Description of studies

For the first version of this review 208 articles identified by the search strategy, eighteen articles were selected provisionally, on the basis of the inclusion criteria mentioned above.
We excluded eleven studies after reading the full published papers: two studies were open‐label (Rudick 1998; Herndon 1999); in three studies participants were treated with natural interferon‐alfa (IFN‐alfa) (Knobler 1984; AUSTIMS 1989; Milanese 1990); in one study only immunological, not clinical, outcomes were reported (Hirsch 1986); one study was a dose‐comparison trial of interferon beta‐1a without a placebo group (Pozzilli 1997); one study was performed on progressive MS participants and included only two participants with relapsing‐remitting MS (Rudge 1995); one study reported data only relating to the placebo group of a RCT on interferon beta‐1a (IFNB‐1a) (Simon 1999) and in one study (Jacobs 2000) participants were treated after a first exacerbation, prior to a definite diagnosis of MS. Finally, one study was excluded because neither blinding criteria nor clinical outcomes were clearly described (Schwartz 1997). (See table of excluded studies).

During the update of the review we first identified two studies that met the inclusion criteria. One of them (Granger 2003) was excluded because it is a secondary analysis of a study already included in this review (The MSCRG 1996) using FIM instrument instead of EDSS to assess disease progression.

Thus eight trials contributed to this review: the earliest was published in 1993 and the most recent in 2003 (Knobler 1993; IFNB MS Group 1993; Durelli 1994; The MSCRG 1996; The PRISMS 1998; Myhr 1999; The OWIMS 1999; Polman 2003). A total of 1301 experimental participants (in the higher dose arms which were compared with placebo treated participants), including 658 treated with interferon and 643 randomised to placebo, were included in this review. One multicenter study of subcutaneous IFNB‐1a (The PRISMS 1998) accounted for 371 (30.5%) participants while a study of intramuscular IFNA‐2a (Durelli 1994) accounted for only 20 (1.6%) participants. All studies included only relapsing‐remitting MS participants. (See table of included studies).

THE AGENTS COMPARED WITH PLACEBO WERE:

• IFNA‐2a: 9.0 MIU self‐administered intramuscularly every other day for six months (Durelli 1994), 4.5 or 9.0 MIU administered by study physician subcutaneously three times weekly for six months (Myhr 1999). The dose of 9.0 MIU was used for comparison.

• IFNB‐1a: 6.0 MIU administered once weekly by the intramuscular route for 104 weeks, by study nurses or local health professionals under the supervision of study personnel (The MSCRG 1996), 6.0 or 12.0 MIU self‐administered subcutaneously once a week for 48 weeks (The OWIMS 1999), 6.0 or 12.0 MIU self‐administered subcutaneously three times weekly for two years (The PRISMS 1998). In the last two studies the dose of 12.0 MIU was used for comparison with placebo. In one study (Polman 2003), 0.06, 0.6, or 6 MIU were administered every day by the oral route.

• IFNB‐1b: 1.6 or 8.0 MIU self‐administered subcutaneously every other day for two years (IFNB MS Group 1993); 0.8 or 4.0 or 8.0 or 16.0 MIU self‐administered subcutaneously three times weekly for three years (Knobler 1993). The doses of 8.0 MIU and 16.0 MIU were used in the first and second study respectively.

EXCLUSION CRITERIA
Exclusion criteria were specified in seven trials (IFNB MS Group 1993; Durelli 1994; The MSCRG 1996; The PRISMS 1998; Myhr 1999; The OWIMS 1999; Polman 2003). The first trial (IFNB MS Group 1993) excluded participants who had been treated with ACTH or prednisone during the 30 days prior to entry or who had received any previous treatment with azathioprine or cyclophosphamide. The second trial (Durelli 1994) excluded participants who had other severe illnesses, participants who had been treated previously with interferon or cytostatic drugs or with ACTH or corticosteroids during three months before entry, and women who were pregnant. The MSCRG (The MSCRG 1996) excluded patients with prior therapy with immunosuppressants, interferon or ACTH/steroids within two months before study entry; patients with infection or other active disease; patients with progressive MS; those who were pregnant or breastfeeding and those unwilling to practice contraception during the study. The fourth study (The PRISMS 1998) excluded patients who had previously received interferon, lymphoid irradiation, cyclophosphamide or immunomodulatory or immunosuppressive drugs in the 12 months prior to the study. The fifth study (Myhr 1999) excluded patients who were pregnant or breastfeeding; those who were unwilling to practice contraception; those previously treated with interferon or immunosuppressants during the previous year or steroids in the month before randomisation; patients experiencing significant concomitant disease and those who were in the progressive phase of the disease. The OWIMS study (The OWIMS 1999) excluded patients who were pregnant or lactating, patients who were concomitantly ill with other conditions and those who had previously been treated with interferon, cyclophosphamide, lymphoid irradiation or any immunosuppressive or experimental therapy in the preceding twelve months. In the trial testing oral interferon (Polman 2003) patients were excluded if any previous parenteral IFN treatment had been terminated due to inefficacy. In only one study (Knobler 1993) exclusion criteria were not reported.

