Combination Immunotherapy of Relapsed or Refractory Low-Grade or Follicular Non-Hodgkin’s Lymphoma with Rituximab and Interferon-α-2a¹

The NHLs³ are a diverse group of lymphoid neoplasms that collectively rank fifth in cancer incidence and sixth in mortality (1). The estimated prevalence of NHL in the United States is projected to reach 244,000 cases in 1998 (1). Low-grade or follicular NHL has been estimated to account for ∼120,000 of the NHL cases, who have a median survival of 6.2 years (1). Most low-grade NHLs are of B-cell origin and are responsive to initial therapy, but virtually all advanced stage patients experience a continuous pattern of relapse and eventually succumb to the disease or its complications, with subsequent courses of chemotherapy leading to shorter remissions and inevitable relapse (2, 3). Thus, the search for effective novel therapies and combinations of existing therapies continues as an ongoing effort.

Rituximab has demonstrated significant clinical activity in the treatment of relapsed or refractory, low-grade or follicular NHL (4, 5, 6, 7, 8, 9). In a large efficacy study (n =166) in which rituximab was given as four consecutive weekly infusions of 375 mg/m², the ORR was 48% for the intent-to-treat population and 50% for the evaluable population. In evaluable patients, the median TTP and duration of response were 13.1 and 11.2 months, respectively (7, 8, 9).

Rituximab is a mouse/human chimeric antibody containing human constant regions (IgG1 κ isotype) and murine variable regions that specifically target the CD20 antigen (10). The CD20 antigen is expressed exclusively on mature B cells, including those in most B-cell lymphomas, but not on normal plasma cells, pre-pre-(or pro-)B cells, stem cells, or dendritic cells (11). Rituximab treatment rapidly depletes the CD20+ normal and tumor B cells in the peripheral blood and bone marrow. After depletion of CD20+ B cells, the B-cell population is reconstituted from the stem cells and pro-B cells, reaching normal levels within 9–12 months (9). Rituximab is a well-tolerated outpatient treatment completed in 22 days. Treatment-associated adverse events were primarily grade 1 or 2 and infusion related, including transient fever,chills, nausea, and headache (7, 8).

Combination therapy with rituximab has been investigated with various agents. The safety and efficacy of rituximab given in combination with CHOP (cyclophosphamide, doxorubicin, vincristine, and prednisone)chemotherapy was investigated in a Phase II trial. The combination therapy was very effective (ORR of 100% in all treated patients) and demonstrated a toxicity profile consistent with each therapy given separately (12). Interim results from a Phase I/II trial of rituximab and granulocyte-colony stimulating factor combination therapy suggest that the treatment is well tolerated and at least as efficacious as rituximab given alone (13, 14). Rituximab given in combination with an ⁹⁰yttrium-labeled anti-CD20 murine monoclonal antibody (IDEC-Y2B8) to patients with relapsed or refractory NHL has demonstrated an ORR of 67% in all patients, with an ORR of 82% in low-grade patients (15, 16, 17).

Several trials have investigated the efficacy of combining immune system modulators, such as IFN, with chemotherapy for treatment of NHL (18, 19, 20, 21). Preclinical studies in mice revealed a synergistic antitumor effect between IFN and monoclonal anti-idiotype antibodies (22, 23). Mechanisms by which IFN may increase the effectiveness of antibodies include the potentiation of antigen expression (24), increased targeting of antibodies into tumors (25, 26, 27, 28), and enhanced cytotoxicity of immunotoxins (29, 30, 31). A combination trial investigated the effect of IFN therapy on the clinical efficacy of anti-idiotype antibodies in patients with NHL. Patients treated with the combination therapy demonstrated an improved initial response rate (9 of 12 responded)compared with patients who received anti-idiotype antibody alone (8 of 16 responded; Refs. 32 and 33).

IFN, as a single agent or in combination therapies, has demonstrated significant clinical activity in the treatment of low-grade NHL (34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44). A 10-year literature review and analysis of results generated from various single-agent studies revealed that the overall response rate in 237 patients with relapsed, low-grade, or follicular NHL was 36% (45). Adverse events were primarily flu-like symptoms, such as fever and fatigue (35).

