May 2009  << Back  

  Can't open the PDF? Click here for help.

 

 RESEARCH ARTICLE

Anxiety symptom severity and functional recovery or relapse

Dan J. Stein, MD, PhD

Department of Psychiatry, University of Cape Town, Cape Town, South Africa, Mt. Sinai School of Medicine, New York, NY, USA

Borwin Bandelow, MD

Department of Psychiatry and Psychotherapy, University of Göttingen, Göttingen, Germany

Ornah T. Dolberg, MD

H. Lundbeck A/S Copenhagen, Denmark

Henning F. Andersen, MSc

Novo Nordisk A/S Bagsværd, Denmark

David S. Baldwin, MD

Clinical Neuroscience Division, School of Medicine, University of Southampton, Southampton, United Kingdom

BACKGROUND: Anxiety disorders are associated with significant disability. There is growing interest in the question of whether pharmacotherapy that effectively reduces symptoms can also restore function. Recovery could potentially be defined as a lack of disability, with an associated reduction in symptom severity. Conversely, relapse could potentially be defined in terms of either increased disability or increased symptoms.

METHODS: We analyzed a database of randomized controlled trials of escitalopram in generalized anxiety disorder (GAD) and social anxiety disorder (SAD), focusing on the relationship between disorder-specific severity scales, and the Sheehan Disability Scale (SDS). In short-term studies, cut-points on symptom scales were derived for recovered function. In relapse prevention studies, the effects of defining relapse in terms of increased disability scores were examined.

RESULTS: In GAD and SAD, there is a close correlation between primary symptom severity scales and the SDS, both in the short term and during relapse prevention. Thus, functional recovery is associated with relatively low symptom severity scores, and rates of relapse—defined in terms of increased disability—are significantly lower on escitalopram than on placebo.

CONCLUSIONS: These data indicate that recovery and relapse can potentially be defined either in terms of symptom severity or functioning. Thus, the concept of functional recovery and relapse may be useful in defining treatment outcomes. Longer-term treatment of anxiety disorders is needed to ensure functional recovery.

KEYWORDS: generalized anxiety disorder, social anxiety disorder, Sheehan Disability Scale, recovery

ANNALS OF CLINICAL PSYCHIATRY 2009;21(2):81–88

  INTRODUCTION

The anxiety disorders are the most prevalent of the psychiatric disorders, typically have an early onset and chronic course, and are associated with significant comorbidity.1-3 Not surprisingly, the anxiety disorders are accompanied by significant morbidity, with impaired functioning and reduced quality of life.4 Indeed, the impairment associated with some anxiety disorders is more marked than that seen in many chronic general medical disorders.5 Such impairment is associated not only with individual distress, but also with substantial costs to society.6

Current pharmacological and psychological interventions are effective in reducing the symptoms of the anxiety disorders.7,8 However, although the question of efficacy is important, funding decisions also require consideration of effectiveness and cost efficiency.9 There is, therefore, growing interest in whether these interventions can also reduce the high negative impact of these conditions. It can be hypothesized that short-term treatment can decrease anxiety symptoms (response), but that longer-term treatment is needed for comprehensive improvement in symptoms (remission) and function (recovery).

In this study, we used a database of randomized placebo-controlled clinical trials of escitalopram to examine whether effective treatment of anxiety disorders can restore functioning. Each study used the Sheehan Disability Scale (SDS),10 which enabled us to focus on the association between changes in the primary symptom severity outcome measures in different anxiety disorders and improvement in the SDS.

  METHODS

The randomized placebo-controlled studies of escitalopram included in the current analyses were 2 generalized anxiety disorder (GAD) studies and 3 social anxiety disorder (SAD) studies (TABLE 1). All analyses were based on a modified intent-to-treat (ITT) population (ie, patients with at least 1 dose of study drug and 1 valid post-baseline efficacy assessment).


