Types of studies

We considered only randomised controlled trials (RCT) as primary studies in this review and meta‐analysis (O'Connor 2011). We included both full text and abstract publications.

Types of participants

We included studies on male and female patients with previously untreated and histologically‐confirmed diagnosis of early unfavourable or advanced stage Hodgkin lymphoma (HL). We would have accepted trials with protocols for children as well as adults, but we could not identify any randomised trials including children.

If the trial population had been mixed, including participants with haematological malignancies, we would have used data from the HL subgroup. If subgroups had been available (not reported and not provided after personally contacting the author), we would have included only trials with a minimum of 80% of patients with HL, but we could not identify any trial of this type.

We included trials for early unfavourable or advanced stage HL. We considered patients with Ann Arbor stage I/II with risk factors as early unfavourable and patients with Ann Arbor stage III/IV independently from risk factors as advanced stages. We accepted the risk factors used by the respective study group or institution and kept the classification defined and used by the respective study group or institution.

Types of interventions

To assure equal doses of chemotherapy, we considered chemotherapy including at least two cycles of escalated BEACOPP as the intervention condition, and regimens including at least four cycles of ABVD as the control condition, both with or without radiotherapy as first‐line treatment for early unfavourable or advanced stage HL (Hasenclever 2001).

We did not consider chemotherapy with BEACOPP given in the standard dose in this systematic review because the GHSG HD9 trial showed clearly better progression‐free survival (PFS) results with the escalated regimen (Diehl 2004).

If additional radiotherapy was applied, the condition, dosage and field should have been the same in both control and intervention arms.

Types of outcome measures

Electronic searches

For the primary review published in 2011 we searched the following databases in two steps:

First step: MEDLINE (1985 to November 2008, search strategy see Appendix 1), Embase (1985 to November 2008, search strategy see Appendix 2) and Cochrane Central Register of Controlled Trials (CENTRAL) (the Cochrane Library 2008, Issue 3, search strategy see Appendix 3).

Second step: In 2010 we adapted the search strategies and performed another search in MEDLINE (November 2008 to August 2010, search strategy see Appendix 4) and CENTRAL (the Cochrane Library 2010 issue 8, search strategy see Appendix 5). The Embase search was not updated because our account for this database ended in the meantime.

To keep the review up to date we used the adapted search strategies and performed another search in MEDLINE (August 2010 to March 2017, search strategy see Appendix 6) and CENTRAL (the Cochrane Library 2017 issue 03, search strategy see Appendix 7).

Searching other resources

For the primary review, we conducted the search for the years 1990 until August 2010. We searched those years which were included in CENTRAL via CENTRAL. For the updated review we searched the conference proceedings of annual meetings of the following societies for abstracts:        

• American Society of Clinical Oncology (ASCO) (up to the issue of 2016);

• American Society of Hematology (ASH) (up to the issue of 2016);      

• European Hematology Association (EHA) (up to the issue of 2016);       

• International Symposium on Hodgkin Lymphoma (ISHL) (up to the issue of 2016).  

Moreover, we tried to identify trials with the following methods.      

• Handsearching of references

  • Citations from identified trials and relevant review articles.

• We checked references of all identified trials, relevant review articles and current treatment guidelines for further literature.

• Personal contacts

  • We contacted experts in the field in order to locate other trials.

• Databases of ongoing trials:

  • Meta‐register of controlled trials: http://www.controlled‐trials.com/mrct/

• We also searched a number of databases and websites of relevant institutions, agencies, organisations (e.g. study groups), societies and registries for (ongoing) trials (August 2016):

• • Register of controlled trials: www.controlled‐trials.com

  • EU clinical trials register: www.clinicaltrialsregister.eu/ctr‐search/search

  • Clinicaltrials.gov: clinicaltrials.gov

  • www.eortc.be

  • www.ghsg.org

  • www.ctc.usyd.edu.au

  • www.trialscentral.org/index.html

Selection of studies

Two review authors (AW, NS) independently screened titles and abstracts of studies identified from the sources listed above for potentially relevant studies. First, we screened and discarded studies that were clearly ineligible. We assessed selected studies with an eligibility form to determine whether they met the inclusion criteria. We resolved any disagreement between the two review authors by discussion. We would have sought further information from the authors, whose articles contained insufficient data to make a decision about eligibility. The eligibility form contained the following questions.

