SEARCH

SEARCH BY CITATION

Keywords:

  • osteosarcoma;
  • ErbB2;
  • survival;
  • metastasis;
  • chemotherapy

Abstract

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES

BACKGROUND

Elevated ErbB2 expression and gene amplification have been shown to be associated with poor prognosis in many cancers. Recently, it has been demonstrated that overexpression of ErbB2 protein in osteosarcoma is associated with the presence of pulmonary metastasis and decreased survival. By contrast, a previous study showed that the expression of ErbB2 declines in individual osteosarcomas as they become metastatic. In the current study, the authors determined the relation between ErbB2 status and outcome in a large number of selected patients with high-grade osteosarcoma.

METHODS

ErbB2 status was determined immunohistochemically in biopsy specimens of osteosarcoma of the extremities from 81 patients who were treated with surgery and chemotherapy. None of the patients had metastatic disease at presentation (Stage II), and all were followed-up for at least five years. The ErbB2 status was analyzed in relation to the lengths of event-free and overall survival.

RESULTS

Of the 81 tumors examined, 51 (61%) demonstrated high levels of ErbB2 expression. The presence of increased levels of ErbB2 in osteosarcoma was significantly associated with the increased probability of event-free (72.2% v. 45.6% at 5 years, P = 0.03) and overall survival (79.7% v. 58.2% at 5 years, P = 0.03). Cox multivariate analysis showed that the risk of adverse events and death was increased substantially (rate ratio: 2.24 and 2.54; 95% confidence interval, 1.07–4.72 and 1.09–5.67, respectively) among patients with decreased levels of ErbB2 protein in tumor cells, as compared with patients who had increased levels of ErbB2 in tumor cells.

CONCLUSIONS

In patients with high-grade osteosarcoma without metastatic disease at presentation and treated with surgery and chemotherapy, the presence of increased levels of ErbB2 in tumor cells is associated with a significantly increased probability of event-free and overall survival. Further data are needed before this marker can be used in making clinical decisions. Cancer 2002;94:1397–404. © 2002 American Cancer Society.

DOI 10.1002/cncr.10360

The identification of effective chemotherapy for patients with osteosarcoma has led to a significant improvement in patient outcome over the last several decades. Despite an increase in the long term rate of disease-free survival, metastases develop in 30–40% of patients, 90% of whom rarely respond to salvage treatment.1 A need exists for the capability to stratify patients at diagnosis into high- and low-risk groups.

The c-erbB2 (also called neu or HER2) protooncogene, which is located on human chromosome 17,2 encodes a 185-kilodalton transmembrane glycoprotein (ErbB2) that shows significant structural similarity to and functional association with the epidermal growth factor receptor.3–6 Overexpression of ErbB2 occurs in 15–40% of breast carcinomas, and the overexpression is associated with poor patient survival.7 Recently, two studies have shown that overexpression of ErbB2 (HER-2/neu) protein in osteosarcoma is associated with the presence of pulmonary metastases and decreased survival as well as a poor histologic response to preoperative chemotherapy.8, 9 Conversely, we have recently shown that the expression of ErbB2 decreases in individual osteosarcomas as they become metastatic, suggesting that ErbB2 does not play an important role in the development of lung metastasis.10 Since the appearance of lung metastasis is the most important prognostic factor of osteosarcoma, it is still unclear whether the overexpression of ErbB2 is involved in poor prognosis for osteosarcoma.

We aimed to determine whether the presence of increased levels of ErbB2 is a predictor of the clinical behavior of osteosarcoma. We studied 81 patients with high-grade osteosarcoma of the extremities who were treated with chemotherapy and adequate surgery. These patients had no metastatic disease at presentation and were followed-up for at least five years. Unexpectedly, there was a significant positive correlation between an increased level of ErbB2 and a favorable prognosis for osteosarcoma.