TABLE 1

Summary of trials included in these analyses

Anxiety disorder Study number Duration, weeks Agent, dose, mg/day Number of patients (women) Mean age, years Reference
GAD 1 12 PBO 139 (93) 42 Baldwin et al, 200617
ESC, 5 134 (78) 41
ESC, 10 136 (91) 42
ESC, 20 133 (92) 41
PAR, 20 139 (84) 42
GAD 2 12 + 24 to 76 PBO 188 (113) 42 Allgulander et al, 200618
ESC, 20 187 (112) 41
SAD 3 12 PBO 177 (83) 36 Kasper et al, 200519
ESC, 10 to 20 181 (80) 39
SAD 4 24 PBO 166 (85) 37 Lader et al, 200420
ESC, 5 167 (83) 36
ESC, 10 167 (96) 37
ESC, 20 170 (90) 37
PAR, 20 169 (91) 37
SAD 5 12 + 24 PBO 181 (89) 38 Montgomery et al, 200521
ESC, 10 to 20 190 (87) 36
ESC: escitalopram; GAD: generalized anxiety disorder; PAR: paroxetine; PBO: placebo; SAD: social anxiety disorder.

Primary symptom severity scales were the Hamilton Anxiety Scale (HAM-A) for GAD (TABLE 2)11 and the Liebowitz Social Anxiety Scale (LSAS) for SAD (TABLE 3).12 The HAM-A comprises 14 items, each rated from 0 to 4, so that the total score has a range of 0 to 56. The LSAS consists of 24 items and each is rated in terms of both fear and avoidance on a 0 to 3 scale, giving a range of 0 to 144. The SDS rates disability in 3 different arenas (work, social, and family), each rated from 0 to 10, so that the total score has a range of 0 to 30. For each subscale, the severity of symptom disruption is rated 0 (not at all), 1 to 3 (mildly), 4 to 6 (moderately), 7 to 9 (markedly), and 10 (extremely).

Spearman’s correlations were used to determine the association between severity scales (HAM-A, LSAS, and Clinical Global Impressions–Severity [CGI-S]) and the SDS, using all post-baseline data. Equipercentile linking was undertaken, examining each point on each SDS subscale and finding the corresponding value on the disorder-specific scale (HAM-A or LSAS), so that the same percentage of patients having poor scores on the disorder-specific scale had poor scores on the SDS subscale, using all post-baseline data.13-15 For example, if 40% of patients had an SDS work subscale score ≤2 and 40% of patients had an LSAS ≤65, then an SDS work subscale score of 2 equates to an LSAS of 65. This is valid only if the 2 scales are correlated.

Kaplan-Meier survival curves were examined in the relapse prevention studies, with relapse defined as an SDS subscale score ≥4.


TABLE 2

Hamilton Anxiety Scale (HAM-A) for generalized anxiety disorder11

  • Anxiety

  • Tension

  • Fears

  • Insomnia

  • Difficulties in concentration and memory

  • Depressed mood

  • General somatic symptoms—muscular

  • General somatic symptoms—sensory

  • Cardiovascular symptoms

  • Respiratory symptoms

  • Gastrointestinal symptoms

  • Genitourinary symptoms

  • Autonomic symptoms

  • Behavior at interview

Each item is rated from 0 to 4, so that the total score has a range of 0 to 56.

TABLE 3

Liebowitz Social Anxiety Scale (LSAS) for social anxiety disorder12

  • Telephoning in public

  • Participating in small groups

  • Eating in public places

  • Drinking with others in public places

  • Talking to people in authority

  • Acting, performing, or giving a talk in front of an audience

  • Going to a party

  • Working while being observed

  • Writing while being observed

  • Calling someone you don’t know very well

  • Talking with people you don’t know very well

  • Meeting strangers

  • Urinating in a public bathroom

  • Entering a room when others are already seated

  • Being the center of attention

  • Speaking up at a meeting

  • Taking a test

  • Expressing disagreement or disapproval to people you don’t know very well

  • Looking people you don’t know very well in the eye

  • Giving a report to a group

  • Trying to make someone’s acquaintance

  • Returning goods to a store

  • Giving a party

  • Resisting a high-pressure salesperson

Each item is rated from 0 to 3 in each of the symptom and avoidance scales, so that the total score has a range from 0 to 144.

  RESULTS

For both GAD and SAD, there was a high correlation between the primary outcome measure (HAM-A and LSAS) and each of the 3 subscales of the SDS in the short-term studies over the entire range of the subscales (FIGURE 1). On the basis of these correlations, scores on symptom severity scales could be equated with those on the SDS (TABLE 4). For example, a score of 3 on the SDS work subscale equated to a HAM-A score of 13 and an LSAS score of 54 (TABLE 2). Similarly, a score of 0 on the SDS work subscale equated to a HAM-A score of 4 and an LSAS score of 15 (TABLE 2).

