SEARCH

SEARCH BY CITATION

Keywords:

  • Juvenile-onset ankylosing spondylitis;
  • Adult-onset ankylosing spondylitis;
  • Functional outcomes

Abstract

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

Objective

To compare functional outcome of patients with juvenile-onset ankylosing spondylitis (JoAS; defined as AS with symptom onset before 16 years of age) with that of patients with adult-onset AS (AoAS) and to identify variables associated with a poor functional outcome of JoAS.

Methods

A cross-sectional study was performed of 326 JoAS patients who participated in a postal survey conducted by the Spondylitis Association of America. This cohort was compared with 2,021 AoAS patients who participated in the same survey. Simple and multiple logistic regression analyses were performed to identify differences with respect to clinical features, demographic features, and functional outcome (defined by the Bath Ankylosing Spondylitis Functional Index [BASFI]) between the 2 groups. A validation cohort of 255 AS patients was also surveyed.

Results

The mean ± SD BASFI score (controlled for disease duration) for JoAS was 51.3 ± 1.5 compared with 46.4 ± 0.6 for AoAS (P < 0.0001). Multiple logistic regression identified only age (P < 0.0001) and income status (P < 0.0001) as factors associated with functional impairment.

Conclusion

It appears that JoAS is a progressive disease and is associated with significant delay in diagnosis and worse functional outcome compared with AoAS. Furthermore, women do worse than men with JoAS. This would argue for the importance of early diagnosis and treatment of AS, particularly in the subgroup of patients with JoAS.


INTRODUCTION

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

Ankylosing spondylitis (AS) is a chronic inflammatory rheumatic disease of the spine that affects 0.2–0.8% of the population (1). Although AS typically presents in the late teens or early twenties, it can present in childhood. When symptom onset occurs in individuals <16 years of age and individuals go on to develop radiographic sacroiliitis at a later stage, the disease is termed juvenile-onset AS (JoAS) (2). Among patients with AS, prevalence rates for juvenile onset vary from 9% to 21% in white populations, although prevalence rates ≥40% have been reported with Mexican Mestizo and Korean AS patients (3).

Although adult-onset AS (AoAS) and JoAS share many common features, both being characterized by the presence of radiographic sacroiliitis, they also differ in many respects. JoAS presents more commonly with peripheral joint symptoms and adults are more likely to present with axial manifestations. Differences in functional outcome have also been reported that depend on the age of onset. In a study comparing 24 JoAS with 71 AoAS patients, JoAS had worse functional outcome (4). In another study comparing functional outcome in 22 JoAS with 22 AoAS patients, no difference was observed (5). There are several reasons for the discrepancies reported between the studies, such as small sample size and the lack of a common definition of outcomes, making direct comparison between studies difficult. In addition, several phenotypic manifestations of JoAS exist that may account for differences in determining functional outcomes. Burgos-Vargas et al described a characteristic group of JoAS patients with “genuine adult-like AS” as opposed to patients who present with the syndrome of enthesopathic arthritis and evolve subsequently into AS (3).

No comparison of functional outcome of JoAS with AoAS has been performed to date in a North American cohort. The potential for biologic agents to improve functional outcome in AS has recently been demonstrated in several short-term randomized controlled trials (6–8). For this reason and because functional outcome has been shown to be the most important predictor of total costs in AS (9), it is important to determine why some patients have a poor functional outcome. The objective of this study was to compare functional outcome of JoAS with AoAS and to identify predictors of a poor functional outcome of JoAS.

PATIENTS AND METHODS

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

Study population.

A survey was mailed to subjects registered in the Spondylitis Association of America (SAA) database. The database consisted of 1,500 current members of the SAA, 2,460 lapsed members, and 3,600 individuals who had made enquiries to the SAA after 1992. The questionnaire was administered by Harris Interactive from July through September 2002. In all, there were 30 questions relating to diagnosis, pattern of disease, treatment, effect of AS on quality of life, and demographic information, such as age, sex, education, and socioeconomic status. The questionnaire took ∼30 minutes to complete. Potential respondents were sent a letter a few days prior to receiving the survey introducing the survey in the hope that they would be more likely to respond if they were informed about the study and its aims in advance.