PATIENTS AND METHODS

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES

Between 1984 and 1995, 155 patients with newly diagnosed high-grade osteosarcomas were seen at two cooperative musculoskeletal oncology centers in Sapporo, Japan. Inclusion criteria for the current study included: 1) high grade osteosarcoma located in an extremity, 2) no evidence of metastasis at presentation, 3) adequate preoperative and postoperative chemotherapy, and 4) tumor excision with a wide or radical margin. Among the 155 patients, 16 were older than 45 years, 13 had evidence of metastasis, 7 did not have conventional high-grade osteosarcoma, 7 did not receive surgery, 6 did not receive adequate chemotherapy, 5 did not have osteosarcoma of the extremities, 4 were lost to follow-up, and one underwent intralesional tumor resection.

The current study included 81 patients for whom tumor samples from biopsy specimens (obtained before chemotherapy and preserved in archival paraffin-embedded tissue blocks) were available for immunohistochemic analysis. Informed consent was obtained from all patients. Biopsy specimens from the remaining 15 patients were excluded because of inadequate sample amounts (needle biopsy). The surgically resected specimens were not included. Although closed bone tumor biopsy has been reported successful in some centers with extensive experience, we preferred open biopsy to fine needle biopsy. Closed biopsy is unsuitable for more contemporary clinical and research studies of tumor tissues, such as cytogenetic and DNA analyses. Table 1 summarizes the clinical and pathologic characteristics of the 81 patients with adequate tissue for analysis. All tumors were classified as Stage II conventional high-grade osteosarcoma.14 The tumors were classified as intracompartmental (Stage IIA) in 5 patients and extracompartmental (Stage IIB) in 76 patients.

Table 1. Significance of Variables Compared with ErbB2 Expression
VariablesNo.ErbB2 expressionP valuea
negativepositive
  • a

    Calculated by chi-square test.

Age
 ≤ 144116250.74
 > 14401426
Gender
 male5324290.035
 female28622
Site of tumor
 femur4519260.69
 tibia25817
 humerus624
 fibula514
Histologic subtype
 osteoblastic4919300.80
 fibroblastic19613
 chondroblastic1257
 teleangiectatic101
Surgical staging
 Stage IIA4130.61
 Stage IIB772948
Response to chemotherapy
 good3812260.34
 poor431825

Chemotherapy was given before and after surgery according to the four regimens, NSH-5, which is equivalent to T-10 regimen by Rosen et al.,11 NSH-6,12 NSH-7,12 and NSH-8.13 The doses and cycles of drugs of each chemotherapy regimen are listed in Table 2. In each case, the surgical procedures took into account the location and extent of the tumor and the life expectancy of the patient. Limb-salvage procedures were performed in 41 patients (51%), whereas 40 patients (49%) underwent amputation. Modern surgical management of patients with osteosarcoma in the extremities usually allows up to 90% of patients to undergo limb salvage procedures. In our institutions, patients frequently underwent amputation before 1990; this is why nearly half of the patients in the current series required amputation.

Table 2. Doses and Cycles of Drugs Administered to Patients in Each Chemotherapy Regimen
RegimenPreoperative armaPreoperative therapyResponse to preoperative chemotherapybPostoperative therapy
HDMTX g/M2ADR mg/M2CDDP mg/M2IFO g/M2OthersHDMTX g/M2ADR mg/M2CDDP mg/M2IFO g/M2Others
  • HDMTX: high-dose methotrexate; ADR: doxorubicin; CDDP: cisplatin; IFO: ifosfamide; BCD: bleomycin (20 mg/M2); cyclophosphamide (600 mg/M2); dactionmycin (0.6 mg/M2); VP16: etoposide (100 mg/M2 × 3).

  • a

    In NSH-7 and NSH-8, preoperative drugs were changed if clinical response was evaluated as progressive at the end of the first cycle (B).

  • b

    The response of chemotherapy was considered good if the extent of tumor necrosis was 90% or greater and poor if it was less than 90%.

  • c

    Figures in parentheses indicate number of cycles.

