Article Text

PDF

Variability in the biological response to anti-CD20 B cell depletion in systemic lupus erythaematosus
  1. D Albert1,
  2. J Dunham2,
  3. S Khan2,
  4. J Stansberry2,
  5. S Kolasinski2,
  6. D Tsai2,
  7. S Pullman-Mooar2,
  8. F Barnack2,
  9. C Striebich3,
  10. R J Looney4,
  11. E T Luning Prak2,
  12. R Kimberly5,
  13. Y Zhang2,
  14. R Eisenberg2
  1. 1
    Dartmouth Medical School, Hanover, New Hampshire, USA
  2. 2
    University of Pennsylvania School of Medicine, Philadelphia, Pennsylvania, USA
  3. 3
    University of Colorado Health Science Center, Aurora, Colorado, USA
  4. 4
    University of Rochester, Rochester, New York, USA
  5. 5
    University of Alabama at Birmingham, Birmingham, Alabama, USA
  1. Robert Eisenberg, University of Pennsylvania, Department of Rheumatology, 756 BRB II/III, 421 Curie Blvd, Philadelphia, Pennsylvania 19104–6160, USA; raemd{at}mail.med.upenn.edu

Abstract

Objective: To study the effects in systemic lupus erythaematosus (SLE) of B cell directed therapy with rituximab, a chimeric monoclonal antibody directed at CD20+ B cells, without concomitant immunosuppressive therapy in mild to moderate SLE.

Methods: Patients (n = 24) with active SLE and failure of ⩾1 immunosuppressive were recruited from three university centres into this phase I/II prospective open-label study. Patients were followed for 1 year to assess safety, efficacy and biological effects.

Results: In total, 18 of the patients scheduled to receive the full lymphoma dose of rituximab were evaluable for B cell levels in peripheral blood. Of these, 17 had effective CD19+ B cell depletion (<5 cells/μl). However, six of the depleted patients showed B cell return before 24 weeks. A total of 70% of patients improved by week 55, as defined by an SLE Disease Activity Index (SLEDAI) score improvement of ⩾2 units from baseline. The degree of CD19+ B cell depletion was correlated with SLEDAI improvement at week 15 (r = 0.84). In general, rituximab infusions were well tolerated. Approximately a third of the patients developed human anti-chimeric antibody (HACA) titres, which correlated with poor B cell depletion. Most patients (9 of 14) did not respond to immunisations with Pneumovax and tetanus toxoid.

Conclusions: Rituximab is a promising new therapy for SLE. The variability of responses in patients with SLE may be related to HACA formation. The failure to respond to immunisations is surprising, in view of the apparently low risk of infections. Better biological markers are necessary to follow these patients during treatment.

Statistics from Altmetric.com

Systemic lupus erythaematosus (SLE) is a pleomorphic disease with the central feature of multiple autoantibodies that are not restricted to a single organ system. B lymphocytes are thought to play a major role in the pathogenesis of SLE.13 It has long been known that the B cell lineage, particularly plasma cells, are the source of autoantibodies that characterise SLE. In addition, activation of T cells is partially controlled by B cell antigen presentation.4 5 Animal models of SLE suggest that intrinsic B cell defects are critical in the development of the disease.6 7 In patients with SLE, several B cell abnormalities are evident, and B cell hyperactivity may be a defining event of SLE.8 9 Thus, therapy directed at B lymphocytes holds promise for the treatment of SLE by interacting with any one of the above mechanisms.10 11

The introduction of monoclonal antibodies that selectively target B lymphocytes has provided an opportunity to develop more rationally focused therapy for SLE.3 1218 Rituximab is a chimeric monoclonal antibody against CD20, a B cell-specific surface marker. CD20 is present on B cells from early development through maturity until the differentiation of B cells into plasma cells. Rituximab selectively depletes CD20+ B lymphocytes through a variety of mechanisms, including antibody-dependent cell-mediated cytotoxicity, apoptosis and complement-mediated cytotoxicity.19 20 Rituximab was first approved for the treatment of relapsed or refractory CD20+ B cell non-Hodgkin’s lymphoma in 1997 and has since been used in over 900 000 patient treatments with an excellent efficacy and safety profile. More recently, it has been approved for use in combination with methotrexate in patients with rheumatoid arthritis who have an inadequate response to one or more tumour necrosis factor (TNF) antagonist therapies.21 The availability of this approved agent has greatly facilitated initial studies of B cell depletion as a treatment for SLE. A number of investigations have been published or presented, with promising observations of clinical improvement and modest toxicity.15 17 18 2225 Two ongoing randomised controlled trials aim to demonstrate efficacy in non-renal and renal SLE, respectively.26