The total SDS score was also significantly correlated with the CGI-S scores for both GAD and SAD (FIGURE 1C). For GAD, a CGI-S score of 1 (“not at all ill”) corresponded to a total SDS score ≤2, a CGI-S score ≤2 (“borderline mentally ill”) corresponded to a total SDS score ≤7, and a CGI-S score ≤3 (“mildly ill”) corresponded to a total SDS score ≤13. For SAD, a CGI-S score of 1 also corresponded to a total SDS score ≤2, whereas a CGI-S score ≤2 corresponded to a total SDS score ≤5, and a CGI-S score ≤3 corresponded to a total SDS score ≤10.

FIGURE 1 Correlation between primary outcome measure (HAM-A and LSAS) and SDS subscales for GAD and SAD

Equating post-baseline scores of the HAM-A and SDS subscale scores in patients with GAD (raw data from Baldwin et al, 200617). Spearman correlation: 0.68 (work subscale); 0.71 (social subscale); 0.67 (family subscale): P < .001 for each subscale.

FIGURE 1 Correlation between primary outcome measure (HAM-A and LSAS) and SDS subscales for GAD and SAD

Equating post-baseline scores of the LSAS and SDS subscores in patients with SAD (raw data from Lader et al, 200420). Spearman correlation: 0.76 (work subscale); 0.79 (social subscale); 0.64 (family subscale): P < .001 for each subscale.

FIGURE 1 Correlation between primary outcome measure (HAM-A and LSAS) and SDS subscales for GAD and SAD

Equating post-baseline scores of the total SDS with CGI-S scores in patients with either GAD or SAD. Spearman correlation: 0.74 (GAD); 0.79 (SAD): P < .001 for each subscale.

CGI-S: Clinical Global Impressions–Severity; GAD: generalized anxiety disorder; HAM-A: Hamilton Anxiety Scale; LSAS: Liebowitz Social Anxiety Scale; SAD: social anxiety disorder; SDS: Sheehan Disability Scale.

For GAD, 30.8% of patients had a CGI-S score of 1 at the last visit, with 64.6% of these patients (68.9% of escitalopram patients and 59.4% of paroxetine patients) also having a total SDS score ≤2 (data not shown). Similarly, 61.4% of patients had a CGI-S score ≤2 at the last visit, with 83.1% of these patients (84.1% of escitalopram patients and 78.6% of paroxetine patients) also having a total SDS score ≤7. For SAD, 27.0% of patients had a CGI-S score of 1 at the last visit, with 70.5% of these patients (73.0% of escitalopram patients and 63.2% of paroxetine patients) also having a total SDS score ≤2. Similarly, 49.3% of patients had a CGI-S score ≤2 at the last visit, with 79.0% of these patients (81.0% of escitalopram patients and 73.4% of paroxetine patients) also having a total SDS score ≤5.

In the short-term studies, the SDS was administered only at weeks 0, 6, and 12, less frequently than the primary symptom scales. Nevertheless, changes in SDS were already apparent at week 6, with significant correlations with the primary outcome measure at this time point (data not shown) and at last assessment (TABLE 5).

Relapse was defined as an SDS subscale score ≥4, corresponding to a CGI-S score ≥3 (“mildly ill”). In the relapse prevention studies, the correlation between the primary outcome measure and improvement in SDS remained apparent in responders, with further decreases in both symptom scores and further improvements in the SDS (data not shown). Similarly, in relapsed patients, the increase in symptoms and the decrease in function remained highly correlated (data not shown). Thus, when relapse was defined using SDS criteria, rates of relapse were significantly higher with placebo than with medication (FIGURE 2).


TABLE 4

Corresponding Sheehan Disability Scale values for total scores of the standard scales

  SCORE
  0 3 6 9
Arena/SDS subscale
1. Work
HAM-A (GAD) ≤4 ≤13 ≤21 ≤28
LSAS (SAD) ≤15 ≤54 ≤82 ≤117
2. Social
HAM-A (GAD) ≤4 ≤13 ≤21 ≤33
LSAS (SAD) ≤15 ≤55 ≤82 ≤121
3. Family
HAM-A (GAD) ≤5 ≤15 ≤22 ≤33
LSAS (SAD) ≤35 ≤76 ≤102 ≤133
GAD: generalized anxiety disorder; HAM-A: Hamilton Anxiety Scale; LSAS: Liebowitz Social Anxiety Scale; SAD: social anxiety disorder; SDS: Sheehan Disability Scale.
Each of the 3 arenas is rated from 0 to 10, so that the total score has a range of 0 to 30. For each subscale, the severity of symptom disruption is rated as follows: 0=not at all; 1 to 3=mildly; 4 to 6=moderately; 7 to 9=markedly; and 10=extremely.