NSH-58–10 (8)c30 × 2 (2)BCDGood8–12 (8)30 × 2 (3)BCD
 (n = 27)Poor30 × 2 (6)120 (6)BCD
NSH-68–10 (6)BCDGood8–10 (4)
 (n = 23)Poor30 (6)120 (6)
NSH-7A8–10 (6)30 × 2 (2)Good8–10 (4)30 × 2 (4)
 (n = 12)Poor30 × 2 (4)120 (6)
B8–10 (3)30 × 2 (3)120 (2)Good30 × 2 (4)120 (4)
Poor2 × 3 (5)VP16
NSH-8A8–12 (6)30 × 2 (2)120 (2)Good8–12 (6)30 (10)120 (2)
 (n = 8)Poor8–12 (4)30 × 2 (4)120 (4)16 (4)
B8–12 (2)30 × 2 (1)120 (1)16 (2)Good8–12 (6)30 (10)120 (2)
Poor8–12 (4)30 × 2 (4)120 (4)16 (4)

The surgical margins of the tumor specimens were histologically defined according to the system of Enneking.14 The margins were radical in 10 patients and wide in 71 patients. The extent of tumor necrosis was evaluated according to a previously described semiquantitative method. The response of chemotherapy was considered good if the extent of tumor necrosis was ≥ 90% and poor if it was < 90%.15

Immunohistochemistry

All patient archival materials were retrieved and stained with monoclonal antibody CB11 (Novocastra Laboratories, Newcastle upon Tyne, UK) at 1/40 dilution to detect ErbB2 receptor protein. The correlation of ErbB2 expression with the prognosis of osteosarcoma was first shown using CB11 antibody. In that study, ErbB2 expression in osteosarcoma evaluated by immunohistochemistry with CB11 antibody was shown to be closely correlated with ErbB2 expression detected by immunoblotting.8 Bony specimens were decalcified overnight. Sections were cut at 4-5 μm, deparaffinized, and rehydrated. Pretreatment consisted of digestion with 0.05% trypsin followed by microwave treatment for 10 minutes. Slides were immersed for 20 minutes in 0.3% absolute methanol and then preincubated with 5% normal rabbit serum (Nichirei, Tokyo, Japan) for 15 minutes at room temperature. Overnight incubation with monoclonal antibody CB11 at 4 °C was followed by rinsing with phosphate-buffered saline for 30 minutes. Biotinylated rabbit antimouse immunoglobulin G (Nichirei, Tokyo, Japan) was applied for 60 minutes at room temperature. After rinsing with phosphate-buffered saline (PBS) for 30 minutes, slides were incubated with the streptavidin-biotin-peroxidase complex (Histofine kit, Nichirei, Tokyo, Japan) for 45 minutes at room temperature, then rinsed again with PBS for 30 minutes. Color was developed by incubating slides with 3,3′-diaminobenzidine tetrahydrochrolide in the presence of 0.05% H2O2 in PBS for 15 minutes. Slides were then rinsed with tap water, counterstained with hematoxylin, rinsed for 2 seconds in 1% acid alcohol, rinsed for 15 seconds in ammonia water, dehydrated, and coverslipped.

Positive controls, made up of cases showing more than an 80% positive staining for ErbB2, as well as negative controls, in which the primary antibody was omitted, were included with each run.9 Archival breast carcinomas known to overexpress ErbB2 were also used as positive controls. Membrane and cytoplasmic staining were not distinguished. The results were expressed according to a semiquantitative scale, which was a modification of the five point scoring system reported by Corbett et al.16 ErbB2-positive tumor samples were graded from 0 to 3+ according to the proportion of tumor cells staining as follows: 0, complete absence of tumor cells staining positive; 1+, 1-30% of the cells staining positive; 2+, 31-75% of the cells staining positive; 3+, 76-100% of the cells staining positive. Intensity was not considered as part of the grading system. The highest degree of positivity found in any area of the section was recorded. Increased levels of ErbB2 were inferred in the samples graded either 2+ or 3+.10 This 30% cutoff point was based on the study of Muss et al.17 All slides were evaluated for ErbB2 overexpression by two pathologists (Y. K. and K.Y.) without any knowledge of patient information. The two pathologists had generally similar estimates of the frequency of stained cells on the slides (less than 5% discrepancy between estimates).

Statistical Analysis

The chi-square test was used to evaluate the association between two dichotomous variables. Kaplan-Meier plots and the log rank test were used to evaluate the association of the percentage of necrosis and the expression of ErbB2 with event-free and overall survival. Cox proportional hazards regression analysis with forward selection of variables was performed to estimate the rate ratios for possible risk factors for the occurrence of adverse events. Data analysis was performed with the Statview statistical package, version 6.0 (Abacus Concepts, Cary, NC).

