The prevalence of systemic lupus erythematosus (SLE) varies with age, gender, and ethnicity, and the highest rates occur in young adult women, particularly of Afro-Caribbean origin, who are in peak childbearing years [1–4]. Nephritis complicates SLE in a significant minority of patients and is associated with renal failure and increased mortality. The tendency for people of Afro-Caribbean origin to have a worse prognosis may be due, at least in part, to poor socio-economic status [5]. In the 1950s, patients with class IV nephritis rarely lived longer than 5 years, whereas now more than 80% survive with good renal function for more than 10 years [6].

The World Health Organization (WHO) classification for lupus nephritis is based on the histological appearance, with progressive changes to the glomerulus and tubulo-interstitium with increasing severity (Additional file 1). Milder disease (WHO classes II and IIIa) affects approximately 35% to 50%, whereas more serious classes IIIb, IV, and V affect 45% to 60% [7]. A significant minority of patients with class III disease (focal segmental proliferative glomerulonephritis) show worsening renal function and may progress to class IV lupus nephritis. Class IV (diffuse proliferative glomerulonephritis) usually presents with clinical evidence of renal disease, including oedema, hypertension, sediment, and worsening renal function with proteinuria. Class V (diffuse membranous glomerulonephritis) involves patients with laboratory and clinical features of nephrotic syndrome.

The aim of treatment is first to stop disease progression (induction phase) and then prevent recurrence (maintenance) while minimising the adverse effects. More specifically with lupus nephritis, the aims of treatment are to reduce the risk of end-stage renal disease, reduce renal and extra-renal lupus activity or symptoms, and reduce the mortality risk.

Treatment with immunosuppressive therapy is better than prednisolone monotherapy at preserving renal function, and combination therapy is better still [8–10]. Cyclophosphamide is used mainly to induce remission in patients with proliferative disease and has been considered to produce the best renal outcomes [11]. A systematic review of immunosuppressive treatments for lupus nephritis highlighted some of the problems [10]. These included ovarian failure, affecting 47% of women treated with cyclophosphamide plus steroid, and major infection in 20%. Nor were therapies entirely effective using cyclophosphamide plus steroid; doubled serum creatinine was observed in 24%, end-stage renal disease in 16%, and mortality in 21%. Weaknesses of these estimates included small numbers of trials and patients (maximum 290 in comparisons with steroid alone) and pooling data from patients with different disease severity who were being administered different regimens of immunosuppressive therapy. Few mentioned any blinding, but this is more difficult in long-term comparisons of parenteral and oral therapies.

Mycophenolate mofetil (MMF) is a relatively well-tolerated immunosuppressive agent used after kidney, liver, and heart transplants. MMF selectively and noncompetitively inhibits inosine monophosphate dehydrogenase in the de novo purine synthesis pathway. This enzyme facilitates the conversion of inosine monophosphate to xanthine monophosphate, an intermediate metabolite in the production of guanosine triphosphate. Because MMF results in the depletion of guanosine nucleotides, it impairs RNA, DNA, and protein synthesis [12]. Pharmacokinetics and metabolism of MMF suffer no major perturbations in renal impairment [13], but this may not be the case for severe renal insufficiency.

One year after renal transplantation, MMF at 2 or 3 g daily was at least as effective as azathioprine regimens at maintaining graft survival. Patient survival was similar, with a tendency to higher survival rates at longer duration [14, 15]. MMF may be particularly beneficial in African-Americans [16]. MMF has also been shown to have some benefits over calcineurin inhibitors after liver transplantation [17]. MMF reduced mortality and graft loss compared with azathioprine over three years after heart transplantation in a large randomised trial [18]. Adverse events with MMF have tended to be different from those of other commonly used immunosuppressants. Diarrhoea may be more common [15, 18] whereas ovarian failure is less common with MMF than with cyclophosphamide. The aim of this review was to systematically examine the available evidence concerning MMF in lupus nephritis from randomised trials and observational studies.

QUOROM (quality of reporting of meta-analyses) and MOOSE (meta-analysis of observational studies in epidemiology) guidelines were followed where applicable [19, 20]. Studies of any design and in any language were sought concerning MMF in the treatment of patients with complications of lupus nephritis or SLE. A series of free text searches were undertaken to identify eligible reports from MEDLINE (to September 2005), Cochrane Library (Issue 3, 2005), and PubMed (to October 2006) by using the terms MMF or mycophenolate and the terms lupus nephritis or lupus erythematosus. Additional trials were sought in review articles [9, 10, 21–29] and reference lists of retrieved articles. The company (Aspreva Pharmaceuticals Ltd, Bagshot, Surrey, UK) developing CellCept for lupus nephritis was asked about the availability or knowledge of published or unpublished clinical trials. Authors of papers were not contacted for unpublished reports or additional information from published reports.