TABLE 5

Sheehan Disability Scale subscale scores in social anxiety disorder and generalized anxiety disorder

    Baseline, Mean ± SD Last assessment, 12 weeks, Mean ± SD
Study Agent, dose, mg/day Work Social Family Work Social Family
Kasper et al, 200519 PBO 6.7 ± 1.9 7.0 ± 1.8 4.7 ± 2.6 4.97 ± 2.65 4.93 ± 2.67 3.09 ± 2.36
  ESC 6.9 ± 2.2 6.8 ± 1.9 4.2 ± 2.4 4.38 ± 2.74 4.22 ± 2.65 2.76 ± 2.30
Montgomery et al, 200521 ESC OL 6.78 ± 2.05 7.32 ± 1.6 5.01 ± 2.47 3.31 ± 2.34 3.28 ± 2.23 2.26 ± 2.19
Lader et al, 200420 PBO 6.63 ± 1.90 7.33 ± 1.64 4.50 ± 2.54 4.58 ± 2.70 4.65 ± 2.72 2.24 ± 2.26
  ESC, 5 7.04 ± 1.68 7.19 ± 1.85 4.30 ± 2.56 3.49 ± 2.81 3.46 ± 2.66 1.58 ± 1.79
  ESC, 10 6.93 ± 1.93 6.93 ± 1.71 4.07 ± 2.53 3.79 ± 2.81 3.60 ± 2.81 1.27 ± 1.64
  ESC, 20 6.78 ± 1.93 7.25 ± 1.60 4.63 ± 2.35 3.10 ± 2.69 3.13 ± 2.73 1.29 ± 1.86
  PAR, 20 7.05 ± 1.73 7.08 ± 1.65 4.57 ± 2.26 3.36 ± 2.85 3.56 ± 2.84 1.93 ± 2.17
Allgulander et al, 200618 ESC OL 5.98 ± 2.13 5.98 ± 2.15 5.58 ± 2.14 2.39 ± 2.53 2.32 ± 2.44 2.18 ± 2.36
Baldwin et al, 200617 PBO 6.03 ± 2.36 6.11 ± 2.06 5.36 ± 2.17 3.35 ± 2.76 3.30 ± 2.67 3.01 ± 2.42
  ESC, 5 6.20 ± 2.09 5.98 ± 1.85 5.47 ± 1.95 3.10 ± 2.68 2.77 ± 2.47 2.58 ± 2.35
  ESC, 10 5.86 ± 2.34 5.66 ± 2.30 5.42 ± 2.27 2.63 ± 2.55 2.31 ± 2.57 2.23 ± 2.49
  ESC, 20 6.05 ± 2.25 6.14 ± 2.23 5.61 ± 2.29 2.84 ± 2.73 2.71 ± 2.71 2.48 ± 2.57
  PAR, 40 5.85 ± 2.26 6.18 ± 2.14 5.64 ± 2.14 3.00 ± 2.63 3.07 ± 2.75 2.60 ± 2.45
ESC: escitalopram; OL: open label; PAR: paroxetine; PBO: placebo.

FIGURE 2 Kaplan-Meier estimates of relapse in SAD

Kaplan-Meier estimates of relapse in SAD (raw data from Montgomery et al, 200521). Relapse defined as an SDS work subscale score ≥4 for patients with a score of <4 at randomization.

FIGURE 2 Kaplan-Meier estimates of relapse in SAD

Kaplan-Meier estimates of relapse in SAD (raw data from Montgomery et al, 200521). Relapse defined as an SDS social subscale score ≥4 for patients with a score of <4 at randomization.

FIGURE 2 Kaplan-Meier estimates of relapse in SAD

Kaplan-Meier estimates of relapse in SAD (raw data from Montgomery et al, 200521). Relapse defined as an SDS family subscale score ≥4 for patients with a score of <4 at randomization.