RESULTS

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES

Of the 81 patients we examined, 14 did not show immunoreactivity for ErbB2 (Grade 0) and 16 had only scattered positive cells (Grade 1+). These 30 tumors (37%) were considered ErbB2 negative. Of the remaining 51 tumors, 23 were graded 2+ and 18 were graded 3+. These 51 tumors (63%) were considered to have increased levels of ErbB2 (Fig. 1).

Association of the Expression of ErbB2 with Clinical and Pathologic Features

Immunostaining for ErbB2 protein was positive more frequently in female patients than in male patients. This association was statistically significant (P = 0.035). No significant differences in the incidences of ErbB2 positivity were found among histologic subtypes.

No relation was found between the level of expression of ErbB2 and the extent of tumor necrosis; overexpression of ErbB2 was found in tumors from 26 out of 38 patients (68%) with good responses and 25 out of 43 patients (58%) with poor responses (P = 0.34).

In the 81 patients, increased levels of ErbB2 were significantly associated with an increased probability of remaining event-free after diagnosis (P = 0.011) (Fig. 2); the probability of event-free survival at five years was 72% in patients with increased levels of ErbB2 protein and 46% in patients without such overexpression. In addition, increased levels of ErbB2 were significantly associated with an increased probability of overall survival (P = 0.028) (Fig. 3). The histologic response to preoperative chemotherapy was significantly related to the probabilities of event-free survival (P = 0.011) but not to the probability of overall survival (P = 0.057). With regard to operative procedures, limb-sparing surgery was significantly associated with increased probabilities of event-free and overall survival (P = 0.0009 and P = 0.0033, respectively). This may have been due to a bias toward performing limb salvage procedures for patients with more manageable or smaller tumors. Other clinicopathologic features, including age, gender, anatomic site, and histologic subtype, were not associated with outcomes (Table 3).

thumbnail image

Figure 1. Immunohistochemic detection of ErbB2 in osteosarcoma with monoclonal antibody CB11. A) No detectable staining (Grade 0); B) 1-30% of the cells stained positive (Grade 1+); C) 31-75% of the cells stained positive (Grade 2+); D) 76-100% of the cells stained positive (Grade 3+). Grades 2+ and 3+ staining were considered positive.

Download figure to PowerPoint

thumbnail image

Figure 2. A) Kaplan-Meier plots for event-free survival and B) overall survival for ErbB2 positive cases compared with negative cases.

Download figure to PowerPoint

Table 3. Results of Univariate Analyses of Event-Free and Overall Survival
VariableNo. of patientsEvent-free survivalOverall survival
Rate ratio95% CIP valueRate ratio95% CIP value
  • CI: confidence interval.

  • a

    Reference category.

Time of diagnosis
 1980–89432.51.2–7.40.0192.30.8–6.20.10
 1990–95a381.01.0
Age of patients
 ≤ 14411.70.8–3.40.171.20.5–2.80.61
 14a401.01.0
Gender
 malea531.01.0
 female280.480.2–1.10.0910.800.3–1.90.63
Histologic type
 osteoblastic491.90.7–5.00.212.50.7–8.40.15
 chondroblastic122.00.6–6.80.291.90.4–9.30.44
 fibroblastica191.01.0
Site of tumor
 femur441.70.7–3.90.25
 tibiaa251.0
 humerus61.10.2–5.30.91
 fibula51.30.3–6.40.72
Response to chemotherapy
 good380.360.2–0.80.0110.420.2–1.00.057
 poora431.01.0
Operative procedure
 amputation413.91.8–8.80.00094.41.6–120.0033
 limb-sparinga401.01.0
Chemotherapy regimen
 NSH5, 6a461.01.0
 NSH7, 8350.320.1–0.80.0130.330.1–1.00.045
ErbB2 expression
 negative302.431.2–4.90.0142.441.1–5.40.034
 positivea511.01.0

Univariate analyses showed that the time of diagnosis, chemotherapy regimen, response to preoperative chemotherapy, type of operation, and ErbB2 immunoreactivity were associated with the probability of event-free survival (Table 3). Multivariate analysis revealed that only negative staining of ErbB2 persisted as an independent risk factor for a poor outcome. Patients with negative ErbB2 had a significantly higher rate ratio (2.53) for decreased overall survival than patients with increased levels of ErbB2 (Table 4).

