SEARCH

SEARCH BY CITATION

Abstract

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

Objective

To determine whether there are factors, such as the diffusing capacity for carbon monoxide (DLCO) or pulmonary artery pressure (PAP) on echocardiogram, that can predict the development of pulmonary hypertension (PHT) in patients with limited scleroderma.

Methods

Using the large Pittsburgh Scleroderma Databank, 106 patients who had the diagnosis of PHT after January 1, 1982, were matched with 106 controls by scleroderma subtype, age, sex, race, disease duration, and the mean time to the diagnosis of PHT after the initial Pittsburgh visit. Autoantibodies, vascular features, use of calcium channel blockers, extent of pulmonary function, and echocardiogram findings were determined at any time prior to the diagnosis of PHT (or prior to the matched time in controls).

Results

Patients with PHT had a mean DLCO of 52% of predicted at an average of 4.5 years prior to the diagnosis of PHT. This was markedly decreased compared with the values in controls, whose mean DLCO was 81% of predicted (P < 0.0001). The estimated mean PAP on echocardiogram was only slightly higher in the PHT patients compared with controls (34 mm Hg versus 29 mm Hg; P not significant). Nineteen PHT patients had 4 serial measurements of the DLCO during the 15 years prior to the diagnosis of PHT. The initial mean DLCO was 80% of predicted, which decreased in a linear manner to a mean of 35% of predicted at the time of diagnosis of PHT, whereas the value in controls remained at ∼80% of predicted (P < 0.0001). PHT patients had more severe Raynaud's phenomenon and more severe digital tip ulcers, but they used calcium channel blockers significantly less frequently (37% versus 61% of controls; P < 0.01). The predominance of nucleolar autoantibodies and the absence of anti–Scl 70 antibody were associated with PHT.

Conclusion

A decreasing DLCO is an excellent predictor of the subsequent development of isolated PHT in limited scleroderma. The DLCO may be significantly decreased for many years prior to the diagnosis of PHT. The presence of autoantibodies and the PAP may also be helpful predictors. The long-term use of calcium channel blockers may be protective, but newer agents that are more effective in treating PHT may also be helpful in altering the natural history of this serious complication in limited scleroderma.

Pulmonary hypertension (PHT) is the leading cause of scleroderma-related deaths in patients with systemic sclerosis (SSc) and limited cutaneous involvement (1). Until recently, this complication nearly always was rapidly progressive and uniformly fatal. However, in the past few years, potent pulmonary vasodilators and lung transplantation have improved patient well-being and may have improved survival.

In 1977, findings from a series of 10 SSc patients with PHT were reported by our colleagues at the University of Pittsburgh (2). All patients had limited scleroderma and had either minimal or no pulmonary interstitial fibrosis. Autopsy studies confirmed that noninflammatory intimal proliferation of the medium-size pulmonary arteries was the cause of this type of PHT (3). This bland vasculopathy results in an illness very similar to primary PHT. Many patients with limited scleroderma have mild dyspnea on exertion during their disease course, but this symptom is usually attributed to deconditioning or other aspects of the disease. Echocardiograms may yield completely normal findings, even after the onset of dyspnea. However, in some patients, the echocardiogram shows mild elevation of the pulmonary artery pressure (PAP) for many years, regardless of whether the patient ultimately develops the severe form of PHT (4).

PHT is seen in SSc patients in a variety of other settings, but when it occurs without other etiologies, it is termed isolated PHT. PHT can also occur secondary to severe pulmonary fibrosis, renal crisis, or cardiomyopathy with diastolic dysfunction. Such secondary causes of PHT are more likely to occur in patients with diffuse scleroderma.

Patients with isolated PHT have, at some point, rapidly progressive dyspnea on exertion, leading to right heart failure and death from hypoxemia and ventricular arrhythmia. As medications are developed that have the potential to modify or even prevent PHT, it is important to identify, early in the course of SSc, those patients who are most likely to develop PHT. The present study uses case–control methods to ascertain the risk factors that may predict the subsequent development of PHT.

PATIENTS AND METHODS

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

Cases.