All trials applied diagnostic criteria for MS (Poser 1983).

Data on the time period of patient recruitment were missing in the Myhr study (Myhr 1999) and in the Polman study (Polman 2003). The scheduled follow‐up periods were six months (Durelli 1994), 48 weeks (The OWIMS 1999), 12 months (Myhr 1999), two years (IFNB MS Group 1993; The MSCRG 1996; The PRISMS 1998) and three years (Knobler 1993).

OUTCOMES
A clinical outcome was the primary outcome measure in five studies: clinical relapse (IFNB MS Group 1993; Knobler 1993; Durelli 1994;The PRISMS 1998) or disability progression (The MSCRG 1996).
Magnetic resonance imaging was used as the primary outcome measure in two studies (The OWIMS 1999; Myhr 1999) in which clinical outcomes were reported as the secondary outcomes. In the Polman study (Polman 2003) primary outcome was the cumulative number of newly active lesions seen on brain MRI scans; secondary outcomes included volume of enhancing lesions on TI‐weighted images each month, lesion volume on T2‐weighted images at month 3 and 6 and safety measures.

CLINICAL OUTCOME
In five trials (Knobler 1993; Durelli 1994; The PRISMS 1998; Myhr 1999; The OWIMS 1999) data were available on the number of participants who continued to experience exacerbations during the first treatment year; and in three trials (IFNB MS Group 1993; The MSCRG 1996; The PRISMS 1998) exacerbations were reported over the two years following randomisation. Comparable definitions of relapse were used in the trials and included:

• acute or subacute objective deterioration of neurological status attributable to the disease, lasting at least 24 hours in the absence of fever and followed by complete or partial resolution (described by Poser 1983)(Durelli 1994);

• a new symptom or worsening of an old symptom accompanied by a new neurologic abnormality, lasting at least 24 hours in the absence of fever and preceded by stability or improvement for at least 30 days (described by Schumacher 1968) (IFNB MS Group 1993; Knobler 1993; The PRISMS 1998; Myhr 1999; The OWIMS 1999);

• a new symptom or worsening of an old symptom of at least 48 hours, which followed clinical stability or improvement of at least 30 days duration (described by Jacobs 1995) (The MSCRG 1996).

Definitions of disease progression were also similar:

• deterioration from baseline = 1.0 point or more on the Expanded Disability Status Scale (EDSS) persisting for at least six months (The MSCRG 1996);

• increase in EDSS of at least 1.0 point sustained over at least three months (IFNB MS Group 1993; The PRISMS 1998).

In two trials (IFNB MS Group 1993; The PRISMS 1998) the mean disability score (EDSS) was reported as a change of disability with respect to the baseline.

The number of patients who were unable to walk without aid (EDSS score greater than 5.5 ) at the end of the follow‐up period was not reported in any trial.

Time from randomisation to first exacerbation was reported as the median time.

Two studies (Durelli 1994; Myhr 1999) reported the number of participants who underwent steroid treatment in the first year after randomisation and one study (The PRISMS 1998) reported participants with steroid treatment within two years of randomisation.

Hospitalisation within two years was included as an outcome in two studies (Durelli 1994; The PRISMS 1998).