The present study of 24 patients found a suggestion of favourable responses and modest toxicity in line with previously reported experience. However, several aspects of the mechanistic investigations that accompanied the study highlight the complexity of the biological responses of patients with SLE to this agent. In particular, the high incidence of human anti-chimeric antibody (HACA) responses appears paradoxical in the face of the failure of most patients to respond to immunisations.

PATIENTS AND METHODS

Patients

Patients aged 18–70 years who fulfilled the American College of Rheumatology (ACR) criteria for SLE were recruited from three university medical centres (University of Pennsylvania, Philadelphia, Pennsylvania, USA; University of Colorado, Denver, Colorado, USA; University of Rochester, Rochester, New York, USA). Patients had SLE for a duration of ⩾6 months, had failed at least one immunosuppressive drug and had currently active disease, as defined by an SLE Disease Activity Index (SLEDAI) score of >127 and active visceral disease. Patients were not permitted to receive an immunosuppressive drug for 1 month before starting the study. All patients gave written informed consent before starting the study. The protocol and consent form were approved by the institutional review board (IRB) at each university.

Procedures

Rituximab was provided by Genentech Inc. (South San Francisco, California, USA) and Biogen Idec (San Diego, California, USA). The protocol was originally structured as a dose-escalation safety assessment of rituximab in SLE, with an initial 100 mg dose of rituximab followed by a repeat dose at 100 mg/m2 1 month later. After treatment of the second patient, the protocol was altered to follow the full dosage approved for use in lymphoma (four once-weekly intravenous (IV) infusions of rituximab 375 mg/m2). To minimise infusion reactions, patients were pretreated with diphenhydramine 25 mg IV and oral acetaminophen 650 mg 30 min before rituximab administration. After the first two patients, the diphenhydramine dose was changed to 50 mg orally because of mild transient hypotension, and methylprednisolone 100 mg IV was added 30 min before each infusion. Patients were permitted to continue hydroxychloroquine and non-steroidal anti-inflammatory drugs (NSAIDs), as well as prednisone daily.

Patients were followed for 1 year after the four rituximab infusions to assess the safety and the efficacy of rituximab. Safety assessments included haematological, renal, hepatic and immunological assays, as well as observations for adverse events. Efficacy outcomes included the SLEDAI and the Systemic Lupus Activity Measure (SLAM),28 both of which are validated instruments for measuring SLE disease activity.29 SLEDAI and SLAM scores were recorded at baseline and at weeks 4, 7, 11, 15, 19, 27, 39 and 55. Serum autoantibodies and immunosuppressive and steroid medication requirements were also determined at baseline and at weeks 4, 12, 24, 36 and 56. Multiparameter flow cytometry analysis of peripheral blood lymphocytes was performed at baseline and at weeks 4, 12, 24 and 36 to monitor the B cell count and to evaluate B cell subset distributions after depletion with rituximab. Pharmacokinetics and HACA titres were measured by ELISA at Genentech. Serum was tested for antibodies to pneumococcal polysaccharides and tetanus toxoid (TT) at screening, 5 months after rituximab treatment and 1 month after immunisations with Pneumovax and TT at month 7. FcγRIIIa genotyping was performed as described previously.30

Analysis of data

All data were analysed by descriptive statistics using SAS V 8.1 (SAS Institute, Cary, North Carolina, USA). The correlation between SLEDAI and SLAM scores and B cell number was determined using the Pearson correlation coefficient test. B Cell counts were calculated by multiplying the lymphocyte count obtained from the complete blood count by the fraction of CD19+ lymphocytes. Categorical data were analysed in contingency tables by the Fisher exact test.