Cases were defined as all SSc patients from the Pittsburgh Scleroderma Databank who were diagnosed as having isolated PHT after January 1, 1982. This year was chosen because it was the time when an echocardiogram became a routine part of our initial patient evaluation, and because this was the time when calcium channel blockers began to be used for the treatment of Raynaud's phenomenon in SSc. Isolated PHT was defined as a pulmonary artery systolic pressure >30 mm Hg on echocardiogram (4, 5), mild to moderate dyspnea on exertion (World Health Organization class III), and at least 1 additional objective finding of PHT, including clinical signs of right heart failure with peripheral edema or ascites or right heart abnormalities on echocardiogram. An asymptomatic patient with increased PAP but no other evidence of PHT was not diagnosed as having PHT, unless they later became symptomatic or developed signs of right heart failure.

We chose not to include these asymptomatic patients with mild, isolated increases in the PAP because the natural history in such patients is unknown. Several echocardiographic studies have shown that up to 65% of SSc patients have increased PAPs, but clearly, the severe, progressive, deadly form of PHT is not that frequent (4, 6). Thus, we focused on only those patients who had definite life-threatening PHT. These patients did not have significant interstitial fibrosis on chest radiograph, and the forced vital capacity (FVC) on pulmonary function testing was required to be >70% of predicted. Patients with other causes of PHT, including valvular heart disease, thromboembolic disease, or use of anorectants, were excluded. The diffusing capacity for carbon monoxide (DLCO) in most patients was very low at the time of diagnosis of PHT, as we have previously described (7).

Controls.

Controls were patients with SSc from the Pittsburgh Scleroderma Databank who did not have PHT during the study time and who were first evaluated after January 1, 1982. Controls were matched with cases using a case-matching program from our data management system, MEDLOG. One control was identified for each case. Cases and controls were matched on the following features: 1) scleroderma subtype (limited scleroderma or diffuse scleroderma) (8); 2) date of initial Pittsburgh visit (±1 year); 3) age at disease onset (±5 years); and 4) disease duration from onset to initial visit (±5 years). Each control had a visit or time that best correlated to the time when the case was diagnosed as having PHT, referred to as the “matching time.” When a patient had the diagnosis of PHT made at the time of the initial visit, the initial visit was considered to be the “matching time” for the control. Controls for cases who developed PHT after their first visit were matched with these cases at a time that was closest to the interval between the initial Pittsburgh visit and the time that the case was diagnosed with PHT.

Data analysis.

The Pittsburgh Scleroderma Databank comprises a prospective cohort of >2,500 consecutive patients with scleroderma seen at the University of Pittsburgh since January 1, 1972. All patients undergo a standardized evaluation on the first visit, and all available information related to their illness is collected and entered into a computerized database (9). Biannually, all patients are contacted and requested to complete a standardized questionnaire, which includes the scleroderma Health Assessment Questionnaire (HAQ) that has been used since 1980 (10). Information on severe organ involvement, hospitalizations, and results of pulmonary function tests and echocardiograms is requested and obtained with the patient's consent. All deaths are investigated by obtaining information from the patient's family, primary physician, or rheumatologist and from hospital data, to accurately determine the cause of death. We have maintained a 95% accountability of these patients (11).

After cases and controls were matched, the results of pulmonary function tests and echocardiogram data were examined prior to the development of PHT in the cases and prior to the matching time in controls. We also studied the frequency of digital tip ulcers and digital gangrene, and we used the patient-generated visual analog scale to determine the severity of Raynaud's phenomenon and digital tip ulcers (scale 0–3) (10). Lastly, we determined the frequency and duration of the use of calcium channel blockers, prior to the diagnosis of PHT in cases and prior to the matching time in controls.

For data analysis, we used standard descriptive statistics, McNemar's matched-pair analysis, and survival analysis.

