PRISMS-4: Long-term efficacy of interferon-β-1a in relapsing MS

Patients.

PRISMS-2 enrolled 560 patients in 22 centers in Europe, Canada, and Australia. All patients had Expanded Disability Status Scale (EDSS7) scores of 0 to 5.0 with at least two relapses in the previous 2 years. Patients from PRISMS-2 were eligible for the two 1-year extensions if their medical and neurologic condition was deemed appropriate for continuing blinded treatment. Patients who had stopped treatment were allowed to re-enter during years 3 and 4.

Study design and treatment.

In PRISMS-2, patients were originally assigned in equal allocation to 22 mcg IFN-β-1a, 44 mcg IFN-β-1a, or placebo using a computer-generated randomization list prepared by Serono. The list was stratified by center with a block size of six. The study was extended twice, for a third and then a fourth year. Before the first patient had reached the end of year 2 and before results were available, placebo recipients were re-randomized in equal allocation to one of the two IFN-ß-1a doses in blinded fashion for PRISMS-4 (figure 1). Patients who had received active medication during PRISMS-2 continued to receive the same blinded dose.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1. Patient enrollment and disposition. Patients receiving placebo in the original study (PRISMS-2) were randomized to treatment with IFN-β-1a at the start of year 3 and before the analysis of PRISMS-2.

Treatment was discontinued for World Health Organization grade 4 toxicity or for pregnancy and could have been discontinued for protocol violations, other adverse events, or noncompliance. Patients who stopped treatment prematurely were followed, with consent, until the end of the planned study period, and their data were included in the analyses.

Immunosuppressive treatments were not permitted during the study. Acetaminophen was recommended for IFN side effects, and steroid use was limited to treatment for acute exacerbations (1.0 g/day intravenous methylprednisolone for 3 consecutive days). Contraception was mandatory for fertile women.

Blinding was maintained during the extension period. As IFN side effects are easily recognized, physicians responsible for general patient management, including safety assessments, were different from those responsible for efficacy assessments. The two doses were provided in identical vials, each containing 0.5 mL of solution. Doses were increased gradually for all patients at baseline and again at the beginning of year 3 to allow adjustment to the medication, thereby enhancing blinding. MRI scans were evaluated without access to treatment information.

Assessments.

Physical and neurologic examinations (including Kurtzke Functional Systems and EDSS7) and laboratory investigations were performed every 3 months during the first 3 years and twice in year 4, with additional neurologic assessments in case of relapses. MRI assessment during the extension period was limited to annual proton density/T2-weighted scans.

Adverse events and concomitant medications were recorded throughout the study. Patients were tested twice yearly for neutralizing antibodies (NAb) to IFN-β using a bioassay in which IFN-β-1a is added at a concentration of 10 LU/mL (LU = laboratory unit, defined as the concentration of IFN-β-1a that protects 50% of WISH [Wistar Institute Susan Hayflick] cells from viral cytopathic effect).8 Blood for the NAb assay was collected at least 36 hours after the previous injection of study drug.

The primary outcome was relapse count per patient over 4 years. A relapse was defined according to Schumacher et al.9 as the appearance of a new symptom or the worsening of an old symptom attributable to MS, accompanied by an appropriate new neurologic abnormality or focal neurologic dysfunction lasting at least 24 hours in the absence of fever, and preceded by stability or improvement for at least 30 days. Other relapse-related measures included time to second relapse, the percentage of patients free from relapses, relapse severity, and duration of relapses. Severity was defined based on changes in neurologic examination or impact on the activities of daily living.3 Relapse duration was defined as time to resolution of the relapse or 90 days, whichever was shorter. Steroid use and hospitalization for MS were also assessed.

The principal disability measure was time to first confirmed EDSS progression, defined by an increase of at least one point on the EDSS that was sustained over at least 3 months. A disadvantage of this measure is that once patients have experienced their first progression, no further disability information is used. Consequently, two further measures of disability assessment were preplanned. The first was the Integrated Disability Status Score (IDSS), a summary measure defined by the baseline-adjusted area under the curve (AUC) of a time–EDSS score plot.10 This measure incorporated all changes in disability, transient or sustained, including both worsening and improvement. The second measure was the number of confirmed one-point EDSS changes per patient, defined as the number of times a patient experienced a sustained one-point worsening in EDSS (or 0.5 point for EDSS >5.5); when a change had been confirmed, the increased score served as “baseline” for the next potential change.