RESULTS

Clinical effects

In total, 24 patients from the 3 institutions were enrolled over a 48-month period from January 2001 to January 2005. Characteristics of the patients at baseline are detailed in table 1. The majority of patients had longstanding SLE, and all patients had immunosuppressant-resistant, moderate severity SLE. There were similar numbers of Caucasian and African–American patients in the study (13 vs 11), but the female-to-male ratio was 22:2. A total of 17 patients received the full rituximab dose (ie, that approved for use in lymphoma). Two patients received a lower dose in the escalation phase (patients 001 and 002), and three patients did not receive the full dose because of a serious adverse event (SAE) (patients 008, 010 and 020). Two patients were withdrawn after one infusion because of non-compliance (patients 021 and 024). Of the 22 patients who were scheduled to receive the full dose of rituximab, we had sufficient post-treatment data to determine B cell depletion in 19 patients, including three that did not actually receive the full dosing (008, 020 and 024). Of these patients, 18 achieved effective B cell depletion (<5 cells/μl of blood; fig 1). The remaining patient (015) only partially depleted to 50 cells/μl of blood, from an initial value of 900 cells/μl (partial depleter). The lowest B cell numbers were found between between 4 and 12 weeks after the last dose of rituximab. The duration of B cell depletion varied, ranging from 4 weeks to at least 55 weeks (median of 28 weeks). Seven patients had an early return of B cells before 24 weeks (short-term depleters). Two patients (024 and 008) who received only one or two infusions, respectively, nevertheless showed full and persistent B cell depletion, while patient 020, whose treatment was stopped after two infusions because of serum sickness, had an early return of B cells. The other two patients who did not received the full schedule dosing also did not have sufficient follow-up of their B cell counts (010 and 021).

Figure 1 B Cell counts over time for individual patients. Patient data are organised with the latest reconstituting patients toward the front, for better visibility. B Cell counts <5 cells/μl (assessed by multiparameter flow cytometry) were considered to indicate B cell depletion.
Table 1 Patient characteristics at study entry

Disease activity scores tended to improve after rituximab treatment in most patients. The mean SLEDAI score across all patients was decreased from 8 points at baseline to 4 points by week 15, and 5 points by week 55. In total, approximately 70% of patients improved their SLEDAI score by ⩾2 points at some point within the 55 weeks of the trial. A week 55 landmark analysis was a predetermined endpoint, although it reflects a possible long-lasting impact on the disease activity. Similar results were found for SLAM scores. Area under the curve cumulative analyses of SLEDAI and SLAM scores were similarly favourable in terms of decreased disease activity (data not shown). At week 15, 10 out of 16 evaluable patients had an improvement in SLEDAI score by ⩾2 points. An analysis of responses for the specific clinical manifestations of SLE is shown in table 2. Skin and joint disease showed the greatest response to rituximab in this limited number of patients. Following rituximab treatment, most patients were able to reduce their prednisone dose regardless of whether or not the patients had immunosuppressive drug therapy reinstated post rituximab treatment. Multiple correlations of all clinical and serological parameters showed that the best association was with SLEDAI score and B cell number, with a peak r value of 0.84 (p<0.001) at week 15. Specifically, there was no correlation with ANA titre (which did not decrease significantly), complement levels (which did not increase significantly) or other laboratory measures such as ESR and CRP.

Table 2 Clinical outcomes of systemic lupus erythaematosus (SLE) disease manifestations: number of patients exhibiting specific symptoms

Immunological studies

Of the 24 patients, 10 produced an antibody response to rituximab (HACA), including 7 of the 16 patients who received the full 4-week treatment course (table 3). Three patients (001, 002 and 020) who developed HACAs had received lower rituximab doses. Patients 001 and 002 were treated in the initial dose escalation protocol, where they received 100 mg as a test dose. Patient 001 then received a second dose of 100 mg/m2 1 month later, and developed an adverse event and a high HACA titre (see below). As a result, patient 002 was not re-treated; however, this patient was found to have a modest HACA titre not associated with any ill effects. Patient 020 developed a serum sickness-like reaction after her second dose, associated with an extremely high HACA titre. Among the evaluable patients who were scheduled to receive the full rituximab course, the development of HACA was associated with failure to deplete fully or maintain depletion (table 4; p = 0.024 by Fisher exact test). Patient 015 was unusual in being the only patient not to show B cell depletion at any time point. Curiously, her serum rituximab levels actually fell with each infusion: from 58 μg/ml to 48 μg/ml to 8.2 μg/ml 1 week after the first, second and third infusions, respectively. At 3 weeks after the fourth infusion, serum rituximab levels were undetectable and, at 15 weeks, the patient had a modest HACA titre (73 ng/ml). In the 17 patients for whom data were available, the absence of rituximab in the serum at 15 weeks showed a slight but insignificant trend towards an association with HACA titre (p = 0.64 by Fisher exact test). HACA positivity appeared late (weeks 31, 39 or 55, respectively) in three patients who had rituximab in their serum at 15 weeks (fig 2). In patients who had no detectable rituximab in the serum at week 15 and developed HACA, four of five did so before 24 weeks. Finally, the reintroduction of immunosuppressives in nine patients within 6 months of starting rituximab treatment did not appear to influence the development of HACAs (table 4).