SIDE EFFECTS AND ADVERSE EVENTS
Data concerning the numbers of participants with side effects or adverse events (attributed to interferon during the study according to the authors) were reported in six trials (IFNB MS Group 1993; Durelli 1994; The MSCRG 1996; The PRISMS 1998; Myhr 1999; The OWIMS 1999). One study (Knobler 1993) did not report data for each treatment group, but only gave cumulative numbers for groups that had received different doses of interferon. Finally, it was not possible in any of the trials to assess the length of time for the development of side effects or adverse events, either due to short follow‐up periods or because of incomplete reporting.

MAGNETIC RESONANCE IMAGING (MRI)
Seven trials reported MRI data using a number of different assessments, measures and time of follow‐up. One study (IFNB MS Group 1993) reported MRI data for only 60% of the randomised participants. In another study (The OWIMS 1999) the authors reported the results of MRI, but without the numbers of participants who underwent complete MRI evaluations.

Risk of bias in included studies

RANDOMISATION
Concealment of treatment allocation had been adequate in three trials (Durelli 1994; The PRISMS 1998; The OWIMS 1999) and unclear in the other five (Knobler 1993; IFNB MS Group 1993; The MSCRG 1996; Myhr 1999; Polman 2003).

BASELINE PARTICIPANTS' CHARACTERISTICS
All participants were clinically stable at study entry except in one study which did not report whether or not participants were in a stable phase (The PRISMS 1998). Equivalence between treated and control participants for age, disability score at entry and frequency of exacerbations in the pre‐study period was confirmed in all studies except in one which did not report clinical characteristics for each arm (The MSCRG 1996). Disease duration was longer in treated than in control participants in three studies (IFNB MS Group 1993; The PRISMS 1998; Myhr 1999).

BLINDING
All trials were intended to be double‐blind. However the well documented side effects of interferon injection, mainly injection‐site reactions and influenza‐like symptoms, make it likely that patients could become unblinded during trials. Analysis of blinding in two studies (IFNB MS Group 1993; The MSCRG 1996) identified a strong tendency for treated patients to become unblinded. Specifically, 80% of participants in the 8.0 MIU IFNB‐1b arm, 51% in the 1.6 MIU IFNB‐1b arm and 30% in the placebo arm had correctly guessed their treatment at the end of follow‐up (IFNB MS Group 1993). Thus many, if not most, treated patients had become aware of the treatment they were receiving during the course of the trial, and these trials should be regarded as single‐blind.

PARTICIPANTS LOST TO FOLLOW‐UP
Overall, 246 (19%) participants were excluded after randomisation or were lost to follow‐up (Table 2). In one study (The MSCRG 1996), 73 (46%) of the 158 randomised participants in the treatment group and 56 (39%) of the 143 in the control group did not complete the scheduled two years of follow‐up because the study ended prematurely. In this study, the occurrence of acute exacerbations or disease progression were not reported at one year; at two years these primary outcomes were available for only 57% of the randomised participants. In one trial (IFNB MS Group 1993), withdrawals and losses to follow‐up were agglomerated into a single figure: 48 participants (19%).
Participant‐ or investigator‐ perceived worsening and side effects or adverse events were the most frequent causes of withdrawn or loss to follow‐up in all trials.

INTENTION‐TO‐TREAT ANALYSIS
Although an intention‐to‐treat analysis was mentioned in seven trials (IFNB MS Group 1993; Durelli 1994; The MSCRG 1996; The PRISMS 1998; Myhr 1999; The OWIMS 1999; Polman 2003) in most cases patients who were withdrawn and lost to follow‐up had been excluded from the analyses.

Effects of interventions

PRIMARY OUTCOMES

(1) THE NUMBER OF PARTICIPANTS WHO CONTINUED TO EXPERIENCE EXACERBATIONS DURING THE FIRST YEAR OF TREATMENT

This information was available from five trials (Knobler 1993; Durelli 1994; The PRISMS 1998; Myhr 1999; The OWIMS 1999) and 667 participants representing 51% of those included in this review. Eighty five per cent of these participants were from two trials (The PRISMS 1998; The OWIMS 1999). Overall, the risk of new exacerbations in participants receiving interferon was 50% and the risk of new exacerbations in participants not receiving interferon was 68%. There was significant heterogeneity among the trials (chi 2 = 8.54, df = 4, p = 0.07). Using a random effects model the pooled relative risk with interferon administration was 0.73, 95% CI 0.55 to 0.97, p = 0.03, and the pooled risk difference was ‐23%, 95% CI ‐8% to ‐39% which means that if 100 MS patients were treated with interferon, 23 (95% CI 8 to 39) would be free from exacerbations during the first year.