MRI measures for the 4-year analysis were determined using PD/T2-weighted dual echo sequences with 5-mm transverse contiguous slices (or with a minimal inter-slice gap <1 mm) with scanning parameter ranges that included repetition time (repetition time, 2000 to 800 ms; echo time [TE1], 20 to 35 ms; and TE2, 70 to 100 ms). Results were determined using the baseline scan and annual scans from years 1 through 4. MRI activity was assessed in terms of numbers of new T2 lesions and percentages of scans showing such lesions. Burden of disease (BOD), defined as the summed cross-sectional area (in mm²) of lesions in T2 scans, was analyzed as percent change from baseline. Blinded MRI analysis was performed by the University of British Columbia MS/MRI Analysis Group, Vancouver, as in PRISMS-2.5

Statistics.

All data up to the 48-month visit were considered for these analyses. For analyses for years 3 and 4 alone, only data after the 24-month visit were considered.

The efficacy sample included all patients who were randomized and received at least one dose of placebo or IFN-β-1a during year 1 (the intent-to-treat [ITT] sample). Analyses specific to years 3 and 4 included all patients who received at least one dose of IFN-ß-1a during year 3. There were 560 patients in the ITT sample at the beginning of year 1 and 506 patients in years 3 and 4. Patients who stopped treatment prematurely continued to be assessed as planned, with their consent; information from the follow-up period was included in the analyses. For patients who stopped treatment prematurely and either did not consent to follow-up or were lost to follow-up before the 48-month visit, only available data were considered. No imputation was made for missing data.

Baseline values and characteristics for the study extension were defined as the last values obtained either at month 24 or during year 3 before the first dose of study drug.

The primary efficacy analysis included all patients who received any study drug, grouped by their original assignment: crossover, Rx22, or Rx44. This analysis biases against treatment effect because patients in the crossover groups had received active treatment for 2 years. However, it provides the best estimate of dose effect because the Rx22 and Rx44 groups received constant treatment for 4 years. A four-group analysis was also performed to compare the effects of the same doses initiated at year 1 and at year 3: the groups were placebo/22, placebo/44, Rx22, and Rx44.

The relapse count and the number of one-point EDSS changes were analyzed using a Poisson regression model with effects for treatment, center, and their interaction and an offset for time on study. For the proportions of patients remaining relapse-free or progression-free, a logistic regression model with effects for treatment and center was performed. Patients lost to follow-up without relapse or progression were not included in these analyses. The proportions of patients excluded were comparable between groups (ITT analysis).

Time to second relapse and time to first EDSS progression were estimated using Kaplan–Meier curves and analyzed using a Cox proportional hazards model with effects for treatment and center.

Hospitalizations and use of steroids were compared between the Rx22 and Rx44 groups only using an analysis of variance (ANOVA) model on ranked data with effects for treatment and center. Patients lost to follow-up without hospitalization or without use of steroids were not included in these analyses. Excluded patients were equally distributed between groups.

The IDSS was calculated using the trapezoidal rule to estimate the area under the EDSS curve and above the baseline EDSS score for each patient. The EDSS curve for each patient included all EDSS scores from baseline to the 48-month visit. For calculations including only data from years 3 and 4, the 24-month visit was used as the baseline. The IDSS score was tested for differences among treatment groups using an ANOVA model on ranks with effects for treatment and center.

The group median of the mean number of new T2 lesions per patient per scan and the percentage of scans showing new T2 lesions were analyzed using an ANOVA model on ranked data with effects for treatment and center, because the data did not satisfy parametric model assumptions. The change in BOD was analyzed using an ANCOVA model on ranked data with effects for treatment and center, adjusting for baseline BOD.

The rates of adverse events and development of NAb were compared between groups using Fisher’s exact test. Safety parameters, including laboratory test results, were analyzed for all patients who received at least one injection of study drug.

To assess long-term safety, the incidence of adverse events was compared for the Rx22 and Rx44 groups. The incidence, prevalence, and severity of adverse events were also compared between patients in the crossover groups and those treated for 4 years.

A complete analysis plan was generated before database lock and is presented here. This plan included an analysis of dose effect, as this was identified as an important issue based on PRISMS-2 results. The Serono US Biometrics Department performed the statistical analysis using SAS software (SAS version 6.12, SAS, Cary, NC). All tests were two-sided, and results were considered significant at the 5% level.

As the primary outcome was relapse count per patient, with other outcomes considered exploratory, and the primary comparison was high dose versus placebo, no adjustment was made for multiple comparisons.

Ethics.