Figure 2 Human anti-chimeric antibody (HACA) titres (A) and rituximab serum levels (B) over time for individual patients.
Table 3 Immunological studies in patients with systemic lupus erythaematosus (SLE) treated with rituximab
Table 4 Statistical evaluations of biological tests

We measured serum titres of antibodies to 14 serotypes of pneumococcal polysaccharides and TT before starting rituximab, 6 months later, and then 1–2 months after immunisation with Pneumovax, and TT at about 7 months. Of the 14 evaluable patients, all showed a persistent level of antibodies between the first and second bleeds, as others have seen.31 Strikingly however, nine patients showed no response (defined as a twofold rise in titre) to any of the antigens measured, while two patients showed a response to one of the immunisations, and three patients responded to both. Three of the five patients who showed some response (including the two “double” responders) were short-lived depleters, while six of the nine non-responders were full depleters. This apparent trend did not reach statistical significance. In addition, the addition of cytotoxic therapy after the rituximab course did not significantly influence the response to immunisations (data not shown). Thus, in spite of a failure to deplete γ globulin levels (IgG or IgM) or existing antibody titres, immunisations were often not effective.

The degree of B cell depletion did not correlate with ethnicity or presence of antibodies to extractable nuclear antigens (table 3). The association with the FcγRIIIa genotype polymorphism at position 176 showed a surprising trend towards greater depletion with the low affinity allele (F), but this did not reach statistical significance (p = 0.34). Interestingly, of the 13 patients in whom we could judge the rituximab levels at the time of B cell return, two still had detectable drug. The addition of a cytotoxic drug after rituximab therapy had no effect on the persistence of B cell depletion (data not shown). Extensive multiparameter flow cytometry analyses were performed on all patients, and these are the subject of a separate report.32

Adverse reactions

There were three SAEs considered to be possibly or probably related to rituximab infusion. One serious reaction occurred after the second infusion (see below), and one delayed infusion reaction occurred 10 days after the second dose.

Patient 001 experienced a SAE that was probably related to the study drug. This patient had received an initial low dose of rituximab (100 mg), followed by 100 mg/m2 1 month later. At 6 h after the second infusion, the patient developed hypotension and fever, which responded to fluids and corticosteroids. However, the patient then developed a marked, persistent sinus bradycardia (heart rate <35 beats/min at rest), without hypotension or symptoms, which lasted for approximately 36 h during observation as an inpatient. Although this SAE resolved without sequelae, the patient was withdrawn from the study. Subsequently, the patient was found to have developed a high HACA titre (1890 ng/ml) (fig 2A).

Patient 013 developed headache, fever and abdominal pain 8 weeks after the fourth infusion of rituximab 375 mg/m2. A lumbar puncture revealed a white blood cell count of 590×103/ml and normal glucose and protein levels. Although the patient’s plated cerebrospinal fluid (CSF) cultures were negative, the culture broth grew coagulase-negative staphylococci. The patient was diagnosed with possible lupus cerebritis and possible meningitis. The patient received treatment with antibiotics and corticosteroids and the episode resolved without sequelae. This patient did not develop a detectable HACA titre.

Patient 020 developed a serum sickness-like reaction 1 week after the second infusion of rituximab, which was considered likely to be related to the study drug. The patient had a macular rash, polyarthritis, fever and low complement C3 and C4. Outpatient treatment with an increased dose of oral steroids led to resolution without sequelae. This patient did not receive further rituximab and, at week 15, the patient developed an extremely high HACA titre (28 700 ng/ml).