A validation cohort was also surveyed and consisted of 255 AS patients with diagnosis confirmed by the treating physician. The validation cohort was recruited from the practices of 44 physicians who were approached and asked to request the participation of AS patients in their practice. This group was completely separate from the survey sample. The study was approved by the local ethics review committee.

Patients with disease onset at <16 years of age were classified as having JoAS and those with symptom onset >16 years of age were classified as having AoAS. Therefore, these disorders are not clinically distinct and differ only in the age of onset of symptoms.

Primary outcome measure.

The Bath Ankylosing Spondylitis Functional Index (BASFI) was used to measure functional impairment (10). This instrument comprises 8 questions relating to the functional anatomy of AS patients and 2 additional questions that assess the AS patient's ability to cope with everyday life. Each question is answered on a 100-mm visual analog scale (0 = easy, 100 = impossible). The mean of the 10 scales gives the BASFI score. This tool has also been validated in a cohort of 163 AS patients. It takes on average 47 seconds to complete. When distinguishing between patients with high and low disease severity, it has been shown to have a sensitivity of 94% and specificity of 87% (10). The BASFI correlates with other external criterion for disease activity, such as patient and physician global assessment of disease activity and function. Severe functional impairment was defined as a BASFI score >50.

Secondary outcome measures.

Individuals were considered to have AS-related work disability if they were ever on either temporary or permanent work disability resulting from AS. Self-perceived spinal deformity was assessed to obtain information about AS disease severity and damage. Patients were asked to what extent their AS caused them to stoop over. Patients were shown 4 graphics representing 0°, <25°, 25–45°, and >45° of deformity and asked to choose the one that best represented their posture (Figure 1). This instrument was developed for the purpose of this survey, to ensure it met the construct of face validity for the assessment of self-perceived spinal deformity. A focus group of individuals, some of whom had AS, met to plan the development of the instrument and ensure it met the principles of truth and feasibility as outlined by the Outcome Measures in Rheumatology Clinical Trials filter. Subsequent to this, an AS patient, a methodologist, and several lay individuals who work at the SAA met with the graphic design team at Harris Interactive to finalize the graphics for the instrument.

thumbnail image

Figure 1. Bruckel instrument for patient self assessment of spinal deformity. This was a tool developed for the purpose of this study to assess the degree of spinal deformity by patient self report. Four cartoon figures denote various stages of advanced spinal deformity. Study patients were asked to select the cartoon figure that they felt best represented the degree of spinal deformity they perceived they had. The instrument was named after Jane Bruckel, cofounder of the Spondylitis Association of America.

Download figure to PowerPoint

Statistical analysis.

Comparisons between the validation cohort and total cohort of JoAS and AoAS patients for demographic and clinical characteristics were performed using descriptive statistics and parametric and nonparametric tests, as appropriate. To provide some correction to the skew in the data, time since diagnosis (i.e., disease duration) was truncated at 35 years, delay in diagnosis was truncated at 50 years, and the minimum for age of first symptoms (JoAS) was set at 9 years.

Simple and multiple logistic regression analyses were performed to identify significant variables associated with functional impairment. In the univariate analysis, the mean BASFI for JoAS and AoAS patients was compared and adjustments were made for disease duration using a two–way analysis of variance test. Functional impairment was the dependant variable and was considered a continuous variable in this analysis. Thus, the influence on functional impairment of age, disease duration, delay in diagnosis, education, household income, sex, and age at onset was statistically assessed. Variables with P values < 0.05 in the simple logistic regression analyses were entered to the multiple regression analysis. All statistical analyses were performed using SAS version 8.02 (SAS Institute, Cary, NC).

RESULTS

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

Response to survey.

Questionnaires were mailed to 7,500 individuals on the SAA database, of which 545 could not be delivered and were returned by the post office. Of the 6,955 individuals who received the questionnaire, 2,384 responded. There was broad representation from all the States in the United States, with a mean number of 53 patients recruited per state and a range of 4–410. The percentage response rate by category was as follows: for current members, 877 of 1,498 (59%); for lapsed members, 678 of 2,174 (31%); and for those who inquired after 1992, 684 of 3,600 (19%). Among the validation cohort, 194 of 245 (75%) individuals responded.