For completeness, we included studies in three separate groups: randomised trials, cohort studies with patients with SLE and lupus nephritis, and cohort studies with patients with SLE, only some of whom had lupus nephritis or undefined renal involvement. Full publications and abstracts were allowed. With multiple publications of the same study, we used the largest body of data or the most informative or the most recent (as appropriate). Excluded were reviews with clinical information published in a fuller form elsewhere, studies with purely biochemical (biomarker), pharmacokinetic, or immunological information, and studies in which MMF was used for treating other conditions.

Each randomised trial was scored for quality by using a three-item quality scale [30]. A maximum of five points were awarded to studies according to whether they were randomised, double-blind, and accounted for withdrawals or drop-outs and to whether the methods of randomisation and double-blinding were described and appropriate. No scoring system was used for cohort studies.

Two reviewers extracted information from the trials independently, and disagreement was resolved by consensus. Information extracted included the number of patients treated (per group in randomised trials), demographic information given (age, gender, and ethnicity), definition or classification of lupus nephritis (WHO was used), dosing regimens, duration of therapy, information about patients studied and their diagnoses, study design, the number of patients with efficacy and/or safety outcomes, and discontinuations, especially because of adverse events or lack of efficacy. Efficacy outcomes sought were those of complete or partial response to treatment as defined by the original authors (mainly urine protein excretion, serum creatinine or creatinine clearance, or a combination of these) and subsequent relapse. Adverse events sought included mortality, infection (especially severe infection), leucopaenia, gastrointestinal problems, amenorrhoea (in women), and hospital admission because of adverse events.

For analysis of efficacy, only those studies or trials with patients with both SLE and lupus nephritis were included. Any definition of lupus nephritis, as provided by authors, was accepted. If studies included patients who did not have lupus nephritis, efficacy information was only used for analysis if reported separately for nephritis patients. We used information on all patients for evaluation of adverse events because there was no prior indication that adverse events would be different in patients with SLE with or without nephritis or with different grades of nephritis.

For cohort studies, we calculated the percentage of patients with beneficial or harmful events by using as the denominator the total number of patients in studies mentioning the event. For randomised trials in which MMF was compared with another treatment, relative benefit and relative risk estimates were calculated with 95% confidence intervals by using a fixed effects model [31]. Because they have previously been shown to be unhelpful, heterogeneity tests were not used [32]; homogeneity was assessed visually [33]. Publication bias was not assessed using funnel plots, likewise because these tests have been shown to be unhelpful [34, 35]. Relative benefit or risk was considered statistically significant when the 95% confidence interval did not include 1. The number needed to treat (NNT) and number needed to harm (NNH) with confidence intervals were calculated by the method of Cook and Sackett [36]. NNT or NNH values were calculated only when the relative risk or benefit was statistically significant.

The following terms were used to describe adverse outcomes with respect to harm or prevention of harm:

• The NNT to prevent one event (NNTp). We used this term when significantly fewer adverse events occurred with MMF than with the comparator.

• The NNH to cause one event. We used this term when significantly more adverse events occurred with MMF than with the comparator.

Additional file 2 has details of all the relevant studies found; there were no exclusions after screening, but not all studies were used in analyses. Six randomised trials in eight papers [37–44] reported on 370 patients in total, 197 of whom were treated with MMF. Ten cohort studies [45–54] reported on the use of MMF in 212 patients with lupus nephritis, and 8 cohort studies [55–62] reported on 284 patients with SLE treated with MMF, only some of whom had nephritis or renal problems. For several studies, there were multiple reports; duplicate information was avoided through detailed attention to authors, place of research, design, and study numbers. No study reported results separately for different classes of lupus nephritis or by ethnicity. Ethnicity was inconsistently reported (Additional file 2).

Efficacy

Randomised trials

Results for complete response and complete plus partial response in induction and maintenance therapy are shown in Table 1, in which the comparator was cyclophosphamide. There was a consistent response between studies. Figure 1 plots the proportion of patients in each trial with a complete and partial response with MMF or cyclophosphamide and shows the variability between individual trials.