RESULTS

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

We identified 106 cases of isolated PHT. Ninety percent of these patients had limited scleroderma, 85% were female, and the mean age at the onset of scleroderma was 41 years. The mean disease duration from onset of first symptom attributable to SSc to the time of PHT diagnosis was 14.4 years. At the time of PHT diagnosis, the mean FVC was 78% of predicted and the mean DLCO was 40% of predicted. Eighty percent of the 79 patients who had pulmonary function tests performed at or near the time of PHT diagnosis had a DLCO of <55% of predicted. The mean pulmonary artery systolic pressure measured at right heart catheterization, or estimated from echocardiogram if catheterization was not performed, was 76 mm Hg (range 50–106 mm Hg). During this time period, there were 43 patients with limited scleroderma who had an increased pulmonary artery systolic pressure (>30 mm Hg) on echocardiogram but did not fulfill our criteria for PHT. Eighteen of these patients fulfilled the criteria for PHT during followup and were included as PHT cases. The remaining 25 patients with PAPs >30 mm Hg were eligible to be controls, and 6 were chosen to be controls using the computer matching program. Of these 25 patients, 6 developed PHT after this study was completed, including 1 of the 6 controls.

Table 1 shows the expected close similarity between the cases and matched controls with respect to the matching variables. It is important to note that the mean time of the first visit in cases and controls was within 1 month of each other. This finding increases the likelihood that both of the patients in the matched pairs had an echocardiogram performed and were prescribed a calcium channel blocking agent.

Table 1. Demographic information on pulmonary hypertension (PHT) scleroderma cases and non-PHT scleroderma controls
 Cases (n = 106)Controls (n = 106)
  • *

    Among patients who had PHT diagnosed after the first visit (or matching time [MT] in controls).

Sex, % female8582
Limited scleroderma, %9090
Mean age at disease onset, years40.640.4
Disease duration at first visit, years11.611.3
Mean date of first visitDecember 1986January 1987
Mean time to PHT diagnosis (or MT) after first visit, years (n = 42)*7.457.62

The clinical features of scleroderma at the time of diagnosis of PHT and the serum autoantibody results in the 2 groups are shown in Table 2. There was little difference between these groups in the Rodnan total skin score (9), frequency of gastrointestinal involvement, and frequency of pulmonary interstitial fibrosis via chest radiograph. Both groups had a similar degree of restrictive lung disease based on the mean FVC, although this parameter was not part of the matching criteria. Pericardial effusions and right bundle branch block on electrocardiogram were significantly more common in the PHT patients, but the occurrence of documented cardiac arrhythmias was similar. When the data at the time of the diagnosis of PHT were excluded from the analysis, the results showed similar trends, except that there were no statistically significant differences between cases and controls.

Table 2. Clinical features in patients with limited scleroderma with (cases) and without (controls) pulmonary hypertension*
 CasesControls
  • *

    GI = gastrointestinal; EKG-RBBB = electrocardiogram showing right bundle branch block.

  • P < 0.001 versus controls.

Maximum skin score, mean6.65.8
GI involvement, %4743
Fibrosis on chest radiograph, %3540
Mean forced vital capacity, % predicted7885
Pericardial effusion, %5513
EKG-RBBB, %256
Cardiac arrhythmia, %2021
Serum autoantibodies, no./total (%)  
 Anticentromere43/98 (43)42/97 (43)
 Anti-Th/To15/48 (31)10/34 (29)
 Anti–U3 RNP13/50 (26)1/30 (3)
 Anti-U1 RNP10/50 (20)9/53 (17)
 Anti–Scl-700/88 (0)12/93 (13)

The most frequently encountered serum autoantibody was the anticentromere antibody, which was found in 43% of both the cases and the controls. Thus, although the anticentromere antibody is associated with limited scleroderma, it does not predict the development of PHT in that clinical subset (12). The anti–U3 RNP antibody was tested in small numbers of patients but was almost exclusively seen in the cases (P < 0.001) versus controls. The anti–Scl 70 antibody appears to be “protective,” since no PHT cases had this antibody (P < 0.001 versus controls). The anti-Th/To and anti–U1 RNP antibodies did not distinguish between those with PHT and those without; but anti-Th/To was examined in only one-third of the patients.

Table 3 summarizes the data on the DLCO and pulmonary artery systolic pressures estimated from echocardiogram prior to the diagnosis of PHT in cases and prior to the matching time in controls. The time of the respective test in relation to the matching time and the number of patients studied are included. Results on the DLCO were available for 62 cases and 47 controls at 4.5 years and 6 years prior to the diagnosis of PHT or the matching time in controls, respectively. The mean DLCO was significantly lower in PHT cases compared with controls (52% versus 81%; P < 0.0001). Possibly for technical reasons, smaller numbers of cases and controls had estimations made of the PAPs on echocardiogram. Among the PHT cases, the estimated mean pulmonary artery systolic pressure was slightly increased at 34 mm Hg, but this was not significantly different than the mean of 29 mm Hg in controls.