One death occurred during the study. The patient (patient 008) had a history of transaminitis prior to study entry and developed transaminitis (alanine aminotransferase and aspartate aminotransferase >300 IU; bilirubin normal; prothrombin normal) after the third infusion of rituximab 375 mg/m2, and was withdrawn from the study. This patient was lost to follow-up until 6 months later when she presented with atrial fibrillation, inflammatory myositis and bronchiolitis obliterans-organising pneumonia, which were treated with steroids and mycophenolate. Subsequently, the patient developed abdominal pain, was admitted to hospital, and was diagnosed with pancreatitis. While in the hospital, the patient developed aspiration pneumonia and died secondary to sepsis. The patient’s death was attributed, in part, to uncontrolled SLE. The patient never developed a detectable HACA titre. She never had B cells in her peripheral blood, even before therapy.

DISCUSSION

This uncontrolled but prospective study provides additional suggestive evidence that rituximab is possibly effective for the treatment of SLE. Disease activity scores tended to improve after rituximab treatment, with 70% of patients achieving a ⩾2 improvement in the SLEDAI during the 55 weeks of the study.33 However, since the study had no control group, we cannot rule out regression to the mean. In addition, steroids and cytotoxic medicines were varied after the completion of the rituximab infusions, so it is clearly not possible to draw any firm conclusions about clinical efficacy at this time. The currently ongoing randomised controlled trials will determine definitively if the uncontrolled experience of our trial and others is valid.26

Our trial was specifically designed to achieve B cell depletion without cyclophosphamide. We did this for two reasons: first, to observe the effect of rituximab alone on SLE disease, as well as its effect on serological and immunological parameters; and second, from a practical standpoint many of the actual and potential complications of SLE therapy are due to cyclophosphamide. Therefore, eliminating this drug would be a potential goal of rituximab therapy. By contrast, the addition of an immunosuppressive agent to the rituximab regimen might lead to more persistent B cell depletion. In this regard, therefore, it is of interest that a similar open-label trial that did not allow concomitant immunosuppressives found clinical improvement in most patients. These investigators also saw a relatively high incidence of HACA formation (4 out of 15 cases), but they had less variability in the patterns of B cell depletion.34

Several aspects of our study raise important issues about the potential use of rituximab in SLE, beyond the basic question of whether the drug is efficacious. First, the degree or persistence of B cell depletion was more variable than that found in patients with rheumatoid arthritis.35 This is in accord with the University of Rochester study, although only four patients in that series received the full 4-week course of 375 mg/m2 (three of these showed full depletion).8 15 Nevertheless, those authors did find an association of B cell depletion with clinical improvement. In our study, the degree of B cell depletion correlated highly with improvement in the clinical scores (SLEDAI or SLAM) at 15 weeks. Thus, it seems likely that the peripheral blood B cell count reflects, to some degree, the biological effects of rituximab that ameliorate SLE disease activity and, consequently, a failure to achieve complete or persistent B cell depletion may predict a poor therapeutic response.

A second important related issue in our study is the high level of HACA formation. This is in clear contradiction to what has been seen with rituximab in patients with lymphoma, as well as in those with rheumatoid arthritis.21 35 36 The development of HACA in the first two patients in this study may have been due to the low of rituximab doses used. This is supported by the even higher levels of HACA found in the University of Rochester study, in which 12 of the 16 patients received less than the full lymphoma dose of rituximab dose.15 Starting with the third patient in our study, we infused 100 mg of methylprednisolone before each rituximab infusion. It is not clear if this had any effect on the HACA response, although it may have helped prevent infusion reactions. In addition, there was no apparent impact of adding cytotoxic therapy after rituximab (or with rituximab in the University of Rochester study) on the formation of HACA, as has been reported for HACAs with infliximab.15 37 In two of our subjects (patients 001 and 020), the development of HACAs may have been causally related to adverse events. Serum sickness in association with HACAs has been reported previously in patients with Sjögren syndrome who were treated with rituximab,38 and overall serum sickness appears to occur more frequently in rituximab-treated patients who have an autoimmune disease.39 Typically, it occurs about a week after the second dose of rituximab, as in our patient.38 40 We also found that HACA formation was associated with failure to achieve full and persistent B cell depletion, although from our limited data it was not possible to determine the direction of any potential causal relationship between these two sets of findings. It is possible that HACA formation decreases the drug effect, and therefore the extent of B cell depletion; alternatively, patients that deplete poorly may be capable of producing a HACA response. This is consistent with the trend we observed for patients who did not fully deplete B cells to respond better to vaccination. The initial detection of HACAs late in the course (week 23 or later) in five of seven HACA+ patients who received the full rituximab course suggests that the HACAs were not responsible for decreased drug effect in those individuals. However, of the three patients who had HACA detected at week 15 (patients 013, 015 and 020; excluding 01 and 02 who had low dose treatment), all failed to have full and persistent depletion. Thus, in those individuals, the HACAs are more likely to be mechanistically related to the response of their B cells to rituximab. It is possible that concomitant cytotoxic therapy with cyclophosphamide would decrease HACA formation and thus increase rituximab’s effectiveness.