Outcome	Number of trials	Number of patients	Percentage with MMF	Percentage with cyclophosphamide	Relative benefit or risk (95% CI)	NNT (95% CI)
Efficacy
Complete response	4	266	36	23	1.5 (1.1 to 2.1)	7.6 (4.2 to 43)
Complete or partial response	5	306	66	54	1.2 (1.03 to 1.4)	8.0 (4.3 to 60)
Subsequent relapse	2	102	27	34	0.8 (0.4 to 1.4)
Adverse events						NNTp (95% CI)
Death	5	306	0.7	7.8	0.2 (0.07 to 0.7)	14 (8 to 48)
Hospital admission	2	220	1.7	15	0.1 (0.04 to 0.5)	7.4 (4.8 to 16)
Adverse event discontinuations	3	246	1.6	5.6	0.3 (0.08 to 1.4)
All infections	4	280	39	73	0.5 (0.4 to 0.7)	3.0 (2.3 to 4.4)
Serious infections	4	304	3.9	15	0.3 (0.1 to 0.6)	8.7 (5.5 to 21)
Leucopaenia	3	122	1.6	25	0.1 (0.03 to 0.5)	4.3 (2.9 to 8.3)
Amenorrhea	5	312	1.9	12	0.2 (0.08 to 0.6)	9.5 (6.2 to 20)
Hair loss	3	240	0.0	16	0.1 (0.01 to 0.4)	6.4 (4.4 to 11)
						NNH (95% CI)
Diarrhoea	4	260	16	4.0	4.0 (1.5 to 10)	8.5 (5.3 to 21)

CI, confidence interval; MMF, mycophenolate mofetil; NNH, number needed to harm; NNT, number needed to treat; NNTp, number needed to treat to prevent one event.

MMF was significantly better than cyclophosphamide, with NNTs of approximately 8 for MMF compared with cyclophosphamide for one additional complete or complete plus partial response (that is, for every eight patients treated, one more achieved a response with MMF than with cyclophosphamide). Restricting the analysis to the four induction trials with 266 patients (omitting [41]) improved the NNT to 6.6 (95% confidence interval 3.7 to 30). The proportion of patients with complete or partial response in patients given MMF was 66%. Subsequent relapse after successful induction was reported in two trials, in 14/52 patients on MMF and 17/50 patients on cyclophosphamide.

Cohort studies

Complete or partial response to therapy with MMF was reported in 7 of the 10 cohort studies that recruited only patients with lupus nephritis. Complete or partial response was reported in 80% (Table 2) and failure or no response was reported in 20%. All 10 studies reported some measure of clinical or biochemical improvement. Reduced protein excretion was mentioned in 10, reduced corticosteroid dose in 7, with 3 each mentioning reduced serum creatinine and disease severity or increased serum albumin. Where numerical data were given, reduction in urinary protein excretion appeared to be of a major magnitude; average reductions were often greater than 50%, and reduction in average corticosteroid dose was also approximately 50% or more.

	All cohort studies			Only definite nephritis			Any SLE patients
Outcome	Number of cohorts	Number of events/patients	Percentage with outcome	Number of cohorts	Number of events/patients	Percentage with outcome	Number of cohorts	Number of events/patients	Percentage with outcome
Efficacy
Complete or partial response	7	121/151	80	7	121/151	80	No data
Failure/No response	8	37/176	21	7	30/151	20	1	7/25	28
Adverse events
Adverse event discontinuations	9	40/285	14	4	6/68	8.8	5	34/217	16
Lack of efficacy discontinuations	5	24/235	10	1	4/24	17	4	20/211	9.5
All infections	10	76/327	23	4	18/70	26	6	58/257	23
Serious infections	8	8/185	4.3	4	4/87	4.6	4	4/98	4.1
Gastrointestinal (diarrhoea, nausea, vomiting)	14	122/406	30	8	52/160	33	6	58/257	29
Leucopaenia	8	6/188	3.2	4	1/61	1.6	4	5/127	3.9
Vertigo, dizziness	3	6/81	7.4	No data			3	6/81	7.4
Hair loss	2	5/42	12	1	1/13	7.7	1	4/29	14

SLE, systemic lupus erythematosus.