The study was conducted in accordance with the Declaration of Helsinki. Approval was obtained from the ethics committees of all centers before study initiation. All patients gave written informed consent before undergoing any study procedures at study start and before entering the year 3 and year 4 extensions.

Study population and patient disposition.

Ninety percent (506/560) of the patients originally randomized in PRISMS-2 entered PRISMS-4. This included 13 patients who had stopped therapy during years 1 and 2 (five placebo, four Rx22, and four Rx44). During year 3, four patients in Rx44 and one patient in each of the other three groups did not take treatment. Of these seven patients, one patient in each of the placebo/44, Rx22, and Rx44 groups received treatment during year 4. During year 4, three Rx 44 patients and one placebo/22 patient remained on-study but off therapy. Those patients who discontinued the study at any time during the 4 years were evenly distributed among ITT treatment groups, and their baseline characteristics did not differ from those of other patients. However, dropouts had a higher mean relapse rate during the first 2 years than those who continued in the study (1.6/year versus 1.0/year, p < 0.001, Wilcoxon rank sum test) and also had a slightly higher proportion who had progressed in disability: 35% of dropouts progressed after a mean of 1.5 years before dropping out versus 32% of those who stayed on study for 2 years (p = 0.76, Fisher’s exact test).

During years 3 and 4, 73 patients stopped treatment prematurely (see figure 1), of whom 12 did so at the end of the year 3 extension. Patients receiving high-dose IFN-β-1a discontinued treatment more often than those receiving low-dose treatment (45/251 or 18% verus 28/251 or 11%; p = 0.043), and patients in the crossover groups stopped therapy more often than patients assigned to active treatment for 4 years (36/171 or 21% versus 37/331 or 11%; p = 0.005; the denominator excludes patients who took no drug in years 3 and 4 and were thus not at risk for stopping therapy). Over the 4 years of the study, total treatment discontinuation rates were similar for Rx44 (35/184, 19%) and Rx22 (30/189, 16%). Eighteen Rx44 and eight Rx22 patients discontinued because of adverse events whereas one Rx44 and six Rx22 patients discontinued because of progressive disease.

Demographic characteristics were balanced at study start3 and at the beginning of year 3, whereas significant neurologic and MRI differences at the beginning of year 3 reflected the effects of treatment during PRISMS-2 (table 1).

Efficacy: relapse count per patient (primary outcome).

Four years of treatment with IFN-β-1a (Rx22 or Rx44) significantly reduced the number of relapses per patient per year compared with 2 years of placebo followed by 2 years of IFN-β-1a (placebo/22 and placebo/44 combined; table 2). Furthermore, the smaller number of relapses per patient per year (over 4 years) in the Rx44 group compared with the Rx22 group approached significance (p = 0.069). During years 3 and 4, relapse rates were significantly lower for Rx44 than for the other groups. Relapse rate decreased progressively with each year on therapy, with values for Rx44 of 0.92, 0.82, 0.57, and 0.44 relapses/year for years 1 through 4.

Relapses decreased for crossover patients during years 3 and 4 compared with years 1 and 2. The relapse rates for placebo/22 were 1.3 during years 1 and 2 and 0.63 in years 3 and 4 (mean per patient reduction, 50%; 95% CI, 36 to 63%; p < 0.001; Wilcoxon signed-rank test), and corresponding rates for placebo/44 were 1.29 and 0.68 (mean per patient reduction, 46%; 95% CI, 31 to 61%; p < 0.001; Wilcoxon signed-rank test).

Efficacy: relapse-related secondary outcomes.

In the three-group (ITT) analysis, the proportion of patients who were free from relapses after 4 years was greater in the Rx44 group (19.0%) and in the Rx22 group (14.4%) compared with the crossover groups combined (6.7%; p < 0.001 crossover versus Rx44; p = 0.016 crossover versus Rx22). The Rx44 and the Rx22 groups did not differ significantly (p = 0.159).

The time to first relapse favored active therapy in PRISMS-2. In PRISMS-4, the median time to second relapse was 16.9 months longer for the Rx44 group (p < 0.001) and 8.3 months longer for the Rx22 group (p = 0.006) than for the crossover groups combined; the median time to second relapse for the Rx44 group was also longer than that for the Rx22 group (p = 0.046).

There was less steroid use for MS over 4 years in the Rx44 group (0.4 ± 0.9 courses per patient per year) than in the Rx22 group (0.5 ± 0.8; p = 0.032). This advantage was also observed over years 3 and 4 alone (p = 0.018). No significant dose effect was seen for the duration or severity of relapses or for rates of hospitalization.

Efficacy: disability-related secondary outcomes.