Data from trials of combination therapy reveal that the addition of IFN to anthracycline-containing chemotherapy regimens has been associated with significantly prolonged progression-free survival (TTP); however,IFN did not consistently increase response (37, 38, 42, 44, 46). Three large multicenter trials from the Eastern Cooperative Oncology Group, the Groupe d’Etude des Lymphomes de l’Adulte (GELA),and the European Organization for Research and Treatment of Cancer investigated the administration of IFN in conjunction with chemotherapy or as maintenance therapy after chemotherapy. After a median follow-up period of 3 years, an increase in disease-free survival (TTP) was reported in the IFN-treated arms for the Eastern Cooperative Oncology Group, Groupe d’Etude des Lymphomes de l’Adulte, and European Organization for Research and Treatment of Cancer studies of 47, 60,and 45 weeks, respectively (38, 42, 44). IFN has been approved in the United States for use as initial treatment of clinically aggressive follicular NHL in conjunction with anthracyclinecontaining combination chemotherapy because of a significant increase in the TTP (37, 38, 42, 44, 46).

Given the efficacy of rituximab and IFN as single agents and the preclinical results indicating synergistic effects between IFN and anti-idiotypic antibodies, the safety and efficacy of rituximab given in combination with IFN-α-2a (Roferon-A) was investigated. In a 12-week treatment course, patients with relapsed low-grade or follicular NHL received IFN-α-2a, 5 MIU s.c. three times weekly(weeks 1–12), and 4 infusions of rituximab, 375 mg/m² given weekly (weeks 5–8).

This was an open-label, single-arm, multicenter Phase II safety and efficacy study of combination therapy with rituximab and IFN-α-2a(Roferon-A) in patients with relapsed or refractory, low-grade or follicular, NHL. Rituximab was administered as an i.v. infusion for 4 doses (375 mg/m² given once weekly on weeks 5–8)to patients who were receiving concurrent treatment with IFN [5 MIU administered s.c., three times weekly for 12 weeks (weeks 1–12)]. The treatment schedule for the combination therapy is displayed in Fig. 1. This design should allow for establishment of the effects of IFN-α prior to infusion of rituximab,and continues combination therapy for 2 months (1 month during rituximab treatment and for 1 month after, when rituximab should still be circulating in the patient’s serum).

Patients with histologically confirmed low-grade or follicular B-cell lymphoma (IWF types A–D) who relapsed or failed primary therapy and had progressive disease requiring further treatment were enrolled provided they met the following conditions: bidimensionally measurable disease; no more than four relapses after standard therapies; a demonstrable monoclonal CD20-positive B-cell population in lymph nodes or bone marrow; no known hypersensitivity to IFN; and a prestudy performance status of 0, 1, or 2, according to the WHO scale. The study was approved by the Institutional Review Board at each site, and all patients were required to give written informed consent.

The patients’ characteristics are shown in Table 1. Of the 38 patients enrolled in the study, 28 (74%) patients were male and 37 (97%) were Caucasian. The median age was 53 years (range, 31–80 years). Patients were a median of 3.7 years from diagnosis, and 28 had stage III/IV disease at diagnosis. The majority of patients (92%) had low-grade NHL (IWF types A, B, or C; Ref. 3); 4 patients (11%) were IWF type A,and 31 patients (81%) were IWF type B or C. Three patients (8%) had follicular intermediate-grade NHL (IWF type D). All patients had been treated with at least one prior lymphoma therapy regimen (median, 2;range, 1–6) and had at least one relapse prior to enrollment. A total of 11 patients had at least one prior course of radiation therapy for lymphoma. Five patients had undergone autologous bone marrow transplantation.

IFN-α-2a (Roferon-A, recombinant human leukocyte IFN; kindly provided by Roche, Nutley, NJ) was to be administered as a s.c. injection three times weekly at a dose of 5 MIU for 12 weeks. Self-administration was permitted. If significant adverse events known to be associated with IFN were documented, IFN dose modification was required. A 50% dose reduction to 2.5 MIU/injection was implemented if any of the following adverse events were observed: ANC <1,000/mm³,platelets <75,000/mm³, or AST or ALT >6–11 times normal. The patient was instructed to discontinue IFN treatment for 1 week if ANC was <500/mm³, platelets were<50,000/mm³, or if AST or ALT values were >11 times normal. After 1 week, IFN treatment could be resumed at 2.5 MIU/injection only if the ANC was >3,000/mm³,platelets were >100,000/mm³, and AST or ALT values were <3 times normal. IFN treatment was to be permanently terminated if abnormal thyroid stimulating hormone or T4 values were observed.