Table 3. Evidence of pulmonary function or echocardiographic abnormalities in cases and controls, prior to the time of pulmonary hypertension (PHT) diagnosis in cases or at the matching time (MT) in controls
 CasesControls
  • *

    DLCO = diffusing capacity for carbon monoxide.

  • P < 0.0001 versus controls.

Mean DLCO, % predicted*52 (n = 62)81 (n = 47)
 Mean time prior to PHT or MT, years4.56.0
Mean systolic pulmonary artery pressure, mm Hg34 (n = 40)29 (n = 27)
 Mean time prior to PHT or MT, years3.93.5

We next considered patients who had undergone pulmonary function tests during each of 4 specific time periods, as follows: 1) at the time of or up to 0.5 years before the diagnosis of PHT (or matching time for controls); 2) 0.5–5 years before PHT diagnosis; 3) 5–10 years before PHT diagnosis; and 4) 10–15 years prior to diagnosis of PHT. Nineteen cases and 14 controls had at least 1 PFT study done in each of these 4 intervals. Figure 1 shows that both groups began this time period (10–15 years pre-PHT or matching time) with a mean DLCO of 80% of predicted, but the mean DLCO in cases decreased during each successive time period, reaching a mean of 35% of predicted at the time of PHT diagnosis. In comparison, the mean DLCO in controls stayed close to 80% of predicted throughout this matched time period. This difference is highly significant (P < 0.0001) and strongly suggests that in patients who subsequently develop isolated PHT, there is a progressive deterioration in the DLCO over a 15-year time period prior to the diagnosis of PHT.

thumbnail image

Figure 1. Distribution of the mean diffusing capacity for carbon monoxide (DLCO; in % of predicted [% Pred.]) among patients with limited scleroderma with (cases) or without (controls) a subsequent diagnosis of pulmonary hypertension (PHT) who underwent at least 1 pulmonary function test in each of the 4 time periods over 15 years prior to the diagnosis of PHT (or matching time in controls).

Download figure to PowerPoint

We also examined some of the peripheral vascular features and the use of calcium channel blocking agents in these patients (Table 4). Both cases and controls had a similar frequency of Raynaud's phenomenon, digital tip ulcers, and digital gangrene. However, visual analog scales for Raynaud's phenomenon, which are part of the scleroderma-modified HAQ that we have developed and validated (9), showed that cases had significantly higher values for both the severity of Raynaud's phenomenon and the severity of digital tip ulcers. Interestingly, even though PHT cases had more severe digital vascular problems, they used calcium channel blocking agents, at any time during the pre-PHT period, significantly less frequently (37%) compared with the controls (61%; P < 0.001). Also, the duration of the use of calcium channel blocking agents was significantly shorter in cases (0.9 years versus 2.7 years in controls; P < 0.0001). Thus, only one-third of the PHT cases took calcium channel blockers for a mean of slightly less than a year. As noted above, we believe that both cases and controls had an equal likelihood of having these drugs prescribed.

Table 4. Vascular features in cases and controls, prior to PHT diagnosis (or MT in controls)*
 CasesControlsP
  • *

    NS = not significant (see Table 3 for other definitions).

Raynaud's phenomenon, %9796NS
Digital tip ulcers, %3534NS
Digital gangrene, %138NS
Raynaud's severity scale (0–3), mean score1.030.71<0.05
Digital tip ulcer severity scale (0–3), mean score0.750.45<0.05
New ulcers at time of PHT (or MT), %3810<0.05
Past use of calcium channel blockers, %3761<0.001
Mean duration of calcium channel blockers, years0.912.70<0.0001

The PHT cases had severe PHT with marked increases in pulmonary artery systolic pressures. They had extremely poor survival, consistent with our prior experience with isolated PHT. Figure 2 shows the cumulative survival rates for cases and controls from the time of diagnosis of PHT or the matching time in controls. The PHT cases had a 2-year cumulative survival from the time of PHT diagnosis of 50% and a 10% 5-year cumulative survival. Controls had 2-year and 5-year cumulative survival rates of 88% and 80% from their matching time, respectively. None of these SSc patients were treated with epoprostenal, although 3 of them received a lung transplant and have lived an average of 5 years thereafter.

thumbnail image

Figure 2. Cumulative survival rates among pulmonary hypertension (PHT) cases and controls from the time of diagnosis of PHT or the matching time (MT) in controls.