• Is the study described as a randomised trial for patients?

• Were the participants newly diagnosed with early unfavourable or advanced stage Hogdkin lymphoma?

• Were the participants in the intervention group treated with at least two cycles of escalated BEACOPP?

• Were the participants in the control group treated with at least four cycles of ABVD?

To be eligible, studies had to meet all of the criteria stated above. We also identified duplicate reports.

We obtained all full‐text versions of all eligible studies for quality assessment and data collection. At every stage of searching and screening of the literature, we documented the overall number of studies identified, excluded and included with reasons in a flow diagram (Moher 2009). We resolved disagreements arising at any stage by discussion and consensus. If an agreement had not been reached, we would have asked a third review author (content expert) to give his or her opinion, in accordance with Chapter 7 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011), but this procedure was not necessary.

Data extraction and management

We used both full‐text versions and abstracts (if full‐text was not published) of eligible studies to retrieve the data. In case these did not provide sufficient information, we contacted authors of individual studies for additional details (e.g. HD 2000; GHSG HD9; GHSG HD14). We conducted data extraction according to the guidelines proposed by Cochrane. For this review, two review authors (AW, NS) extracted data independently from the studies using a standardised data extraction form containing the following items.

• General information: first author of the publication, title of study, study ID, author's contact address (if available), source, publication date, country, language.

• Quality and 'Risk of bias' assessment: sequence generation, allocation concealment, blinding (participants, personnel, outcome assessors), incomplete outcome data, selective outcome reporting, other sources of bias.

• Trial details: trial design, objectives of the trial, types of outcomes reported, location, setting, sample size calculation, randomisation, treatment allocation, inclusion/exclusion criteria, reasons for exclusion, comparability of groups, length of follow‐up, stratification, stopping rules described, statistical analysis, results, power calculation, recruitment period, funding.

• Characteristics of participants: age, gender, ethnicity, total number recruited/randomised/analysed (total and for each arm), stage of the disease (Ann Arbor stage, B‐symptoms and risk factors), tumour response (if it was a criteria for randomisation), lost to follow‐up, dropouts by trial arm and reasons.

• Interventions: type of chemotherapy (name, dose and frequency of drugs from the chemotherapy regimen), number of cycles, length of a cycle and route of interventions, compliance to interventions, supportive therapy given, failure to maintain dose intensity, protocol violations, additional radiotherapy (dosage, field).

• Outcomes measured: overall survival (OS), progression‐free survival (PFS) or similar tumour control outcomes, treatment‐related mortality, secondary neoplasms (including overall secondary malignancies and AML and MDS), infertility, adverse events, especially haematological toxicity, cardiopulmonary toxicity, quality of life, response rate, freedom from treatment failure (FFTF), number of patients evaluated for each outcome, length of follow‐up for survival outcomes and definitions of the outcomes.

We extracted trials reported in more than one publication on one form only. We resolved potential disagreements by consensus. If an agreement had not been reached, we would have asked a third review author (content expert) to give his or her opinion, but this procedure was not necessary.

Assessment of risk of bias in included studies

Two review authors (AW, NS) assessed eligible studies obtained in the process of study selection as described above for methodological quality. To assess quality, we used the Cochrane 'Risk of bias' table. To improve transparency of judgements, we included quotations from the full‐text article or from supplemental information from the author in the notes section of the 'Risk of bias' table.

We included the following items in the 'Risk of bias' assessment for randomised trials. For details of bias assessment we referred to Chapter 8 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011a).