A total of four infusions of 375 mg/m² rituximab(Rituxan, Mabthera; kindly provided by IDEC Pharmaceuticals Corp., San Diego, CA) were to be administered once weekly four times (weeks 5–8 of the treatment period) on an outpatient basis. Rituximab, diluted to 1 mg/ml with normal saline, was to be administered as an i.v. infusion at an initial dose rate of 50 mg/h for the first h. If no signs of toxicity were observed, the dose rate was to be escalated gradually (50 mg/h increments at 30-min intervals) to a maximum of 400 mg/h (300 mg/h maximum for the first infusion only). No dose reductions were allowed. The production of rituximab has been described previously (10).

Toxicity was evaluated using the National Cancer Institute’s Common Toxicity Criteria (Recommendations of Representatives of the National Cancer Institute’s Clinical Cooperative Groups and the Cancer Treatment Evaluation Program, February 1988). Disease assessment included the following evaluations: physical examination and assessment of performance status and B-symptoms (baseline, weeks 8 and 12, and every 3 months through 2 years, then every 6 months), chest X-ray(baseline), bone marrow biopsy (baseline and to confirm a CR, if positive at baseline), and computed tomography or magnetic resonance imaging (baseline, week 12, and every 3 months through 2 years, then every 6 months).

Laboratory analyses included the following: serum chemistry and serum immunoglobulin assays at baseline, weeks 8 and 12, and then every 3 months through the first year of follow-up; HACA [detected using one-site (sandwich) ELISA (7, 8)] assays at baseline,week 12, and the third month of follow-up; complete blood count at baseline, weeks 5–12, and every 3 months through 2 years, and then every 6 months; and urinalysis at baseline, weeks 8 and 12, and the third month of follow-up. In addition, peripheral blood was analyzed by flow cytometry at baseline, weeks 6 and 8, and every 3 months for the first year of follow-up, and then every 6 months. Thyroid function (T4 and thyroid stimulating hormone) was assessed at baseline and at the sixth month of follow-up. Serum samples for pharmacokinetic analyses of rituximab were obtained for all patients prior to and immediately after each infusion, at weeks 9–12, and at the third month of follow-up.

Patients were evaluable for efficacy if they completed the four infusions of rituximab, satisfied all prestudy entry criteria, and met criteria for evaluation of response. Response categories consisted of CR, PR, SD, and progressive disease (9, 47, 48, 49, 50). CR required the following: all lymph nodes visible on computed tomography scan ≤1 cm × 1 cm in size, any node previously palpable on physical examination (and considered to be involved by lymphoma) must not be palpable or must be negative for lymphoma on biopsy or fine-needle aspiration; bone marrow, if initially positive at baseline,must be histologically negative for lymphoma; and the liver and spleen,if abnormal at baseline, should be normal in size and radiographic appearance. PR was defined as 50% or greater decrease from baseline in the sum of the products of the greatest perpendicular diameters of all of the measured lesions (SPD) with no simultaneous increase in size of any other lesion or no new lesions. SD referred to patients who did not exhibit at least a 50% decrease or increase in SPD. Progressive disease was classified as the observation of a 50% increase from nadir in SPD or the appearance of a new lesion. Response classifications of PR and CR were confirmed by reassessment 28 days after the initial determination of response. TTP(responders) and response duration were measured from the first injection of IFN and from the initial observation of response,respectively, until progression of disease.

The Kaplan-Meier (51) product-limits method was used to analyze the TTP and duration of response; curves were generated using PROC LIFETEST (52, 53). The Wilcoxon rank sum test (51) was used for the comparison of serum concentration data by clinical response. Clinical adverse event data were assigned preferred terms using COSTART (54, 55) and were analyzed by calculating the number and percentage of patients and events. Investigators classified adverse events by their relationship to study treatment and severity of the event (grade). If an individual event had more than one grade, the most severe grade was used to characterize this unified event.