SAD: social anxiety disorder; SDS: Sheehan Disability Scale.

  DISCUSSION

The short-term data on escitalopram in GAD and SAD emphasize that improvement in primary symptom scales in anxiety disorders is accompanied by a complementary improvement in functioning. Although it might be expected that certain SDS subscales (eg, work) could take longer to show signs of improvement, the data here emphasize the direct negative impact of symptoms on a range of different aspects of function (work, family, social) and the relatively early effects of treatment on decreasing symptom severity and reducing functional impairment.

This finding has important implications for the treatment of anxiety disorders. Although patients often present in the hope of reducing symptoms, in other cases a primary motivation for treatment is associated functional impairment. Evidence that symptom decrease and functional impairment are often correlated provides an incentive for initiating pharmacotherapy and for remaining adherent to medication over time. Improvements in both symptom severity and functional impairment can be expected even early on.

Because there is a direct relationship between improvement in symptoms and function, it is possible to derive symptom scale cut-points for functional recovery. It can be expected that the range of impairment across the SDS subscales would be somewhat different in each of the individual anxiety disorders; indeed, symptom scale cut-points for functional recovery in SAD patients indicate that absence of work impairment requires the LSAS to be ≤15, but absence of family impairment requires the LSAS to be only ≤35.

This point illustrates how constructs of disability and functional recovery may usefully augment symptom measures. Patients are usually classified as “responders” if their scores improve during the trial period by 50% or more, and “remitters” if their final score is under a certain cut-point. This finding has been examined not only with the SDS but also with other scales. Bandelow and colleagues examined the correlation of various disorder-specific rating scales with global outcome measures (Clinical Global Impression–Severity and Clinical Global Impression– Improvement) and found a similar direct correlation, suggesting cut-points for remission and complete remission on the global outcome scale that correlate with the original disorder-specific rating scale.16

It is notable that anxiety symptoms continued to decrease and function to improve during long-term treatment. Although relapse is typically defined in terms of symptom severity in relapse prevention trials, the data here demonstrate good separation of medication and placebo when relapse is defined in terms of functional impairment. There is growing emphasis on the value of symptom remission during treatment of anxiety disorders17; it may be valuable to augment this construct with the notions of functional recovery and relapse. Longer-term treatment of anxiety disorders may be useful in order to achieve full functional recovery.

DISCLOSURES: Dr. Stein is supported by the Medical Research Council of South Africa and receives research grants and/or consultancy honoraria from AstraZeneca, Eli Lilly and Company, GlaxoSmithKline, Johnson and Johnson, H. Lundbeck A/S, Orion, Pfizer Inc, Pharmacia, Roche, Servier, Solvay, Sumitomo, Tikvah, and Wyeth. Dr. Bandelow is a consultant to AstraZeneca, Jazz, Eli Lilly and Company, H. Lundbeck A/S, Pfizer Inc, Solvay, Wyeth Pharmaceuticals, and Xian-Janssen; is on the consultant/advisory board for AstraZeneca, Cephalon, Eli Lilly and Company, H. Lundbeck A/S, Pfizer Inc, Roche, Servier, and Wyeth Pharmaceuticals; is a speaker for AstraZeneca, Bristol-Myers Squibb, Danippon Sumitomo Pharma, Janssen-Cilag, Eli Lilly and Company, H. Lundbeck A/S, Pfizer Inc, Solvay, and Wyeth Pharmaceuticals. Dr. Dolberg is an employee of H. Lundbeck A/S. Drs. Stein, Bandelow, and Baldwin have received occasional consultancy honoraria from H. Lundbeck A/S. Mr. Andersen is a former employee of H. Lundbeck A/S and owns H. Lundbeck A/S bonds. Mr. Baldwin receives grant/research support from H. Lundbeck A/S and is a consultant and speaker for H. Lundbeck A/S, Eli Lilly and Company, Servier, and Pfizer Inc.

ACKNOWLEDGMENTS: The studies on which this article is based were sponsored by H. Lundbeck A/S, which was involved in the design of the clinical trials and the collection of the data.