Validation cohort.

The validation cohort was similar to the remainder of the cohort with respect to age, age at disease onset, sex, ethnicity, delay in diagnosis, and functional impairment due to AS (data not shown).

Demographic features.

There were 326 individuals who reported symptom onset before 16 years of age and 2,021 who had onset after 16 years of age. The diagnosis of AS was made by a rheumatologist in 63% of individuals, by a generalist in 15%, by an orthopedic surgeon in 13%, an ophthalmologist in 5.4%, and a chiropractor in 4.5%. There was no difference in sex distribution between the groups; 60% were male. The majority of patients (94%) sampled were white. However, a significantly greater proportion of AoAS patients (95%) were white as compared with JoAS patients (89%) (Table 1). JoAS patients were younger at the time of the survey (current mean age 46.5 ± 0.76 versus 50.9 ± 0.27 years; P < 0.001), were better educated (P = 0.003), less likely to be married or in a stable relationship (P < 0.001), and less likely to be from a high income category (P = 0.0002) when compared with AoAS patients at the time of the survey (Table 1).

Table 1. Comparison of demographic characteristics of juvenile- versus adult-onset disease
VariableJuveniles (n = 326)Adults (n = 2,021)P
Women, %39.139.70.838
Age, mean ± SEM years46.5 ± 0.7650.9 ± 0.27< 0.001
Marital status, %  < 0.001
 Married60.474.6 
 Single22.410.7 
 Divorced/widow12.612.8 
 Living with partner4.61.9 
Education, %  0.003
 High school or less10.517.2 
 College or more89.582.8 
Race, %  < 0.001
 White89.394.6 
 Other10.75.4 
High income, %88.693.80.002

Clinical characteristics.

JoAS patients were more likely to present with peripheral joint symptoms (46.6% versus 33.2%; P < 0.001) (Table 2). A greater percentage of AoAS patients presented with axial symptoms, low back pain, and stiffness (71.5% versus 66.0%; P = 0.043). There was no difference between the 2 groups in the frequency of enthesopathic manifestations, iritis, or neck pain at presentation. At the time of survey, JoAS patients had a mean disease duration of 18.3 ± 0.70 years compared with 13.4 ± 0.23 years for AoAS patients (P < 0.001). JoAS patients had a significantly longer delay in diagnosis compared with patients with AoAS (15.3 ± 0.79 versus 7.6 ± 0.2 years; P < 0.001).

Table 2. Comparison of clinical characteristics for juvenile- versus adult-onset disease*
VariableJuveniles (n = 326)Adults (n = 2,021)P
  • *

    Except where otherwise indicated, the values are mean ± SEM. BASFI = Bath Ankylosing Spondylitis Functional Index; min = minutes.

  • Adjusted for disease duration.

  • Self-perceived spinal deformity.

Age at symptom onset, years13.6 ± 0.1430.1 ± 0.25< 0.001
Delay in diagnosis, years15.3 ± 0.797.6 ± 0.20< 0.001
Disease duration, years18.3 ± 0.7013.4 ± 0.23< 0.001
BASFI (0–100)51.3 ± 1.546.4 ± 0.57< 0.001
 Putting on socks/tights3.7 ± 0.163.2 ± 0.060.013
 Bending to pick up pen5.0 ± 0.184.3 ± 0.07< 0.001
 Reaching to high shelf4.2 ± 0.173.7 ± 0.070.004
 Getting out of chair4.7 ± 0.184.4 ± 0.070.067
 Off floor from lying on back5.6 ± 0.185.4 ± 0.070.171
 Standing unsupported for 10 min5.2 ± 0.194.7 ± 0.070.016
 Climbing 12–15 steps4.5 ± 0.194.2 ± 0.070.143
 Looking over shoulder6.8 ± 0.186.2 ± 0.080.004
 Physically demanding activities6.0 ± 0.155.4 ± 0.06< 0.001
 Full day's activities5.5 ± 0.165.1 ± 0.070.001
AS-related work disability, %27.420.90.012
Spinal deformity, %  < 0.001
 Not at all26.234.8 
 <25°53.852.7 
 25°–45°17.210.8 
 >45°2.81.6 
Presenting symptoms, %   
 Neck pain/stiffness31.333.90.346
 Back pain/stiffness66.071.50.043
 Stiffness in joints46.633.2< 0.001
 Heel pain13.212.20.622
 Sternum pain16.916.50.877
 Iritis13.815.30.473
 Other32.221.8< 0.001