In the ITT analysis, the time to first confirmed EDSS progression was 42.1 months for the Rx44 group compared with 24.2 months for the crossover group (40th percentile [median not reached in all groups]; p = 0.047). The time to first confirmed progression did not differ significantly between the Rx22 group (35.9 months) and the crossover group (p = 0.289) or between the Rx44 and Rx22 groups (p = 0.33).

In the four-group model, the time to first EDSS progression increased with IFN-β-1a dose and time on treatment (figure 2): 40th percentiles were 24.9 months (placebo/22), 24.2 months (placebo/44), 35.9 months (Rx22), and 42.1 months (Rx44). Only the placebo/22 group had 50% of patients reach the endpoint. Over years 3 and 4 (with “baseline EDSS” reset at the start of year 3), only the placebo/22 and Rx22 groups had 25% of the patients reach the endpoint. Close review of figure 2 shows that the progression rate during years 3 and 4 was slightly faster for Rx44 than for placebo/44 and for Rx22 than for placebo/22, but these differences were not significant. A dose response was observed for time to first progression between Rx44 and Rx22 in years 3 and 4 (p = 0.036).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2. Kaplan–Meier curves for time to confirmed progression in disability for years 1 through 4 (all patients). Proportions of patients are those free from progression. The patients receiving highest cumulative dose of therapy have the lowest rate of progression as opposed to those receiving the lowest dose who had the highest rate of progression. Late treatment was not associated with a catch-up of benefit to early therapy.

In PRISMS-2, 61.7% of patients in the placebo group were free from progression compared with 70.3% for Rx22 and 73.2% for Rx44. At the end of the fourth year (PRISMS-4), 74/161 (46%) of the crossover groups, 88/173 (51%) of the Rx22 group, and 92/164 (56%) of the Rx44 group remained free from progression. These differences were not significant using the ITT sample although a trend was seen for Rx44 compared with crossover (p = 0.070).

To assess cumulative change in disability, the number of confirmed one-point EDSS changes over the duration of the study was examined. Patients in the Rx44 group experienced fewer confirmed EDSS changes over 4 years (0.17/patient/year) compared with the crossover groups combined (0.24/patient/year) and the Rx22 group (0.22/patient/year; p = 0.005 Rx44 versus crossover and p = 0.030 Rx44 versus Rx22); the Rx22 and crossover groups did not differ significantly. The high dose was associated with 29% fewer one-point EDSS changes over 4 years compared with the crossover group. The value 0.24/patient/year combines rates of placebo years with treatment years, making an accurate assessment of treatment effect difficult. The rate of EDSS changes for crossover patients during PRISMS-2 was 0.34/patient/year.

The median IDSS over 4 years was significantly lower for the Rx44 group (138.3 EDSS step-years) than for the crossover groups combined (333.5 EDSS step-years; figure 3). For comparison, the placebo group in PRISMS-2 had a median area under the curve (AUC) of 140 EDSS step-years after 2 years, similar to that of Rx44 after 4 years. The medians for the Rx44 (138.3 EDSS step-years) and Rx22 groups (268 EDSS step-years) did not differ significantly over 4 years, but a dose effect was seen over years 3 and 4, in which median AUC was 0.0 EDSS step-years for Rx44 and 105.0 EDSS step-years for Rx22 (p = 0.005).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3. Median area under the curve for the Expanded Disability Status Scale–time plot. Significant benefit is seen for the Rx44 group compared with the placebo crossover group (p = 0.034) but not for the Rx22 (p = 0.141) group. Approximate 95% CI intervals are based on median values.

Efficacy: MRI-related secondary outcomes.

MRI scans were available for 97% of patients (542/560). Over 4 years, patients in the Rx22 and Rx44 groups had fewer new T2 lesions compared with the crossover groups combined (p < 0.001 in each case), and the Rx44 group had fewer new T2 lesions than the Rx22 group (p < 0.001; table 3). The proportion of scans showing new T2 lesions over 4 years was also less in the Rx44 group than in the Rx22 group (p < 0.001). The proportion of scans showing T2 active lesions was less for crossover patients in both dose groups (p < 0.001) after switching to active drug compared with placebo years.

During years 3 and 4, both the median number of new T2 lesions per patient per scan and the proportion of scans showing new T2 lesions were lower for Rx44 than for either placebo/44 or Rx22 (p < 0.001 in each case). The Rx22 and placebo/22 groups did not differ.