All patients received four complete infusions of rituximab. Patients enrolled in the study were to receive 36 s.c. injections of 5 MIU of IFN. The mean total injections received was 31, and the mean total units received was 141 MIU (maximum, 180 MIU). Twenty-seven patients received all 36 injections of IFN. Of these 27 patients, 18 received the total complete dose of 180 MIU (5 MIU/each injection), and 9 received a reduced dose of 2.5 MIU/injection for at least one cycle(three injections) because of adverse events. Eleven patients did not receive all 36 doses; 2 patients each received 36 and 4 doses, and 1 patient each received 34, 31, 30, 24, 15, 6, and 5 doses.

Adverse events were classified as having occurred during the treatment period (the time interval between the first IFN injection and 30 days after last treatment) or during the long-term follow-up period (the time interval between 31 days after last treatment and 1 year after the first IFN injection). Adverse events were classified further by their relationship to study drug. Events were classified as being possibly or probably related to rituximab, IFN, or to both drugs. If the event was not considered to be related to treatment, it was classified as having an unknown relationship or as related to neither drug (such as related to a concurrent illness or study disease). The most frequent adverse events are summarized in Table 2. No grade 4 events were reported.

During the treatment period, the events with the highest incidence included: asthenia (35 patients; 92%), chills (31 patients; 82%),fever (30 patients; 79%), headache (28 patients, 74%), nausea (23 patients, 61%), myalgia (22 patients, 58%), thrombocytopenia (16 patients, 42%), leukopenia (15 patients, 40%), malaise (12 patients,32%), diarrhea (11 patients, 29%), and neutropenia (11 patients,29%). The majority of related adverse events were classified as IFN related. Of the 734 adverse events reported, 124 (17%) were rituximab related, 395 (54%) were IFN related, 61 (8%) were rituximab/IFN related, 10 (1.4%) were of unknown relationship, and 144 (19%) were attributed to neither drug. All 38 patients experienced an IFN-related adverse event; 34 experienced a rituximab-related event; and 19 experienced a rituximab/IFN-related adverse event. Thirty-one patients experienced adverse events related to neither drug (i.e.,concurrent illness or study disease related), and five experienced events with an unknown relationship.

During the treatment period, 28 grade 3 adverse events were reported in 12 patients. The majority of the grade 3 events (23 of 28; 82%) were reported as IFN related. The most common grade 3 events were malaise,asthenia, and myalgia. Five patients either had reductions in their IFN dosing units to 2.5 MIU/dose for at least one cycle (three injections)or did not complete all 36 injections because of these adverse events. Two patients experienced serious adverse events that required hospitalization: concurrent grade 1 dyspnea and grade 2 pneumonia; and a grade 3 neutropenic fever. Both patients recovered after receiving medication and were discharged from the hospital within 4 days. No deaths were attributed to the study treatment or occurred during the treatment period. Deaths were reported in seven off-study patients, all of whom had progressive disease and had received additional therapy after completing this study.

Patients were monitored for infections for a year after their first injection of IFN. Patients treated with antibacterial agents in whom no organism was isolated were presumed to have a bacterial infection. Fifteen grade 1 or grade 2 infections were reported in 10 patients; of these 15 infections, 87% were bacterial and included bronchitis,pneumonia, sinusitis, upper respiratory infection, and a staph skin infection. One of the 10 patients was hospitalized with grade 2 pneumonia, received antibiotics, and was discharged after 3 days.

The peripheral blood B-cell normal range is considered to be 32–341 cells/μl as determined by fluorescence-activated cell sorter analysis. Rituximab treatment resulted in a rapid depletion of peripheral blood B cells, as measured by CD19-positive cells; the CD19(pan-B) antigen rather than the CD20 antigen was used as a B-cell marker because rituximab inhibits the binding of CD20-detecting antibodies. The median B-cell count had declined to near zero prior to the second rituximab dose. Recovery began between 6 and 9 months after completion of rituximab treatment. Median absolute T-cell counts and the median absolute NK cell counts in peripheral blood remained relatively stable throughout the study.

Serum immunoglobulin levels (IgG, IgA, and IgM) were measured at baseline and at specified intervals for 1 year. Mean serum IgG and IgA levels did not fall outside of the normal range throughout the study. The mean IgM level remained within the normal range, except when it fell below normal at one time point at month 4.