Disability and functional impairment.

Among JoAS patients, 27.4% reported AS-related work disability, compared with 20.9% of individuals with AoAS (P = 0.012) (Table 2). Among JoAS patients, 17.2% reported spinal deformity (25°–45°) compared with 10.8% of AoAS patients. Only 26.2% of JoAS patients and 34.8% of AoAS patients reported no spinal deformity.

The distribution of the BASFI scores for the JoAS patients was reasonably normal. The mean BASFI for JoAS was 51.3 ± 1.5 with an interquartile range of 72–30 compared with 46.4 ± 0.57 for AoAS patients with an interquartile range of 66–25 (Table 2). This held true after adjustment for disease duration. This adjustment was made by running a 2-way analysis of variance (JoAS/AoAs and disease duration, BASFI as the dependant variable) with interaction. Both disease group and disease duration were significant (P < 0.0001 and P = 0.028, respectively), but the interaction term was not (P = 0.814). Thus, JoAS patients had significantly greater functional impairment after controlling for disease duration as defined with these groups (Figure 2). In addition, there was a significant effect of disease duration as defined with these groups. However, the effect of one did not depend on the other.

thumbnail image

Figure 2. Comparison of functional impairment (defined by the Bath Ankylosing Spondylitis Functional Index [BASFI]) versus disease duration stratified by age at disease onset. Juvenile-onset ankylosing spondylitis (JoAS) patients had significantly greater functional impairment after controlling for disease duration as compared with adult-onset ankylosing spondylitis (AoAS) patients. This adjustment for disease duration was made by running a two-way analysis of variance (JoAS/AoAs and disease duration, BASFI as the dependant variable) with interaction. Both disease group and disease duration were significant (P < 0.0001 and P = 0.028, respectively) but the interaction term was not (P = 0.814). In addition, there was a significant effect of disease duration as defined with these groups. However, the effect of one did not depend on the other.

Download figure to PowerPoint

Among JoAS patients, 171 were classified as having severe functional impairment and 160 did not meet the criteria for severe functional impairment. There was no difference in sex distribution, marital status, education, race, or age at disease onset between JoAS patients with severe functional impairment and those without. However, the cumulative percentage of JoAS patients with severe functional impairment appeared to increase with longer disease duration (Figure 3A). Furthermore, women with JoAS were more likely to have severe functional impairment, as defined by a BASFI score >50, earlier in the disease course as compared with men (Figure 3B). JoAS patients with severe functional impairment were also noted to have a greater delay in diagnosis, 16.7 ± 0.9 versus 11.9 ± 0.8 years as compared with JoAS patients without severe functional impairment (data not shown). There appeared to be a linear association between delay in diagnosis and severe functional impairment in JoAS (Figure 3C).

thumbnail image

Figure 3. A, Cumulative percentage of juvenile-onset ankylosing spondylitis (JoAS) patients with severe functional impairment (defined as Bath Ankylosing Spondylitis Functional Index [BASFI] score >50) plotted against disease duration represented in years. B, Cumulative percentage of JoAS patients with severe functional impairment (defined as BASFI score >50) plotted against disease duration represented in years. JoAS patients are stratified by sex (thin line = men, thick line = women).C, Cumulative percentage of JoAS patients with severe functional impairment (defined as BASFI score >50) plotted against delay in diagnosis represented in years.

Download figure to PowerPoint

Multiple logistic regression.