Over 4 years, increases from baseline in BOD were observed in all groups except Rx44 (figure 4), which had a reduction of 6.2% in BOD compared with increases of 3.4% for Rx22, 7.2% for placebo/22, and 9.7% for placebo/44 (p = 0.003 for Rx44 versus placebo/44; p = 0.009 for Rx 44 versus Rx22; and p = 0.113 for Rx22 versus placebo/22). During years 3 and 4, BOD decreased by 2.6% from the month 24 value for placebo/22 and by 5.1% for placebo/44 while it increased by 3.9% for Rx22 and 1.6% for Rx44 (p = 0.024, placebo/22 versus Rx22).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4. Change in burden of disease (median %) over 4 years: analysis of variance on ranks. Change in burden of disease over 4 years is progressively higher with lower total treatment duration/dose. The benefit for the Rx44 group is significant compared with all other treatment groups.

Effect of neutralizing antibodies.

A cut-off of 20 neutralizing units per milliliter (NU/mL) was used to define positive IFN-β NAb titers. Persistent NAb, defined as NAb present at a patient’s final evaluation, were observed in 23.7% of the Rx22 group and 14.3% of the Rx44 group (p = 0.02). Transient antibodies developed in an additional 5.9% of the Rx22 and 4.9% of the Rx44 groups.

To investigate NAb impact over 4 years, efficacy outcomes were evaluated for Rx44 and Rx22 patients, considering patients as NAb-positive throughout the study if they had positive titers at any time. Over 4 years, NAb-positive patients experienced slightly more relapses than NAb-negative patients in both dose groups. During years 3 and 4, relapse rate was 0.50 for Rx44 NAb-negative patients and 0.81 for NAb-positive patients (p = 0.002). Exclusion of patients with NAb from analysis strengthened the dose effect on relapses: 0.76 for Rx22 and 0.65 for Rx44 (p = 0.004). Although NAbs were seen less frequently in the Rx44 group, their impact on relapse rates was greater.

Differences were more apparent for MRI measures. The median number of T2 active lesions was 0.3 for Rx44 NAb-negative patients compared with 1.4 for NAb-positive patients (p < 0.001). Changes from baseline in BOD were assessed for patients with baseline and month 48 scans. In the Rx44 group BOD decreased by 8.5% in NAb-negative patients (n = 83) and increased by 17.6% in NAb-positive patients (n = 26; p < 0.001).

Safety.

Adverse events during the extension were similar to those observed in PRISMS-2 (table 4), and most were mild. Fifty-four patients experienced 67 serious adverse events during years 3 and 4, and the incidence of serious adverse events was similar between groups. One patient in the Rx22 group died after a myocardial infarction. Withdrawals due to adverse events are noted in figure 1.

Common laboratory abnormalities included asymptomatic lymphopenia, thrombocytopenia, and elevated hepatic transaminases, which were more frequent for the Rx44 group than for the Rx22 group (p = 0.07 for lymphopenia, p = 0.02 for thrombocytopenia, p = 0.03 for AST, p = 0.07 for ALT). All cases of thrombocytopenia were World Health Organization grade 1 (less than 25% reduction compared with the lower limit of normal). Over 4 years, one patient from the Rx44 group stopped therapy because of lymphopenia and a further two patients in the same group stopped because of elevated liver enzymes.

As expected, the new onset of adverse events during years 3 and 4 was higher in the crossover groups, particularly for “IFN-type” events such as application site reactions, fatigue, headache, and myalgia. Prevalence figures demonstrated that many patients from the Rx22 and Rx44 groups had ongoing adverse events from PRISMS-2. Comparison of the overall rates of adverse events for the Rx22 and Rx44 groups in years 1 through 4 and in years 3 and 4 suggests that many events, particularly laboratory abnormalities, resolved while therapy continued (table 4).

Injection site necrosis occurred once per 14,100 injections among low-dose patients (Rx22 and placebo/22), and once per 9,300 injections among high-dose patients. No patient discontinued because of necrosis. Seventeen of the 27 patients withdrawing because of adverse events in the extension period reported application site disorders among the events contributing to their discontinuation (one patient from each of the placebo/22 and Rx22 groups, 10 patients from the placebo/44 group, and five patients from the Rx44 group).

The percentage of patients reporting depression at least once was between 23 and 29% during years 3 and 4, including cases with onset during PRISMS-2. In PRISMS-2, depression was reported more commonly in patients receiving placebo than in those receiving IFN-β-1a. There were two suicide attempts during years 3 and 4, one in the placebo/22 group and one in the Rx44 group. One patient in the Rx44 group stopped treatment because of depression.