Most hematological effects from treatment were mild and transient. Anemias in eight patients were classified as grade 1 or 2, and hemoglobin levels for all patients recovered to at least 9.9 g/dl by the end of the study. One patient received a RBC transfusion for grade 2 IFN-related anemia on day 48. Two grade 3 neutropenic events occurred in two patients. One patient who experienced a grade 3 neutropenic fever on day 176 attributed to rituximab treatment and study disease was hospitalized on day 179 and received antibiotics. The fever resolved, and the patient was discharged after 4 days. Another patient experienced grade 3 neutropenia on the day of the fourth rituximab infusion that was attributed to the combination therapy. The patient recovered within 16 days without receiving medication. Twenty-four thrombocytopenic events were reported; 20 of the events were considered to be related to treatment (10 IFN related, 6 rituximab related, and 4 attributed to combination therapy), 3 events were attributed to lymphoma, and 1 was of an unknown relationship. All events were grade 1 or 2 and transient. Platelet values for all patients recovered to a grade 1 level by the end of the treatment period.

There were no clinically significant abnormalities of grade 2 or greater in laboratory values throughout the study. No patients developed a detectable HACA titer during the treatment or follow-up periods.

All 38 patients were evaluable; 4 of the 38 patients (11%) had a CR,and 13 patients (34%) had a PR, for an overall response rate of 45%. A summary of clinical response is found in Table 3. A univariate analysis of baseline prognostic factors such as age, sex, B symptoms, performance status, stage of disease, extranodal disease, splenomegaly, bone marrow infiltration, elevated lactate dehydrogenase, and response to prior therapy demonstrated that none of these factors had a statistically significant effect upon the outcome of treatment (all Ps ≥ .05).

In all 38 patients, the average decrease in lesion size was 43%. The average decrease in lesion size in the 17 responders was 84%, 97% in the 4 patients achieving a CR, and 80% in the 13 patients achieving a PR. Patients with SD had an average decrease in lesion size of 33%. Lesion size decreased in responders until it stabilized between approximately 4–7 months after study entry. Six patients (16%) had B symptoms or other disease-related symptoms at baseline; all six patients demonstrated resolution of disease-related symptoms by the end of the study. Four of the patients achieved a PR, and the remaining two had SD. The median time to progression in all patients was 8.95 months, with a maximum that has not yet been reached after >31.4 months of observation.

The current estimations for the median duration of response and time to progression for the 17 responders, as determined by Kaplan-Meier analysis, are 22.3 months and 25.2 months, respectively, but final values have not yet been reached. These data are summarized in Table 4. The Kaplan-Meier curve of TTP in evaluable responders in this current study as well as the Kaplan-Meier curve of TTP obtained from a previous Phase III single-agent rituximab trial are presented in Fig. 2. In the Phase III trial, rituximab (375 mg/m²) was given in the same course and schedule (weekly times four) to 166 patients with relapsed or refractory, low-grade or follicular, NHL who were selected by the same inclusion/exclusion study criteria. A comparison of the TTPs achieved in the two trials indicated that the increase in TTP observed in the combination trial has not yet reached statistical significance (P =0.4704 determined by log-rank statistics).

Both the median pre- and postinfusion serum concentrations of rituximab increased with each subsequent infusion, which is an indication that steady-state concentrations were not attained. The median postinfusion antibody serum concentrations were 245.3, 302.2,412.5, and 426.4 μg/ml, respectively, for infusions 1, 2, 3, and 4. Median pre- and postinfusion antibody serum concentrations were higher in responders compared with nonresponders throughout the study;however, these differences were not statistically significant. Rituximab serum levels from 10 patients obtained after the fourth infusion of rituximab were analyzed using a noncompartmental linear regression method to determine the serum half-life(T_1/2). The serum half-life of rituximab ranged from 144.4 h to 443.0 h with a mean T_1/2 of 304.5 h.

IFN-α has pleiotropic effects on the immune system, including increased expression of the MHC class I antigen, cellular adhesion molecules and other tumor antigens, increased production of, and sensitivity to, other cytokines, as well as augmentation of the cytotoxic activity of NK cells, an integral part of ADCC (56, 57). Although the predominant mechanism of action of rituximab in vivo is not known, the antibody is able to mediate both complement and antibody-mediated cell killing (ADCC; Ref. 10). Evidence that IFN-α could stimulate the ability of NK cells to mediate ADCC (58) suggests that simultaneous treatment with IFN-α and antibody may augment efficacy. The regimen used in this trial provided tolerable dosing of IFN-α before, during,and after rituximab therapy to maximize any synergistic activity.

