The addition of rituximab to standard chemotherapy produced better overall response, in patients with CD20-positive malignant lymphomas, without increasing the risk of severe infections. Some of the review processes were not fully reported and overall response was not defined, but it seems likely that the results of this review are reliable and reflect the balance of the evidence available.

To determine whether the addition of rituximab to standard chemotherapy increased the risk of severe infections in adults undergoing induction therapy for CD20-positive malignant lymphomas.

PubMed, EMBASE and Cochrane Central Register of Controlled Trials (CENTRAL) were searched for papers published in English, Spanish, French, German or Italian up to 31 July 2010. Search terms were reported for all databases; references and similar systematic reviews from the previous five years were checked for eligible studies.

Randomised controlled trials (RCTs) of adults with CD20-positive malignant lymphomas were eligible for inclusion if they compared identical chemotherapy regimens with or without rituximab and reported the infection rates. Trials of maintenance purging and sequential treatment, and those that included other monoclonal antibodies in addition to rituximab were excluded.

Almost all of the included trials were phase III efficacy trials. Most included only treatment-naive patients, while a few focused on refractory or relapsed patients. The patient mean age ranged from 50.0 to 69.2 years, and trials included patients with either indolent lymphoma or aggressive lymphoma. HIV status where reported was usually negative. Rituximab was always administered at 375mg per m² and was given either on the same day or the day prior to chemotherapy. The number of cycles of rituximab varied from three to eight.

Study selection was carried out by two independent reviewers and the reasons for rejection were recorded. Where decisions differed, the paper was reassessed and a consensus achieved.

The risk of bias was assessed for the following components: sequence generation, allocation concealment, blinding of participants and assessors, attrition rates (less than 20%) and the use of intention-to-treat analysis. Funding and potential conflicts of interest were recorded. Trials were considered to be: at low risk of bias if all criteria were met and reported; at high risk of bias if one or more criteria were not met; or at unclear risk of bias if one or more items were not reported.

The Grades of Recommendation, Assessment, Development and Evaluation Working Group (GRADE) approach was used to rank the quality of evidence for each outcome as high, moderate, low or very low. The level of evidence was downgraded based on risk of bias, imprecision of results, indirectness of the evidence, or risk of publication bias.

The authors did not report how many reviewers assessed the risk of bias or determined the quality of the evidence.

The baseline characteristics, type of intervention, and outcomes, including adverse events occurring during treatment, were extracted. Adverse events were graded according to the National Cancer Institute (NCI)'s Common Terminology Criteria for Adverse Events, and remission was defined according to international criteria.

Data on the occurrence of grade three and four infections, incidence of infection types, incidence of febrile neutropenia, incidence of infection-related death, and incidence of grade three or four leucopenia, granulocytopenia, or lymphopenia were extracted. All outcomes were binary and were calculated as risk ratios, with associated standard error and 95% confidence intervals.

Authors were contacted for additional data and clarifications as necessary. The paper did not report how many reviewers extracted the data.

Risk ratios were pooled using fixed-effect inverse variance meta-analysis, where outcomes were reported by more than two trials. Where at least four trials reported the outcome, publication bias was assessed in funnel plots and the Egger test.

The meta-analyses for the risk of infections, leucopenia and granulocytopenia included only trials reporting events graded as 3 or 4 according to NCI criteria. Outcomes that could occur more than once per patient were meta-analysed only if patient-level data were reported (risks or rates).

Heterogeneity was assessed based using Ι² and assuming that less than 20% was negligible, 20% to 49.9% was moderate, and more than 50% was strong. Where moderate-to-strong heterogeneity was noted, subgroup analysis was planned for dichotomous covariates (lymphoma type, line of treatment, and HIV status) and simple meta-regression was used for other covariates (mean age, maximum and minimum number of rituximab cycles). Where heterogeneity persisted a random-effects estimate was calculated for comparison.

Sensitivity analyses were used to explore how robust the findings were to changes in type of funding, methods used to manage incomplete data, and randomisation and allocation concealment.

Seventeen RCTs reported in 16 papers were included, with 5,259 patients. Five RCTs enrolled 150 or less patients, six enrolled between 150 and 400 patients, and six enrolled more than 400 patients. All trials were considered to be at high risk of bias, with none of the trials adequately blinding patients and assessors to treatment.

There was no statistically significant increase in risk of infection (10 trials; 3,588 patients), risk of death from infection (three trials; 1,161 patients), and febrile neutropenia (two trials; 702 patients) with rituximab (negligible heterogeneity for all comparisons).

There were statistically significant increases in the risk of leucopenia (RR 1.24, 95% CI 1.12 to 1.37; eight trials; 2,287 patients), granulocytopenia (RR 1.07, 95% CI 1.02 to 1.12; eight trials; 2,381 patients), and overall response (RR 1.12, 95% CI 1.09 to 1.15; 14 trials; 4,703 patients). This suggested that patients receiving rituximab and chemotherapy had a greater risk of toxic effects, but a better treatment outcome, than those receiving chemotherapy alone. Heterogeneity was moderate to strong for these analyses.

Publication bias appeared to be present for some of the analyses, particularly overall response, but subgroup analysis suggested this was strongly linked to type of lymphoma with greater asymmetry in the indolent lymphoma trials. Where heterogeneity was found, it was at least partly explained in subgroup analyses by patient characteristics (stage of treatment or type of lymphoma). Sensitivity analyses found no substantial differences between the overall results and selected trial results.

The addition of rituximab to standard chemotherapy produced better overall response, in patients with CD20-positive malignant lymphomas, without increasing the risk of severe infections.

This review addressed a clear clinical question, with detailed inclusion criteria, and relevant databases were searched. Articles in a range of languages were included; two papers in Chinese were excluded due to the inability to obtain the full publications for translation. Publication and language bias seem unlikely to have influenced the findings of this review. Methods to reduce reviewer error and bias were reported only for study selection, it was unclear if similar steps were taken for data extraction and quality assessment.

The included trials were quality assessed and the details were reported. The authors presented the results of individual analyses in tables, but they did not define the overall response, which makes it difficult to interpret its meaning. The synthesis appears to have been appropriate and thorough; heterogeneity was explored as were potentially important covariates.

Some information on the review processes was not reported and overall response was not defined, but it seems likely that the results of this review are reliable and reflect the balance of the evidence available.

Practice: The authors did not state any implications for practice.

Research: The authors suggested that data on special groups of patients, such as those with HIV or who have hepatitis B virus, were needed. RCTs or large prospective cohort studies were required to evaluate the effect of rituximab on silent and chronic viral infections.

Funded by the Italian Society for Infection Disease and Tropical Medicine (SIMIT).