Univariate meta‐regression analysis, performed by STATA, identified that the length of interferon administration (number of weeks up to one year) and dose (interferon MIU/week) explained the heterogeneity between the studies. We found no evidence that the type of interferon administered had any effect on the outcome. A subgroup analysis based on the five trials identified little benefit with interferon administered for six months (relative risk 0.82, 95% CI 0.66 to 1.02, p = 0.08) (Durelli 1994; Myhr 1999; The OWIMS 1999) than treatment for one year (relative risk 0.69, 95% CI 0.59 to 0.80, p < 0.001) (Knobler 1993; The PRISMS 1998). In four of the trials, the exception was (Durelli 1994), the participants were followed at least for one year. In four of the five studies, participants had MS for five to seven years at randomisation; in the other study (Durelli 1994) mean disease duration was nine years. All participants had a disability score (EDSS) at randomisation between 0 and 5.5.

The results of the Polman study (Polman 2003) did not indicate a benefit for oral interferon‐1a: approximately two‐thirds of patients in each group remained relapse free at 6 months follow‐up.

(2) THE NUMBER OF PARTICIPANTS WHO CONTINUED TO EXPERIENCE EXACERBATIONS DURING THE FIRST TWO YEARS OF TREATMENT

Data from three trials (IFNB MS Group 1993; The MSCRG 1996; The PRISMS 1998) involving 919 participants (71% of participants included in this review) were available for this outcome measure. The pooled relative risk of exacerbations despite interferon administration was 0.80 (95% CI 0.73 to 0.88, p < 0.001), a decrement of 20% in the treatment group compared to the control group. There was no heterogeneity between the three trials (chi2 = 0.43, df = 2, p = 0.81). Overall, the risk of exacerbations during the first two years in participants receiving interferon was 55% and the risk in placebo‐treated participants was 69%. The pooled risk difference was ‐14% (95% CI ‐8% to ‐19%) using a fixed effect model.
The reduction in exacerbations disappeared if interferon‐treated participants who dropped out were assumed to have had exacerbations (worst case scenario). In this case, there was a significant heterogeneity between the trials (chi2 = 46.15, df = 2, p < 0.001). This was explained by the fact that in one study (The MSCRG 1996) 43% of the randomised participants did not contribute to the two year data. Using a random effects model the pooled relative risk of exacerbation for participants treated with interferon was 1.11, 95% CI 0.73 to 1.68, p = 0.6.

(3) THE NUMBER OF PARTICIPANTS WHOSE CONDITION PROGRESSED DURING THE FIRST TWO YEARS OF TREATMENT

Data from three trials (IFNB MS Group 1993; The MSCRG 1996; The PRISMS 1998) involving 919 (71%) participants were available for this outcome. Overall, MS progressed in 20% of the participants in the interferon arm and 29% in the placebo arm over two years. There was no heterogeneity between the three trials. This result underlined a benefit of interferon in reducing progression of the disease (relative risk 0.69, 95% CI 0.55 to 0.87, p = 0.002). It implies that for every 100 people treated, 9 (95% CI 3 to 14) would be expected to remain stable over two years. However, the assignment of dropouts (to progression or not) was essential to the demonstration of efficacy. If interferon‐treated participants who dropped out were assumed to have progressed (worst case scenario) there was a significant heterogeneity between the trials (chi2 = 30.99, df = 2, p < 0.001) explained by the MSCRG study (The MSCRG 1996) and the treatment no longer produced a significant benefit (relative risk 1.31, 95% CI 0.60 to 2.89, p = 0.5).

(4) MEAN CHANGE IN DISABILITY SCORE (Expanded Disability Status Scale) AT TWO YEARS

This outcome was available from only two trials (IFNB MS Group 1993; The PRISMS 1998) involving 618 (51%) participants included in this review. The effect of the treatment on patient disability (as assessed by Expanded Disability Status Scale) was statistically significant (weighted mean difference = ‐0.25, 95% CI ‐0.05 to ‐0.46, p = 0.01). There was no significant heterogeneity between the two studies for this outcome. It should be emphasised that this result is of questionable clinical importance because it would be impossible to measure this very low degree of change on the Expanded Disability Status Scale in the clinical practice.