The safety and efficacy of rituximab and IFN-α, given as single agents, were demonstrated previously in patients with relapsed or refractory, low-grade or follicular NHL (7, 8, 35, 36, 38, 39, 43, 45, 59). This trial demonstrated that rituximab treatment(375 mg/m², i.v., once weekly times four) given in combination with IFN-α-2a (5 MIU, s.c., three times a week for 12 weeks) in patients with relapsed or refractory, low-grade or follicular NHL was active, safe, and well tolerated.

There were no unexpected events resulting from the combination treatment. Most related adverse events were attributed to IFN-α-2a treatment (395 of 734 events; 54%), with few being attributed to the combination therapy (61 of 734; 8%). The majority of adverse events were grade 1 or 2; none were grade 4. The hematological or immunological adverse event profile did not appear to significantly differ from that observed in single-agent rituximab trials (4, 5, 6, 7, 8, 9). Most hematological events were mild and transient. B cells were transiently depleted and recovered within 6–9 months, the same time frame observed in rituximab single-agent trials. T-cell and NK-cell counts remained relatively stable throughout the study (4, 5, 6, 7, 8, 9). There were no significant abnormalities of grade 2 or greater in laboratory values during the study. In addition, no patients developed a HACA response, clear evidence that IFN-α-2a therapy does not promote humoral responses against the chimeric antibody.

The overall response rate was similar to that observed in single-agent studies of patients with relapsed or refractory low-grade or follicular NHL. The response rate of 45% (95% CI, 29–61%) is similar to the 50% overall response rate (95% CI, 42–58%) seen in the pivotal trial of rituximab given as a single agent (7, 8) and compares favorably with the 36% response rate determined by a 10-year literature review of IFN single-agent trials (45). As has been observed in rituximab single-agent trials, lesion size decreased in responders until it stabilized between approximately 4–7 months after study entry (9). No statistically significant relationship between various prognostic factors at baseline and response was identified. This is in contrast to a significant relationship (univariate analysis; P < 0.05) between response and histological type, prior ABMT therapy, baseline bcl-2 status, number of relapses, and bone marrow involvement identified in the rituximab single-agent pivotal trial (9). Although responders were found to have higher serum levels of rituximab compared with nonresponders, this difference did not achieve statistical significance, as had been observed in previous trials (9, 60).

Of the five patients with prior ABMT, only one experienced a serious adverse event. This grade 3 febrile neutropenia occurred ∼6 months after therapy and resolved within 1 week. Three patients had dose reductions of IFN. One of these five patients responded with a response duration of 7 months.

When response rate is assessed by histological type and compared with response of the population as a whole, patients with type A histology had a lower ORR (0 responders of 2 patients), patients with type B had a higher ORR (12 of 22; 54%), patients with type C had a similar ORR(4 of 9; 44%), and patients with type D had a higher ORR (2 of 3;67%). Although the ORR was higher in the patients with types B and D histology and lower in patients with type A histology, the small population size precludes definitive conclusions.

The Kaplan-Meier estimate for median TTP in responders of 25.2 months compares favorably with the 13.1-month median TTP observed in the pivotal single-agent rituximab trial that used the same dose and schedule in a similar patient population (9). These efficacy results are consistent with combination trials of IFN and chemotherapy in patients with relapsed or refractory, low-grade NHL in which IFN demonstrated a prolongation of TTP without necessarily increasing the response rate (38, 42, 44). The potential 10.1-month increase in TTP observed with combination therapy, compared with the TTP observed with single-agent rituximab treatment, is not currently statistically significant. Definitive conclusions regarding the potential benefit of combination therapy require a randomized trial investigating single-agent therapy versus combination therapy. Nonetheless, this trial indicates that it is safe and feasible to combine rituximab and IFN-α-2a for the treatment of NHL, and that combination treatment may prolong the TTP. We believe this result justifies further evaluation of this regimen.

We thank Tina Marie Liles, Chet Varns, Steve Shuey, Brian Link,Robert Dillman, Brian Dallaire, and Tom Sklenar for their contributions in carrying out this study. We also thank David Shen, Eric Ding, and Ken Fite for their contributions to data analysis and Julie Deardorff for assistance in the preparation of the manuscript.