Simple logistic regression analyses indicated age, delay in diagnosis, disease duration, and income status as having a significant effect on functional impairment (Table 3). These variables were then included in the multiple regression analysis, which identified only age (P = 0.046) and income status (P < 0.001) as factors associated with functional impairment. Both disease duration and delay in diagnosis were highly correlated with age, explaining why they were not significantly associated with the outcome in the multiple regression analysis.

Table 3. Predictors of functional impairment in juvenile-onset ankylosing spondylitis by univariate and multiple regression analysis*
VariableUnivariate modelPMultiple modelP
  • *

    Disease duration was truncated at 35 years, delay in diagnosis truncated at 35 years, and the minimum age of first symptoms was set at 9 years.

Age, yearsb = 0.633; SE(b) = 0.100< 0.001b = 1.187; SE(b) = 0.5910.046
High incomeb = −28.716; SE(b) = 4.762< 0.001b = −27.606; SE(b) = 4.463< 0.001
Delay in diagnosis, yearsb = 0.458; SE(b) = 0.113< 0.001  
Disease duration, yearsb = 0.382; SE(b) = 0.1310.004  

DISCUSSION

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

This is the first North American study of AS patients that attempts to compare functional outcome in JoAS patients with that in AoAS patients, and is the largest published series to do so. In this study, we found that JoAS patients experience a greater delay in diagnosis compared with AoAS patients and suffered more functional impairment compared with AoAS patients with the latter holding even when controlled for disease duration. Women with JoAS appear to develop functional impairment earlier than men with JoAS. Multivariate analysis demonstrated that age and low income status were strongly associated with worse functional outcome in JoAS compared with AoAS.

One strength of this study lies in the large number of JoAS patients (n = 326) who participated in the survey, the largest reported in the literature to date. Because this survey targeted patients in the population, it is likely that this cohort represents a spectrum of AS disease severity compared with patients recruited only from a tertiary referral center. A unique aspect of the study was that there was representation from all states in the US, although the greatest numbers came from California and New York. Despite the large number of respondents, it is acknowledged that the overall response rate to the survey was 30%. This may be due to the fact that a large number of patients who were mailed questionnaires from the SAA database were those who had only made an inquiry to the SAA. On the other hand, the response rate from current members of the SAA was high, at 78%. It is possible that nonresponders included a high percentage of patients who in fact had never been diagnosed with AS, particularly since the response rate for SAA inquiries was only 19%.

The large sample size facilitated a multivariate analysis to identify factors associated with outcome in JoAS. This was not possible in previous studies because of smaller sample sizes (4, 5). However, a limitation of a cross-sectional study design is that it precludes drawing any definitive conclusion about prognostic indicators of outcome. Therefore we were only able to identify strongly associated variables that may be predictors of outcome. A prospective observational cohort study with patients enrolled at inception would be required to confirm if the identified variables are true predictors of outcome. To date, a large epidemiologic cohort of AS patients has not been compiled in North America. The low incidence of the disease and the prolonged delay in diagnosis pose logistic problems for conducting such a study.

In this study, the diagnosis of AS was made by self report and this does have limitations with regard to misclassification of cases. It is possible that this would bias the results by including some patients who did not have the disease. The bias would, however, most likely be the same for JoAS and AoAS. We did ask patients which health professional made the diagnosis and there was no difference between groups. Almost 90% of individuals reported that the diagnosis was made by a physician. In addition, there were no differences between the 2 groups and the validation cohort where the diagnosis was confirmed by a physician. This was the reason we decided a validation cohort was important to include in the study design.

We acknowledge that questioning patients about symptom onset may be fraught by recall bias. However we feel that the recall bias in this instance may have lead to an underestimation of the true disease duration for many AS patients. Patients are more likely to remember episodes of disease activity and forget good periods. Therefore we may have biased the results toward underestimation of disease duration or indeed delay in diagnosis. Prospective followup of patients from date of onset of symptoms is the only means of avoiding recall bias.