There are three categories: yes (low risk of bias), unclear, and no (high risk of bias) in response to the following questions.

• Sequence generation

• Allocation concealment

• Blinding (participants, personnel, outcome assessors)

• Incomplete outcome data

• Selective outcome reporting

• Other potential sources of bias

Measures of treatment effect

For binary variables, we estimated treatment effect measures of individual studies as relative effect measures risk ratio (RR) with corresponding 95% confidence intervals by using the Mantel‐Haenszel method in a fixed‐effect model. For time‐to‐event variables, we estimated treatment effects as hazard ratio (HR) with corresponding 95% confidence intervals by using the generic inverse variance method in a fixed‐effect model. For each trial, we extracted information about time‐to‐event outcome using the methods described by Tierney (Parmar 1998; Tierney 2007).

Dealing with missing data

We dealt with missing data as suggested in Chapter 16 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011b), We contacted the original investigators to request missing data and received additional information from Mournier and colleagues. We performed sensitivity analysis to assess how sensitive results were to reasonable changes in the assumptions that had been made. We addressed the potential impact of missing data on the findings of the review in the discussion.

Assessment of heterogeneity

We assessed heterogeneity of treatment effects between trials by using a Chi² test with a significance level at P < 0.1. We used the I² statistic to quantify possible heterogeneity (30% < I² < 75%: moderate heterogeneity, 75% < I²: considerable heterogeneity) (Deeks 2011).

We explored potential causes of heterogeneity by performing sensitivity and subgroup analysis.

Assessment of reporting biases

In a meta‐analysis with more than 10 trials, we would have explored potential reporting bias by generating a funnel plot and statistically testing this by conducting a linear regression test (Sterne 2011). A P value less than 0.1 would have been considered significant for this test. However, we only included five trials so this test was not done.

Data synthesis

According to the recommendations of Cochrane (Deeks 2011), we performed meta‐analyses by using a fixed‐effect model (i.e. generic inverse variance method for time‐to‐event outcomes and a Mantel‐Haenszel method for dichotomous outcomes).

We performed analyses according to the recommendations of Cochrane using the Review Manager (Review Manager 5.3) and listed the predefined outcomes OS, PFS, the adverse events treatment‐related mortality, secondary malignancies, secondary AML/ MDS, and infertility, and quality of life in a 'Summary of findings' table summary of findings Table for the main comparison on absolute risks in each group with the help of GRADE pro 2014.

Subgroup analysis and investigation of heterogeneity

We planned to take the following parameters into consideration for subgroup analysis of the primary outcomes:

• stage of the disease (early unfavourable versus advanced);

• age (e.g. children versus adults ≤ 50 years versus adults > 50 years, or depending on the median age of the trial);

• treatment regimens (i.e. only ABVD regimen versus chemotherapy including ABVD regimen);

• length of therapy (e.g. four versus eight cycles of chemotherapy);

• number of cycles of escalated BEACOPP (i.e. two versus four versus eight);

• radiotherapy (with or without radiotherapy);

• sequence (e.g. BEACOPP followed by other agents versus other chemotherapeutic agents followed by BEACOPP);

• length of follow‐up (i.e. long follow‐up (more than five years) versus short follow‐up (five years or less);

• same total number of cycles of chemotherapy for the control and intervention groups (yes versus no).

We did not conduct planned subgroup analysis by "age", "length of therapy", "sequence", and "total number of cycles of chemotherapy" because the included studies were similar with regard to these characteristics.

We did a test for subgroup differences and considered a P‐value of less than 0.05 as statistical significant for all the outcomes mentioned in the 'Summary of findings' table.

Sensitivity analysis

We would have assessed robustness of the results of the primary outcomes by analyses of the kind of publication (full‐text publications/abstracts) and the type of results (preliminary results/mature results). However, all included trials have been published as full texts.