malignant melanoma, breast, prostate, ovarian, gastric,
and lung carcinomas.5,9–11 In squamous cell carcinomas of
the head and neck, tumor angiogenesis or microvessel
density has only recently begun to be investigated. Sev-
eral studies have investigated the role of tumor angiogen-
esis in predicting local regional recurrence and surviv-
al.12–19 These studies have investigated multiple sites in
the head and neck concurrently, and the results have been
mixed, with some revealing a positive and others a nega-
Kaplan-Meier curves evaluating the likelihood of or-
gan preservation based on low versus high vascularity are
graphically depicted in Figures 2 and 3. There is a definite
trend in both groups toward organ preservation in those
TABLE I.
Results.
Chemoresponsive Chemoresistant P Value
Age (yr)
61.69
9/1
62.18
6/4
.141
.303
.171
.381
.606ñ
Stage III/IV
T stage (T2/T3/T4)
N status (N0/N1/N2/N3)
Site (glottic/supraglottic)
0/10/0
7/2/1/0
3/7
1/7/4
8/0/1/1
2/8
Laryngoscope 112: May 2002
846
Teknos et al.: Tumor Angiogenesis as a Predictive Marker