No recent studies have been performed comparing JoAS to AoAS. The existing publications are older studies, each reporting on a small number of patients. The management of AS has changed markedly in recent years, with biologic therapies increasingly prescribed for many patients, which makes comparison of patient outcomes between our study and these older studies difficult. In addition, several methodologic differences further limit direct comparisons between studies.

In one study, a survey was conducted of 3,000 AS patients of whom 135 had JoAS (11). The results were compared with 135 respondents who had AoAS. Functional impairment was measured using the Arthritis Impact Measurement Score (AIMS) (12), which may not be as sensitive as the BASFI for detecting disease-specific functional impairment in AS. The authors reported that functional impairment, disability, and impact of disease were less marked in JoAS compared with AoAS. However, the JoAS patients had a mean age of 37.3 years at the time of sampling compared with our cohort, which had a mean age of 46.5 years at time of survey. In our study, age was identified as being significantly associated with functional outcome. Therefore, this difference in age at the time of sampling between studies may account for the discrepancy in reported outcomes between studies.

In another study by the same group, 22 JoAS patients were compared with 22 AoAS patients (5). There were some similarities with our study; delay in diagnosis was greater in JoAS compared with AoAS, with women experiencing the longest delay in diagnosis. Patient numbers did not allow an analysis stratified by sex or a multivariate analysis. The results of this study may have been limited by referral bias, as patients with milder disease may not have been included because this study was based on patients recruited from a tertiary referral center.

Garcia-Morteo et al (4) addressed the comparison of the 2 clinical entities in a retrospective study. The authors concluded that functional impairment was more severe in the JoAS group. Nineteen percent (n = 24) of the patients in this study had JoAS, as compared with 14% of our cohort. Seventy-one patients with AoAS were included in the study. Of these, only 46 patients were evaluated for functional outcome. In the study, all 24 patients with JoAS exhibited some degree of functional impairment, as defined by the American College of Rheumatology functional classification system, after mean disease duration of 15 years. In comparison, 80% of patients with AoAS were classified as having functional impairment. The methodology of the study was not adequately outlined to determine why the remaining individuals were not reported on for this outcome.

Our study had some important messages for clinicians. There persists a significant delay in diagnosis that is more marked for JoAS than AoAS, and women had an even greater delay. Furthermore, delay in diagnosis is associated with worse functional outcomes in JoAS patients. The magnitude of the difference between the 2 groups in functional outcomes that we noted was small. To date there have been no studies performed to determine the functional significance of small differences in the BASFI scores. The BASFI total score is a mean of 10 subscales. To better understand the differences between groups in functional outcome, we compared the difference in the mean of each of these subscales between JoAS and AoAS patients. For each of these subscales, patients with JoAS scored higher than AoAS patients. The data presented in Table 2 indicates that those subscales that relate to lumbar spine and cervical spine movements had the largest differences between groups. Thus the clinical relevance of this may be that patients with JoAS have more severe axial symptoms than those with AoAS.

This observation is particularly important now, given that biologics are emerging as effective treatments for AS (13). Earlier diagnosis and treatment may improve long-term functional outcomes, but this awaits formal proof in prospective therapeutic trials with longer followup. There is a need for a higher index of suspicion among physicians so that patients with JoAS are diagnosed as early as possible. With the emerging role of magnetic resonance imaging as a diagnostic tool in AS, this may also facilitate a more prompt diagnosis.

There remain many unanswered questions that need to be addressed in future research efforts. It will be difficult to study the natural history of AS in the future as treatment with tumor necrosis factor blocking agents becomes more widespread. Therefore, identification of clinical prognostic markers will also be difficult. However, research efforts should focus on stratifying patients into prognostic categories so that therapy is initiated not only sooner, but also appropriately for each patient. Based on our results, it appears that JoAS is a progressive disease and is associated with worse functional outcome compared with AoAS. This would argue for a widespread campaign to educate physicians about the importance of early diagnosis and treatment of patients with JoAS. However, it remains unknown if treatment of patients early in the disease will prevent disease progression, and future studies should address this issue.

Acknowledgements

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

The authors would like to acknowledge the assistance of Sandy Appelbaum of Harris Interactive.

REFERENCES

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