(5) THE NUMBER OF PARTICIPANTS WHO WERE UNABLE TO WALK WITHOUT AID (Expanded Disability Status Scale greater than 5.5) AT TWO YEARS
No data were available for this outcome.

SECONDARY OUTCOMES

(a) TIME FROM RANDOMISATION TO THE FIRST EXACERBATION
No data were available.

(b) TIME FROM RANDOMISATION TO PROGRESSION IN DISABILITY
No data were available.

(c) THE NUMBER OF PARTICIPANTS WHO UNDERWENT STEROID TREATMENT

This information was available during the first year from only two small trials (Durelli 1994; Myhr 1999) involving 85 (7%) participants included in this review. These two trials differed in duration of follow‐up (six months in Durelli and 12 months in Myhr). It is not surprising to find that the results of these two trials were different from one another: the first study showing a reduction (not significant) in the use of steroids in participants receiving interferon and the second study not confirming this beneficial effect.

The frequency of steroid administration over the first two years from randomisation was available from only one study (The PRISMS 1998) and 371 (29%) participants included in this review. The result was significant (relative risk 0.70, 95% CI 0.56 to 0.87, p = 0.001).

(4) THE NUMBER OF PARTICIPANTS WHO WERE HOSPITALISED OVER TWO YEARS

This information was available from only two trials (Durelli 1994; The PRISMS 1998) involving 391 (30%) participants included in this review. Again there was heterogeneity between these two trials (chi2 = 3.48, df = 1, p 0.06). There was no significant reduction in the frequency of hospitalisation between participants treated with interferon and those treated with placebo (relative risk 0.44, 95% CI 0.08 to 2.36, p = 0.3). Data from another study (IFNB MS Group 1993) were available only after three years. Participation in the third year was optional in that study.

(5) THE NUMBER OF PARTICIPANTS WITH SIDE EFFECTS OR ADVERSE EVENTS ATTRIBUTED TO TREATMENT

The definitions of side effects or adverse events, as reported by the study authors, are listed in Table 3 .

(a) PARTICIPANTS WHO HAD CLINICAL SIDE EFFECTS OR ADVERSE EVENTS DURING TREATMENT

Not all results concerning side effects and adverse events were described, monitored and reported in a uniform manner. There were some deviations among studies in the timelines against which side effects were measured and in the definition of composite syndromes, such as "flu‐like syndrome" or "flu‐like symptoms". This likely accounts for the heterogeneity (chi2 = 12.37, df = 3, p = 0.006) in the observation of this syndrome, which was interpreted rather vaguely. Heterogeneity was not encountered when its component symptoms (fever, myalgia, fatigue and nausea) were considered individually.

A constellation of "flu‐like symptoms" was reported by four studies (IFNB MS Group 1993; The MSCRG 1996; The PRISMS 1998; The OWIMS 1999) involving 1117 (86%) participants included in this review. These symptoms were experienced by 48% of participants receiving interferon and 28% of those receiving placebo: the relative risk was 1.70, 95% CI 1.23 to 2.37, p = 0.001) using a random effects model. All authors reported that flu‐like symptoms diminished after three months of treatment. One author (Durelli 1994) reported that these symptoms were experienced by "several participants" receiving interferon, but usually disappeared the following morning after injection. Another author (Myhr 1999) did not give the frequency of this side effect but reported that one participant in the higher dose interferon group discontinued treatment because of "flu‐like" symptoms and an exacerbation.

Fever and myalgias or arthralgias were described as isolated symptoms in six of the studies (IFNB MS Group 1993; Durelli 1994; The MSCRG 1996; The PRISMS 1998; Myhr 1999; The OWIMS 1999) that included 1199 (92%) participants considered in this review. Fever occurred in 28% of participants receiving interferon and in 14% receiving placebo; myalgias or arthralgias in 26% of treated and 13% of placebo participants. The risks for fever (relative risk 2.01, 95% CI 1.60 to 2.52, p < 0.001) and myalgias or arthralgias (relative risk 1.92, 95% CI 1.52 to 2.43, p < 0.00001) were increased in the treated group.