    REFERENCES

  1. Kessler RC, Berglund P, Demler O, et al. Lifetime prevalence and age-of-onset distributions of DSM-IV disorders in the National Comorbidity Survey Replication. Arch Gen Psychiatry. 2005;62:593–602.
  2. Kessler RC, Chiu WT, Demler O, et al. Prevalence, severity, and comorbidity of 12-month DSM-IV disorders in the National Comorbidity Survey Replication. Arch Gen Psychiatry. 2005;62:617–627.
  3. Kessler RC, Demler O, Frank RG, et al. Prevalence and treatment of mental disorders, 1990 to 2003. N Engl J Med. 2005;352:2515–2523.
  4. Mogotsi M, Kaminer D, Stein DJ. Quality of life in the anxiety disorders. Harv Rev Psychiatry. 2000;8:273–282.
  5. Kessler RC. The epidemiology of pure and comorbid generalized anxiety disorder: a review and evaluation of recent research. Acta Psychiatr Scand Suppl. 2000;(406):7–13.
  6. Greenberg PE, Sisitsky T, Kessler RC, et al. The economic burden of the anxiety disorders in the 1990s. J Clin Psychiatry. 1999;60:427–435.
  7. Stein DJ, Hollander E. The American Psychiatric Publishing Textbook of Anxiety Disorders. Washington, DC: American Psychiatric Publishing, Inc.; 2002.
  8. Baldwin DS, Anderson IM, Nutt DJ, et al. Evidence-based guidelines for the pharmacological treatment of anxiety disorders: recommendations from the British Association for Psychopharmacology. J Psychopharmacol. 2005;19:567–596.
  9. Chisholm D, Sanderson K, Ayuso-Mateos JL, et al. Reducing the global burden of depression: population-level analysis of intervention cost-effectiveness in 14 world regions. Br J Psychiatry. 2004;164:393–403.
  10. Sheehan DV, Harnett-Sheehan K, Raj BA. The measurement of disability. Int Clin Psychopharmacol. 1996;11(suppl 3):89–95.
  11. Hamilton M. The assessment of anxiety states by rating. Br J Med Psychol. 1959;32:50–55.
  12. Liebowitz MR. Social phobia. Mod Probl Pharmacopsychiatry. 1987;22:141–173.
  13. Leucht S, Kane JM, Kissling W, et al. Clinical implications of Brief Psychiatric Rating Scale scores. Br J Psychiatry. 2005;187:366–371.
  14. Leucht S, Engel RR. The relative sensitivity of the Clinical Global Impressions Scale and the Brief Psychiatric Rating Scale in antipsychotic drug trials. Neuropsychopharmacology. 2006;31:406–412.
  15. Leucht S, Kane JM, Etschel E, et al. Linking the PANSS, BPRS, and CGI: clinical implications. Neuropsychopharmacology. 2006;31:2318–2325.
  16. Bandelow B, Baldwin DS, Dolberg OT, et al. What is the threshold for symptomatic response and remission for major depressive disorder, panic disorder, social anxiety disorder, and generalized anxiety disorder? J Clin Psychiatry. 2006;67:1428–1434.
  17. Baldwin DS, Huusom AK, Mæhlum E. A placebo-controlled, double-blind study of escitalopram and paroxetine in the short-term treatment of generalised anxiety disorder (GAD). Br J Psychiatry. 2006;189:264–272.
  18. Allgulander C, Florea I, Huusom AK. Prevention of relapse in generalized anxiety disorder by escitalopram treatment. Int J Neuropsychopharmacol. 2006;9:495–505.
  19. Kasper S, Stein D, Loft H, et al. Escitalopram in the treatment of social anxiety disorder: randomised, placebo-controlled, flexible-dosage study. Br J Psychiatry. 2005;186:222–226.
  20. Lader M, Stender K, Bürger V, et al. Efficacy and tolerability of escitalopram in 12- and 24-week treatment of social anxiety disorder: randomised, double-blind, placebo-controlled, fixed-dose study. Depress Anxiety. 2004;19:241–248.
  21. Montgomery SA, Nil R, Dürr-Pal N, et al. A 24-week randomized, double-blind, placebo-controlled study of escitalopram for the prevention of generalized social anxiety disorder. J Clin Psychiatry. 2005;66:1270–1278.

CORRESPONDENCE Dan J. Stein, MD, PhD, Department of Psychiatry, University of Cape Town, Groote Schuur Hospital (J-2), Anzio Rd., Observatory 7925, Cape Town, South Africa. E-MAIL: dan.stein@uct.ac.za