Meta-analysis of small studies is problematic [63] because of possible inadequacies in design or reporting or because small numbers of patients and events increase the risk of chance findings [64]. The ideal is meta-analysis of high-quality, large, randomised trials [63], but without them we have to make the best of the information available and of all the information available. That is the justification, as here, for including information from both randomised trials and observational studies and (for adverse events) from observational studies in SLE in which only a proportion had lupus nephritis. Results, as here, can be reported separately to support any conclusions in which they are in concordance or to raise doubts about results in which they are not.

The background in lupus nephritis is one of few studies and small numbers. For cyclophosphamide and azathioprine, the mainstay of immunosuppressant therapy, there are at most 200 to 300 patients in randomised comparisons of immunosuppressant plus steroids with steroids alone [10]. In this analysis of MMF compared with cyclophosphamide for induction therapy, there were 197 patients treated with MMF in randomised trials. Ten cohort studies contributed information from an additional 212 patients.

Despite these differences, the main findings of both randomised trials and cohort studies were much the same. Complete plus partial response with MMF occurred in 66% of patients in randomised trials and in 80% in cohorts, a similar proportion given the relatively small numbers. Mortality with MMF in randomised trials was of the same order of magnitude as in cohort studies (0.2%), with only two deaths. Adverse event rates were also similar (Figure 2), but numbers of events were small in many cases. For instance, hair loss in cohort studies depended on only five cases in two studies.

Adverse event rates in lupus nephritis patients were similar to those in patients with SLE (Table 2) and also to those found in renal [14, 15] and heart [18] transplants. This provides additional reassurance of safety and tolerability of MMF, often at high doses (like the 3 g daily in the heart transplant trial) and over a treatment duration of up to 3 years. Despite this, there is the possibility of residual bias in observational studies and lack of blinding in randomised trials. The amount of information for MMF is greater than for cyclophosphamide and azathioprine in this indication, and trials, if anything, are somewhat better, but this is no defence against features of study design which we know may mislead [65].

Although the analysis provides information about efficacy and safety on average, it does not help determine which patients will respond best to MMF or other immunosuppressants. In most studies, either the requirement for entry was reasonable renal function (creatinine clearance greater than 30 ml/minute [44] or 20 ml/minute [39]) or renal function was relatively unimpaired (mean creatinine clearance was 80 ml/minute [43]). Patients with more severe renal disease were excluded, so we do not have information about efficacy or harm in the sickest patients, nor do we have much information about long-term maintenance rather than induction therapy.

Moreover, information was not reported separately for different racial groups or for different WHO classes of lupus nephritis, both of which are important considerations for determining appropriateness of therapy. Practitioners may be frustrated by the inability of a review to provide results for subgroups or better definitions of outcomes, but analysis can be performed only on average results unless individual results are available. The ideal would be an individual patient analysis based on collaboration between trialists.

Within these limitations, the evidence is that daily oral MMF is more effective than pulsed intravenous or oral cyclophosphamide. It produced more complete responses and complete plus partial responses, and the absolute difference, equivalent to an NNT of 8, was clinically useful. This better efficacy came with lower mortality, fewer hospital admissions, and lower rates of several adverse events, including severe infection. In a patient population of young women of childbearing age, it produced almost no cases of amenorrhoea and no cases of hair loss in randomised trials. Evidence from renal transplantation [66] is that MMF may also reduce cardiovascular risk relative to other immunosuppressants, and there are some theoretical reasons for this [67].

This is the first systematic review of MMF in lupus nephritis, but another is under way [68]. The previous systematic review by Flanc and colleagues [10] examined cyclophosphamide plus steroid versus steroid alone, often in older studies, and was limited, as here, by small numbers of events and patients. In that analysis, mortality with cyclophosphamide plus steroid was 21%, much higher than 1% or less with MMF, and approximately double the 7.8% mortality with cyclophosphamide plus steroid in the randomised trials in this review (Table 1). Serious infections with cyclophosphamide plus steroids occurred in 16% in the review of Flanc and colleagues (15% here), but amenorrhoea was higher (47%).

These results form a basis on which to plan future studies and provide a guide for the use of MMF in lupus nephritis until results of larger studies are available. Systematic reviews do not just provide an answer to a particular research question but should also highlight future research agendas. These would include reporting results not just for induction and maintenance, but for different classes of lupus nephritis and for populations with different risk factors such as ethnicity. At least one randomised trial is under way [69], with results from the induction phase expected in 2007, but it will be several years before the results of the maintenance phase are available.

MMF produced more complete responses and complete plus partial responses, and the absolute difference was clinically useful. There are limitations to the existing data, not the least of which is to short-term results relative to the very long course of lupus nephritis.