Fatigue was reported in five of the studies (Durelli 1994; The MSCRG 1996; The PRISMS 1998; Myhr 1999; The OWIMS 1999), involving 952 (73%) participants included in this review. Fatigue occurred in 17% of participants treated with interferon and 12% treated with placebo. The risk for fatigue was 1.37 (95% CI 1.01 to 1.88, p = 0.05). In one study (IFNB MS Group 1993) fatigue was the cause for the withdrawal of three participants receiving interferon.

Information on the frequency of nausea and vomiting was available in two studies (The MSCRG 1996; Myhr 1999) involving 363 (28%) participants included in this review. Participants receiving interferon were at increased risk for this side effect compared to controls (relative risk 1.59, 95% CI 1.11 to 2.28, p = 0.01) using a fixed effect model.

Headache was reported as an isolated symptom in five studies (Durelli 1994; The MSCRG 1996; The PRISMS 1998; Myhr 1999; The OWIMS 1999) involving 952 (73%) participants. Headache was experienced by 50% of participants taking interferon and 42% taking placebo, and was the reason for one participant withdrawing from one study (IFNB MS Group 1993). The relative risk was 1.16, 95% CI 1.02 to 1.33, p = 0.02, using a fixed effect model.

Injection site reactions: in five studies, interferon was given by the subcutaneous route and in four of these (IFNB MS Group 1993; The PRISMS 1998; Myhr 1999; The OWIMS 1999) the proportion of participants with injection site reactions was reported: 62% in the participants receiving interferon and 14% in those receiving placebo. There was significant heterogeneity between these studies (chi2 = 17.75, df = 2, p < 0.0001). This is probably due to the different definitions of injection site reaction and the clinical ascertainment of cutaneous reactions to interferon injection. The relative risk of such reactions, estimated by random effects model, was greater in treated participants (5.57, 95% CI 2.33 to 13.29, p < 0.0001). The severity of skin reactions was not specified in any of the included trials. Only in two studies which reported them, were there comments on the presence or absence of skin necrosis at the site of injection. This occurred in "1 to 3% of participants" (IFNB MS Group 1993) or not at all (The OWIMS 1999). In two studies, interferon was given intramuscularly (The MSCRG 1996; Durelli 1994). In the first of these, skin reactions were reported in 10 to 15% of participants in both treated and control groups and in the second study in none of the participants. The possibility of inadvertent subcutaneous injection was not mentioned.

Hair loss was only described in two studies (Durelli 1994; Myhr 1999) and occurred in 36% of participants in the interferon group and 2% in the placebo group: an increased risk of 9.78, 95% CI 1.98 to 48.27, p = 0.005). In the Myhr study, this symptom affected 40% of interferon participants over the first six months of treatment.

Dysesthesias, paresthesias and weakness: in one study (Myhr 1999) one participant (1/32) in the control group experienced dysesthesias or paresthesias and six participants (4/30 in the interferon group and 2/32 in the placebo group) had weakness.

Depression was assessed and reported in six trials (IFNB MS Group 1993; Durelli 1994; The MSCRG 1996; The PRISMS 1998; Myhr 1999; The OWIMS 1999) and 1199 participants representing 92% of those included in this review. Depression was reported in 16% of participants and did not differ between the interferon and control groups. Suicide or attempted suicide was reported by three studies (IFNB MS Group 1993; The MSCRG 1996; The PRISMS 1998) and occurred in seven participants. These events did not differ between the interferon and placebo participants. Only two studies reported data on anorexia (Durelli 1994; Myhr 1999) which was present in three participants.

Muscle hypertonia was reported in one small study (Knobler 1993) and was present at the same frequency in treated and control groups.

Hypertension requiring drug treatment was not reported in any trials included in this review.

(b) PARTICIPANTS WHO HAD ABNORMAL VALUES DURING TREATMENT

Lowered hemoglobin levels values were reported in three studies (Durelli 1994; The MSCRG 1996; Myhr 1999) and 383 (29%) participants included in this review. This event occurred more frequently in participants receiving interferon than controls (3.5% versus 1%) but the increased risk was not statistically significant.