Download figure to PowerPoint

DISCUSSION

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

The CREST syndrome (calcinosis, Raynaud's phenomenon, esophageal dysmotility, sclerodactyly, telangiectasias) was originally described as being a benign form of SSc. The report by Salerni et al that described 10 “CREST” patients who developed severe PHT independent of pulmonary fibrosis led to the realization that this variant of SSc was truly systemic and could be fatal (2). Since the improved survival in scleroderma renal crisis following the advent of angiotensin-converting enzyme inhibitor therapy, isolated PHT is now the complication in SSc that has the worst prognosis.

Pathologic studies on patients with isolated PHT clearly show that this complication is caused by a noninflammatory vasculopathy unrelated to pulmonary interstitial fibrosis (13). In an autopsy case–control morphometry study, we showed that pulmonary vessels in all SSc patient groups, especially those with isolated PHT, had a greater degree of luminal occlusion and intimal proliferation compared with nonscleroderma controls (14). Patients with limited scleroderma who did not have isolated PHT also had an increase in luminal occlusion of the pulmonary vessels compared with nonscleroderma controls. The degree of intimal proliferation tended to correlate with disease duration in patients with limited scleroderma.

We previously studied 30 patients with limited scleroderma and “isolated” PHT and compared them with 287 patients with limited scleroderma who did not have PHT (7). A prolonged duration of Raynaud's phenomenon prior to the development of PHT was found in the PHT patients, but no differences in the frequency of telangiectasias, calcinosis, or anticentromere antibody were identified.

Only 5 of the patients in the Salerni et al study had pulmonary function tests performed, but all 5 of them had a decreased DLCO (2). Subsequently, we showed that almost all PHT patients had a decreased DLCO (mean 39% of predicted in PHT patients versus mean 84% of predicted in the limited scleroderma, no-PHT comparison patients [7]). Sixty-five percent of PHT patients had a DLCO <45% of predicted at the time of diagnosis of PHT. In several patients, the DLCO was decreased for a number of years prior to the diagnosis of PHT. In another study, we found that >15% of the patients with limited scleroderma with an isolated reduction in DLCO progressed to develop PHT 1–9 years later. One patient had serial pulmonary function tests showing a steady fall in DLCO from 70% of predicted to 36% of predicted at the time of PHT diagnosis 10 years later.

Medical therapy for PHT has until recently been unsuccessful. A variety of vasodilators have been studied in the catheterization laboratory (15). High-dose calcium channel blockers given at the time of right heart catheterization were capable of reducing PAPs and pulmonary vascular resistance in primary PHT patients but were not effective when administered orally to scleroderma patients (16). Continuous intravenous infusion of epoprostenol significantly improved the overall well-being and physical function (6-minute walk distance) of patients with scleroderma and PHT (17). Most recently, an endothelin receptor blocker (trade name Bosentan, a well-tolerated oral agent) has been shown in a 3-month study of PHT patients, including some scleroderma patients, to result in a significant improvement in well-being and function (18). The possibility that such therapy for PHT can alter the natural history of SSc-associated PHT lends urgency to the identification of risk factors that predict the subsequent development of PHT. Additionally, treatment of patients who are at significant risk for developing PHT may possibly prevent, delay, or attenuate this severe complication.