A third important aspect of our study concerns the immunisations given approximately 6 months after rituximab therapy. Previous published experience with Pneumovax and TT in SLE would have predicted responses in nearly every patient.41 42 As others have reported in patients with SLE, rituximab treatment did not reduce pre-existing antibody titres significantly over 6 months,8 although recent evidence suggests that some antibodies in some patients (including those against tetanus, but not those against measles) may show a decrease.43 However, 9 of the 14 evaluable patients in the current study failed to respond to either immunisation with a twofold rise in antibody titre. Previous data have indicated that rituximab-treated patients do not respond to a neoantigen, and the van der Kolk study also showed impairment of a response to a secondary challenge given 4 weeks after rituximab treatment.44 45 Most studies of the response of patients with SLE to immunisation show a normal or somewhat subnormal response, but not a response failure as we observed.46 Failure to respond tended to be associated with full and persistent B cell depletion, but the number of subjects was too small to show statistical significance. Nevertheless, the inability to mount a secondary antibody response suggests that these rituximab-treated patients with SLE would be highly susceptible to infection. So far, our data and those of others have not shown evidence for a clear increase in the risk of infection with rituximab treatment, which may be due to the persistence of antibody titres. The recent observation of two occurrences of progressive multifocal leukoencephalopathy (PML) in patients with SLE treated with repeated courses of rituximab indicate the need for ongoing vigilance, although the relationship to rituximab therapy in these cases is still unclear.4751

B cell depletion with rituximab in uncontrolled case series has generally been associated with amelioration of the signs and symptoms of SLE and other autoimmune diseases, but the effect on autoantibody titres has been variable.8 15 18 24 5254 In our study, autoantibody titres did not changed significantly in response to treatment with rituximab. This may be due to the presence of autoantibody-producing CD20-negative plasma cells.55 We found that B cell depletion was correlated with the low affinity allele of the FcγRIIIa receptor. This association, although not statistically significant, is opposite to what was observed in the Rochester study, and does not have an obvious mechanistic association. Flow cytometric analysis of B cell subsets is beyond the scope of this paper and is reported separately.32

In conclusion, our study suggests that rituximab is possibly effective for the treatment of SLE and the safety profile appears to be reasonable. However, 3 out of 24 patients experiencing SAEs that may be attributable to the study drug is in contrast to the overall excellent safety profile of rituximab in oncology and in rheumatoid arthritis. The biological complexities of the immune competence of the treated patients, as evidenced by their responses to the administered drug (HACAs), but not to immunisations, requires further study and the development of additional biomarkers. The interactions of patients with SLE with rituximab appear to reflect some unique properties of the immune system in this disease.

Acknowledgments

The authors would like to thank Evan Reider and Lytia Fisher for help with the clinical study, and Adelphi Communications Ltd, Macclesfield, Cheshire, UK, for help with preparation of the manuscript.

REFERENCES

View Abstract

Footnotes

  • Competing interests: This work was in part supported by the US distributors of rituximab, Genentech, and RE, DA, ETLP and RJL have consulted with Genentech on B cell depletion in systemic lupus erythaematosus.

  • Funding: Support for this study was provided by the National Institutes for Health (National Institute of Allergy and Infectious Diseases) through the Autoimmune Centers of Excellence (U19AI-46358), by grants from the Alliance for Lupus Research, the Lupus Clinical Trials Consortium and from Genentech, San Francisco, California, USA. Patients enrolled in Colorado were also supported by their General Clinical Research Center through a grant (MO1 RR00051).

  • Ethics approval: All patients gave written informed consent before starting the study. The protocol and consent form were approved by the institutional review board at each university.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.