Leukopenia was reported in five studies (IFNB MS Group 1993; Durelli 1994; The MSCRG 1996; The PRISMS 1998; Myhr 1999) and 1004 (77%) participants. It was found in 6% of participants in the interferon group and 0.6% in the control group, with a significantly increased risk for treated participants (relative risk 6.47, 95% CI 2.43 to 17.20, p < 0.001).

Lymphopenia was reported in two studies (IFNB MS Group 1993; The PRISMS 1998) and 618 (48%) participants included in this review. It occurred in 27% of participants receiving interferon and 14% receiving placebo, a significantly increased risk for treated participants (relative risk 2.16, 95% CI 1.01 to 4.64, p = 0.05, using a random effects model.

Thrombocytopenia was described in three studies (Durelli 1994; The MSCRG 1996; Myhr 1999) and 383 (29%) participants. The risk for this side effect was greater in treated participants in only one of these studies (Myhr 1999). Myhr reported that the proportion of participants experiencing thrombocytopenia was 23% in the interferon group and 3% in the placebo group.

The values of two hepatic enzymes in blood were reported. Aspartate aminotransferase (AST) was reported in three studies (IFNB MS Group 1993; The MSCRG 1996; The PRISMS 1998) and 919 (71%) participants, and alanine aminotransferase (ALT) in four studies (IFNB MS Group 1993; The MSCRG 1996; The PRISMS 1998; Myhr 1999) and 981 (75%) participants. Increased AST and ALT occurred more frequently in participants who received interferon (4% and 9%) than controls (1% and 3%). Participants treated with interferon had a higher risk both for increased AST (relative risk 2.83, 95% CI 1.14 to 7.06, p = 0.03) and ALT (relative risk 3.57, 95% CI 1.98 to 6.43, p < 0.0001, using a fixed effect model).

Neutralizing antibodies (NAB) against interferon developed in approximately 20% (83/421) of the treated participants in whom this parameter was assessed (IFNB MS Group 1993; Knobler 1993; The MSCRG 1996; The PRISMS 1998; The OWIMS 1999; Myhr 1999). The nature of the biological assays was different and the definitions of positive titres were variable between the studies. When assays were performed the timing and sample sizes were often not reported. Antibodies developed most commonly in the first treatment year. Neutralizing antibodies were only rarely reported in placebo‐treated participants.

(6) MAGNETIC RESONANCE IMAGING (MRI)

There was a significant evolution in the application of MRI technology in the decade in which these trials were conducted and it was not possible to make comparisons across the trials from published data. Different measures of MRI were used and reported in the included trials, often on small samples of randomised participants and at different times of follow‐up. For these reasons, we therefore present a qualitative summary of MRI findings (Table 4). In each trial, lesion activity was measured somewhat differently, commonly without reported measures of variance. An analysis of changes in T2 burden was possible only for the trials of interferon beta 1b (IFNB MS Group 1993) and the PRISMS (The PRISMS 1998).

Unpublished data (mean change in T2 burden, standard deviation) were provided by the sponsor of these two trials who reported a significant reduction in T2 burden aggregating both studies (weighted mean difference ‐24.67, 95% CI ‐33.84 to ‐15.50, p < 0.001). Numerical data on this aspect were not available for the MSCRG study, in which a significant effect on the change in T2 burden was not found at the end of the second year of the study (The MSCRG 1996).

Data on gadolinium enhancing lesions were available from two studies (The MSCRG 1996; Myhr 1999) after one year of treatment and from one study after two years of treatment on a small sample of randomised participants (The MSCRG 1996). A reduction in gadolinium enhancement was evident at one year but not at two years. In individual studies, treatment with recombinant interferon was associated with reductions in other MRI measures, such as the number of new T2 or enlarging lesions.

No significant differences in new active lesions were observed between the oral interferon‐1a and control groups (Polman 2003).

NUMBERS NEEDED TO TREAT

The estimates of numbers needed to treat to prevent one patient having at least one exacerbation at 1 year in relation to the baseline risks are shown in Table 5. This figure shows, for example, that nine patients needed to be treated to prevent one patient having at least one exacerbation at 1 year when the risk of recurrence was 40% in the multiple sclerosis population.