In this case–control study of isolated PHT in SSc, a low or decreasing DLCO was the feature that best predicted future development of PHT. Patients with PHT had a significantly lower mean DLCO (52% of predicted) almost 5 years before the diagnosis of PHT, compared with the closely matched controls, whose DLCO was 81% of predicted 6 years prior to the matching time. Smaller numbers of cases and controls had a series of DLCO results. These PHT cases showed a progressive decrease in the DLCO during the years preceding the appearance of clinically obvious PHT. The control DLCO values remained at ∼80% of predicted throughout that time period. The FVC remained close to normal in all of these patients. A low DLCO alone does not make the diagnosis of PHT. The standard echocardiogram is more specific and sensitive than are pulmonary function tests for making an actual diagnosis. An echocardiogram should be performed on any SSc patient with a low or decreasing DLCO. If the estimated pulmonary artery systolic pressure is increased, a right heart catheterization is indicated. If the PAPs on echocardiogram are normal, these patients should be monitored yearly with pulmonary function testing and echocardiograms to identify increased PAPs as soon as they are present.

The anti–U3 RNP antibody, which has a nucleolar immunofluorescence staining pattern on routine antinuclear antibody testing, was significantly increased in the cases, but anti–topoisomerase I was conspicuously absent. Anticentromere antibodies occurred with equal frequency in both PHT cases and controls, and thus did not distinguish patients with limited scleroderma at increased risk for PHT. The pathogenic significance of these antibody associations is unknown, but such strong associations suggest that they are markers of unique subsets of SSc patients. At some point, they may lead us to a better understanding of the pathogenesis of the disease.

There have been several cross-sectional studies reporting varying frequencies of elevated pulmonary artery systolic pressures on echocardiograms in scleroderma patients, but differences in the definition of PHT, the proportion of patients with limited and diffuse scleroderma included, and the frequency of pulmonary interstitial fibrosis make the results difficult to interpret (4, 6). It appears that increased PAPs may be common, but the proportion of patients whose condition progresses to the deadly form of PHT is small. One study of serial echocardiograms from patients with an elevated PAP found that only 20% of these patients showed evolution to severe PHT (4, 19). The only predictive feature in that study was rapidly increasing PAPs. They did not report the findings of the DLCO or FVC.

In the present study, echocardiogram results were not as useful as the DLCO in predicting future development of PHT, although serial studies were not performed. There were no significant differences in the PAPs 4–5 years prior to the diagnosis of PHT in cases or prior to the matching time in controls, but our conclusions may be limited by the smaller number of cases and controls who had echocardiograms performed. Also, the lack of tricuspid regurgitation in some patients can make it difficult to determine the PAP on routine echocardiography. An alternative explanation is that a resting echocardiogram is not sensitive enough to find early PHT. A recent study in familial primary PHT found that all patients with the gene for hereditary PHT had an increased PAP on an exercise echocardiogram, compared with only 10% of those who did not have the gene (5). Exercise echocardiograms in SSc patients should be studied to determine whether they are a better predictive factor than a resting echocardiogram or a DLCO. It seems likely that patients with a very low DLCO will have exercise-induced PHT and may have subclinical PHT. Serial chest radiographs, other cardiac imaging studies, or other parameters on echocardiograms, such as right ventricular or right atrial enlargement, may also be helpful in predicting development of PHT, but these have not been adequately studied.

We found that SSc patients who develop PHT have more severe peripheral vascular disease, based on the increased severity of Raynaud's phenomenon and increased severity of digital tip ulcers compared with the matched, no-PHT controls. An unexpected finding was the infrequent antecedent use of calcium channel blocking agents in PHT patients, suggesting that these drugs might be protective against PHT. This is consistent with a prior study showing improvement in the DLCO with a short course of treatment with a calcium channel blocker (20). Our data are retrospective, but the methods used to ascertain drug use in cases and controls were similar. Nevertheless, long-term prospective data will be necessary to confirm this finding, which has therapeutic implications for the use of calcium channel blockers and for the use of other types of therapy that might alter the peripheral and pulmonary vasculature.

In summary, we have shown that a low and decreasing DLCO is a valuable predictor of the subsequent development of isolated PHT in patients with limited scleroderma. Other features, such as the presence of anti–U3 RNP antibody, the absence of serum anti–Scl-70 antibody, and possibly echocardiogram results, are somewhat helpful. This study showed that the long-term use of calcium channel blocking agents may protect against the development of PHT. Newer agents such as endothelin-1 receptor blockers may also be effective in altering the natural history of this serious complication of SSc.

REFERENCES

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