Table 4. Multivariate Analyses of Event-Free and Overall Survival
VariableEvent-free survivalOverall survival
Rate ratio95% CIP valueRate ratio95% CIP value
  • CI: confidence interval.

  • a

    Reference category.

Time of diagnosis
 1980–891.20.4–3.30.770.90.3–3.00.90
 1990–95a1.01.0
Chemotherapy regimen
 NSH5, 6a1.01.0
 NSH7, 80.450.1–1.50.180.380.1–1.40.16
Response to chemotherapy
 good0.520.2–1.20.140.710.3–1.80.47
 poora1.01.0
Operative procedure
 amuputation3.391.51–7.640.00063.901.44–9.570.051
 limb-sparinga1.01.0
ErbB2 expression
 negative2.241.07–4.720.032.531.09–5.670.03
 positivea1.01.0

DISCUSSION

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES

We found that high levels of ErbB2 in tumor cells correlated significantly with increased event-free survival as well as overall survival in 81 patients with osteosarcoma who had no metastatic disease at presentation (Stage II). No relation was found between the level of expression of ErbB2 and the extent of tumor necrosis after preoperative chemotherapy. Multivariate analysis revealed that negative staining of ErbB2 persisted as an independent risk factor for poor outcome. These data suggested that a decreased level of ErbB2 could be used to identify tumors with a tendency to progress despite chemotherapy.

Two previous studies suggested that overexpression of ErbB2 may be associated with pulmonary metastasis and poor prognosis for patients with osteosarcoma. In one, 26 osteosarcoma tumor samples from 20 patients without metastatic disease (Stage II) and 6 patients with metastatic disease (Stage III) were evaluated by immunohistochemistry. Overexpression of ErbB2 was observed in 42% of the tumor samples and was associated with pulmonary metastasis and poor patient prognosis.8 In the other study, 53 tumor samples from 43 Stage II and 10 Stage III osteosarcoma patients were evaluated. At the time of the initial biopsy, 20 out of 47 samples (42.6%) showed high levels of ErbB2 expression. Expression of ErbB2 was correlated with a significantly worse histologic response (P = 0.03) and worse event-free patient survival (P = 0.01). If the patients who presented with metastatic disease (Stage III) were excluded, this difference in EFS remained but became less significant (P = 0.05).9

What accounts for the differences between the results of these studies and the current results? In the first study, by Onda et al.,8 patients with metastatic disease at presentation (Stage III) were not excluded. Furthermore, chemotherapy protocols were not satisfactorily described. The most important predictor of outcome at diagnosis is the presence or absence of metastasis, usually pulmonary metastasis. Current protocols for the treatment of osteosarcoma, which use a combination of surgery and chemotherapy, achieve a 5 year disease–free survival rate of approximately 70% in patients who do not have metastatic disease at presentation. In contrast, the 15–20% of patients who present with metastatic disease (Stage III) at diagnosis have an extremely poor prognosis, with approximately only 10% achieving long-term survival. It is possible that the clinical behavior and biologic features of Stage III osteosarcoma could be distinct from those of Stage II osteosarcoma. Thus, these two groups of patients should be studied separately. Gorlick et al.9 expanded on the previous work and examined a uniformly treated cohort of patients. However, neither the overall survival rate nor data for univariate and multivariate analyses were presented in that study. Only a permutation test based on the log rank statistic was used to compare event-free survival rates between patients with positive and negative staining for ErbB2. Slamon et al.18 showed in their study on breast carcinoma that tumor size was a slightly better predictor than HER-2/neu gene amplification in univariate analysis. However, its significance was lost on multivariate analysis and was not independent of nodal status. This clearly indicates that multivariate analysis is indispensable to determine independent predictors of both disease relapse and overall survival.

We propose the following biologic explanations for the association of ErbB2 overexpression with the favorable outcome for osteosarcoma patients who did not have metastatic disease at presentation and were treated with surgery and high-dose neoadjuvant chemotherapy. We previously compared the expression levels of the ErbB2 receptor protein among biopsy specimens, tumors resected after preoperative chemotherapy, and pulmonary metastatic lesions in each of 19 patients with osteosarcoma who had not presented with metastasis at diagnosis. Loss of ErbB2 expression was noted in 14 of the 19 patients as the initial tumor became metastatic. The loss of expression occurred during high-dose adjuvant chemotherapy.10 These observations suggest that the overexpression of ErbB2 does not play an important role in lung metastasis, which determines the prognosis of osteosarcoma. This is supported by laboratory data showing that the activated c-erbB2 gene can enhance experimental but not spontaneous metastatic potential in murine colon carcinoma cells.19 Moreover, it is possible that ErbB2 positive clones were deleted during high-dose adjuvant chemotherapy and that surviving ErbB2 negative clones became metastatic. Therefore, it is conceivable that osteosarcoma patients whose tumors overexpress ErbB2 may benefit from high-dose adjuvant chemotherapy and have longer disease-free and overall survival than those whose tumors do not overexpress ErbB2.

Overexpression of ErbB2 has been shown in 15-30% of patients with breast carcinoma and has been found by most but not all investigators to be associated with shorter survival.18, 20–23 This divergence may be related to differences in methods analysis or in patient selection. In patients with node-positive early breast carcinoma, those randomly assigned to the high-dose regimen of adjuvant chemotherapy had significantly longer disease-free and overall survival if their tumors had ErbB2 overexpression. There was a significant dose-response effect of adjuvant chemotherapy in patients with overexpression of ErbB2 but not in patients with no or with minimal ErbB2 expression.17 A study on the correlation between the dose-response effect of preoperative chemotherapy and ErbB-2 status in osteosarcoma patients is now underway in our department.

The most important finding of the current study is that in patients with high-grade osteosarcomas of the extremities that had not metastasized at the beginning of the study, decreased levels of ErbB2 could be used to identify tumors with a tendency to progress despite chemotherapy. The risk of adverse events was increased over twofold in patients whose tumors had decreased levels of ErbB2. Moreover, in patients whose tumors contained decreased levels of ErbB2, the probability of overall survival was significantly lower than in patients whose tumors had increased levels of ErbB2, regardless of the histologic response to chemotherapy. Another advantage of the use of ErbB2 status over estimation of the extent of necrosis as a prognostic factor is that the former can be determined at the time of initial diagnosis, before any therapy is initiated. Estimates of ErbB2 expression are currently provided by many laboratories as part of the prognostic profile, but further data are needed before this marker can be used in making clinical decisions.

An interesting finding is that the immunostaining for ErbB2 protein was positive more frequently in female patients than in male patients. This gender difference was also found in osteosarcoma patients by Gorlick et al.9 and lung carcinoma patients by Guinee et al.24 A potential explanation for this gender bias comes from a report demonstrating that estradiol can bind and activate the ErbB2 receptor.25

A limitation of the current study was that the patients received four different chemotherapy regimens. The advent of modern chemotherapy has been the most important factor in increasing event-free and overall survival in patients with osteosarcoma. Therefore, it was very important to determine whether the results of the current study would remain consistent despite differences in chemotherapy. As shown in Table 2, NSH-6, 7, and 8 regimens were created based on the NSH-5 regimen or T-10 protocol of Rosen et al.11 In addition, multivariate analysis revealed that there was no significant difference in event-free and overall survival between the patients receiving NSH-5 or 6 regimens and those receiving NSH-7 or 8 regimens. However, the current study should be reproduced with a large number of patients who receive a single chemotherapy regimen.

Monoclonal antibodies against the extracellular domain of ErbB2 protein inhibit the proliferation of murine fibroblasts transformed by the activated or overexpressed c-erbB2 gene in vitro and in vivo.26, 27 The same phenomena were found in human cancers that overexpressed ErbB2.28–30 Most recently, Slamon et al. reported the results of a Phase 3 trial to evaluate the efficacy and safety of trastuzumab, a recombinant monoclonal antibody against ErbB2, in patients with metastatic breast carcinoma that overexpressed ErbB2.31 The application of this strategy to osteosarcoma still needs to be examined in more detail.

REFERENCES

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES