Guidelines for the diagnosis and management of primary central nervous system diffuse large B-cell lymphoma.

Abstract

This guideline is aimed at providing healthcare professionals with clear guidance on the diagnosis and management of primary central nervous system lymphoma (PCNSL), defined as diffuse large B-cell lymphoma (DLBCL) solely confined to the central nervous system (CNS): brain, spinal cord, cranial nerves, eyes and meninges. Secondary CNS lymphoma, immunodeficiency-associated lymphoma and rare forms of non-DLBCL CNS lymphoma are outside the scope of this guideline. It is not the intention of this guideline to provide treatment recommendations for all situations and clinicians are advised to take individual patient circumstances into account when making management decisions. Recommendations included a systematic review of published English language literature from publication of previous British Society for Haematology (BSH) PCNSL guidance (1 January 2007) up to 29 May 2017. MEDLINE, EMBASE, Cochrane databases and Web of Science were searched using the preliminary search terms ‘CNS lymphoma’ and ‘intraocular lymphoma’. The Grading of Recommendations Assessment, Development and Evaluation (GRADE) nomenclature was used to assess levels of evidence and assess the strength of recommendations (http://www.gradeworkinggroups.org). Review of the manuscript was performed by the BSH Guidelines Committee Haemato-Oncology Task Force, the BSH Guidelines Committee and the Haemato-Oncology sounding board of BSH. It was posted on the members section of the BSH website for comment. It has also been reviewed by the patient-focused UK charity ‘Lymphoma Action’; although these organisations do not necessarily approve or endorse the contents. Primary central nervous system lymphoma accounts for 1% of all non-Hodgkin lymphomas and 3% of all brain tumours (Swerdlow et al, 2008; Rigau et al, 2011). It is essential to minimise diagnostic delay, and clinical management requires multidisciplinary team (MDT) support. This includes haemato-oncology but, unique to haematological cancers, also requires input from specialist neurological services. Methotrexate-based protocols should only be delivered at centres experienced in intensive chemotherapy. Primary central nervous system lymphoma presents with a range of symptoms and signs including behavioural change, memory and language impairment, focal motor deficits, seizures, raised intracranial pressure, uveitis and neuropsychiatric symptoms (Zhang et al, 2010; Aki et al, 2013). The diagnosis should, in all cases, be confirmed by specialist haematopathology review of sampled tumour tissue or fluid according to the current World Health Organization classification (Swerdlow et al, 2008, 2016). Stereotactic biopsy is recommended for intracerebral lesions with intraoperative rapid cytology and review of frozen sections to avoid unnecessary extensive surgery (Abrey et al, 2005). Cerebrospinal fluid (CSF) cytology and flow cytometry may be used in cases where a biopsy is not possible or to investigate for leptomeningeal involvement, although sensitivity is low (Schroers et al, 2010). At least 3–10 ml of CSF should be taken (Ferreri et al, 2004) and rapidly analysed (Patrick & Mohile, 2015). The identification of clonal IGH rearrangements by polymerase chain reaction may improve diagnostic yield (Ekstein et al, 2006; Langerak et al, 2012). Corticosteroids should, wherever possible, be avoided prior to biopsy as they have a substantial negative impact on diagnosis; non-diagnostic rates range from 33% after a short course (<1 week) to 57% after a longer course of steroids (Manoj et al, 2014). If corticosteroids have been administered, but an enhancing lesion is still present, they should be discontinued prior to urgent biopsy (Patrick & Mohile, 2015). If a suspected PCNSL lesion resolves following steroid administration, re-imaging by magnetic resonance imaging (MRI) should be performed after a short interval (e.g. 2–4 weeks) following steroid cessation. Close clinical follow-up and serial imaging thereafter is recommended, at a frequency guided by MDT advice, with urgent biopsy at lesion regrowth. Up to 20% of PCNSL patients have intraocular involvement (Coupland et al, 2004), which may resemble chronic uveitis (Coupland et al, 2004; Chan & Sen, 2013). Diagnosis can be difficult and imaging alone cannot reliably establish ocular involvement (Haldorsen et al, 2011). It is recommended that slit lamp examination and ophthalmoscopy are followed, if necessary, by vitreous biopsy. Vitreous biopsy should be combined with a sub-retinal aspirate or chorioretinal biopsy, particularly for those with visible sub-retinal deposits, as vitrectomy specimens have diagnostic failure rates up to 30% (Coupland et al, 2004). Contrast-enhanced MRI of the brain is the neuroimaging modality of choice for both diagnosis and response assessment (Coulon et al, 2002; Abrey et al, 2005; Ferreri, 2011). A diagnosis of PCNSL cannot be presumed on radiological appearance and/or clinical features (such as responsiveness to corticosteroids) alone, as neurosarcoidosis, multiple sclerosis, glioblastoma and vasculitis can mimic disease features (Zaki et al, 2004; Abrey et al, 2005). PCNSL can also manifest with atypical features, including variable contrast enhancement or diffusion characteristics, absence of focal masses or presence of necrosis (Tang et al, 2011). Use of additional imaging studies, such as 18F deoxyglucose positron emission tomography (18FDG-PET), or advanced MRI parameters, such as perfusion metrics, diffusion coefficients, MR spectroscopy and novel contrasts, may have the potential to improve the distinction between PCNSL and glioblastoma but none has shown sufficient evidence of specificity to allow use in routine practice (Barajas et al, 2009; Wieduwilt et al, 2012; Kawai et al, 2013; Valles et al, 2013). Leptomeningeal dissemination occurs in around 16% of cases as judged by cytology (Korfel et al, 2012); it can be difficult to detect on MRI (Tang et al, 2011). All patients should undergo cross-sectional imaging to exclude systemic disease. 18FDG-PET combined with computed tomography (CT) imaging (PET-CT) is a sensitive screening tool for systemic involvement at the time of diagnosis (Mohile et al, 2008). A bone marrow biopsy is not essential in the context of typical histology if PET-CT has excluded systemic disease, full blood count parameters are normal and there is no detectable serum monoclonal protein to suggest a concurrent low-grade lymphoma. There is insufficient evidence that PET-CT has sufficient sensitivity to exclude testicular involvement therefore testicular ultrasound should be performed in men. Serum lactate dehydrogenase, performance status and CSF protein levels form part of established prognostic scores (see Appendix S1) and should be measured, where feasible (Ferreri et al, 2003). The International PCNSL Collaborative Group (IPCG) guidelines on standardised evaluation of patients with newly diagnosed or suspected PCNSL (Abrey et al, 2005) are summarised in Table 1. Optimal therapy of PCNSL incorporates two phases of treatment: remission induction followed by consolidation. This concept largely relates to challenges in delivering sufficient dose intensity across the blood–brain barrier and the high rates of PCNSL relapse when consolidation therapy is not delivered (Ferreri, 2011). Fitness for chemotherapy should be determined by physiological fitness (organ function and comorbidities) rather than chronological age (Zhu et al, 2009; Roth et al, 2012; Welch et al, 2012). Performance status is frequently impaired at diagnosis of PCNSL and should not preclude intensive induction therapy. Performance status often improves following initial therapy which may allow subsequent intensification of therapy. Moreover, both intensive multi-agent induction and high dose therapy with autologous stem cell transplant (HDT-ASCT) approaches are feasible in selected older patients, including some aged >70 years (Schorb et al, 2017a). No validated comorbidity assessment exists to guide treatment choices. High-dose methotrexate (HD-MTX)-based protocols require adequate renal, hepatic and cardiac function and may require dose adjustment in line with institutional and manufacturer's guidance if impaired (e.g., left ventricular ejection fraction >45% and creatinine clearance <50 ml/min). Outcomes for PCNSL patients have improved significantly over the past decade, largely as a result of alterations in management based on evidence emerging from prospective clinical trials. Patients should be offered entry into a clinical trial wherever available. Combination chemotherapy regimens incorporating HD-MTX are considered the standard of care for newly diagnosed PCNSL, resulting in high rates of initial response when combined with other agents (Hoang-Xuan et al, 2015; Ferreri et al, 2016). Penetration of MTX into the CNS is influenced by the total dose and rate of infusion. Most studies employ doses of between 3 and 8 g/m2 (Rubenstein et al, 2013; Ferreri et al, 2016; Glass et al, 2016) although the optimal dose has not been established. It is crucial that MTX is administered as a rapid infusion (2–4 h) at a dose of at least 3 g/m2 to maximise therapeutic CSF concentrations, at an interval of 10–21 days, as part of an established protocol. Modern protocols typically employ four to eight cycles of HD-MTX-based therapy. The optimal number of treatment cycles is likely to be influenced by partner chemotherapy agents, dose intensity of the regimen and intended consolidation approach, but comparative data are not available. The International Extranodal Lymphoma Study Group (IELSG) study IELSG20 demonstrated the value of combining HD-MTX with partner cytotoxic agents: the addition of cytarabine (4 doses of 2 g/m2) significantly improved rates of complete response (CR) and progression-free survival (PFS) as compared to HD-MTX alone (Ferreri et al, 2009). The randomised phase 2 IELSG32 trial subsequently demonstrated that the addition of eight doses of rituximab (375 mg/m2) to HD-MTX/cytarabine improved response rates. Importantly, the addition of both thiotepa and rituximab to HD-MTX/cytarabine as a 4-drug regimen (MATRix; see Appendix S2) resulted in a clear improvement in overall survival (OS) over HD-MTX/cytarabine alone, with a 2-year OS rate of 69% (95% confidence interval [CI] 64–74) vs. 42% (95% CI 36–48), respectively. Patients with stable disease or better after four cycles of MATRix were subsequently randomised to receive consolidation with whole brain radiotherapy (WBRT) or HDT-ASCT. Peripheral blood stem cells (PBSC) were successfully collected in 96% patients after two cycles of MATRix (Ferreri et al, 2016). PBSC collection after MATRix should be attempted but may be less successful if deferred beyond cycle 2 (C.P. Fox and K. Cwynarski, unpublished data). Treatment-related mortality is around 4–7% with MATRix, with most treatment-related deaths occurring during the first treatment cycle (Ferreri et al, 2016; Schorb et al, 2017b). Within the IELSG32 study (median age 57 years), the relative dose intensity of cytarabine and thiotepa during remission induction was 78% and 76% respectively, with protocol-defined reductions predominantly for haematological toxicity. In a recent real world study of outcomes with MATRix chemotherapy (median age 61 years), 23 of 88 included patients would not have met IELSG32 eligibility criteria due to age, performance status or co-morbidities (Schorb et al, 2017b). Consequently, chemotherapy modifications were more frequent (40–50%) but survival rates were similar to those in IELSG32, with 2-year OS of 64%. Severe infectious complications and intensive care support were more common during cycle 1 (16%) than cycle 4 (5%). Thus, for patients considered at risk of increased toxicity (any of: Eastern Cooperative Oncology Group performance status ≥2, co-morbid conditions or age >65 years) we recommend dose reducing the myelotoxic agents in MATRix (cytarabine and thiotepa) for the initial cycle, with dynamic review cycle by cycle. In practice, a 25% reduction of cytarabine (achieved by omitting the 4th dose in the cycle) with or without a 25% dose reduction of thiotepa is a reasonable strategy. Granulocyte colony-stimulating factor (G-CSF) and anti-infection prophylaxis (against herpes simplex and Pneumocytis jirovecii as a minimum) are recommended. A number of alternative remission-induction regimens have been assessed in non-randomised trials, limiting reliable comparisons. Promising results have been reported with HD-MTX, temozolomide and rituximab (MT-R), with a variety of consolidation strategies (2-year PFS of 57–64%) (Rubenstein et al, 2013; Glass et al, 2016). The R-MPV regimen (HD-MTX, procarbazine, vincristine and rituximab) has also been associated with encouraging results in a number of small, non-comparative phase 2 studies (Morris et al, 2013; Omuro et al, 2015a,b). R-MBVP induction chemotherapy (rituximab, HD-MTX, BCNU [carmustine], prednisolone, etoposide) followed by either WBRT or HDT-ASCT resulted in a 2-year OS of 86% in the PRECIS trial (Houillier et al, 2016). A recent randomised study failed to demonstrate that the addition of rituximab to the same MBVP regimen improves outcomes (hazard ratio for PFS: 0·77, 95% CI 0·52–1·13, P = 0·18) (Bromberg et al, 2017). However, a post hoc analysis showed evidence of a PFS benefit in patients aged <60 years, and those aged ≥60 years did not receive consolidation therapy; the full study publication is awaited. Whilst rituximab and HD-MTX are internationally accepted as standard of care (Hoang-Xuan et al, 2015), the optimal induction regimen remains uncertain. Although a number of published protocols are associated with promising outcomes, MATRix is recommended given the higher level of randomised evidence demonstrating clear survival advantages over comparator arms. The additional value of concurrent intrathecal chemotherapy is unproven, with conflicting data from published series (Ferreri et al, 2002a; Khan et al, 2002; Pels et al, 2009; Sierra Del Rio et al, 2012). In view of the risks and potential morbidity associated with repeated lumbar punctures, low level of available evidence and improved efficacy of systemic therapy, we do not advocate concurrent intrathecal chemotherapy. Surgical intervention has a very limited role in PCNSL therapy. Although a retrospective analysis of the German PCNSL Study Group (G-PCNSL-SG)-1 trial suggested that patients with subtotal or total resections had improved outcomes (Weller et al, 2012), this finding is likely to be influenced by patient selection bias; resection was more commonly performed for superficial lesions associated with a better prognosis (Ferreri et al, 2003). These findings remain controversial and have not been adopted by consensus (Hoang-Xuan et al, 2015). Given that PCNSL is considered a whole brain disease (Lai et al, 2002), therapeutic resection of PCNSL lesions should be restricted to critical circumstances where urgent surgical reduction of intracranial pressure is essential. Historically, clinical studies for PCNSL have defined ‘elderly’ patients as >60 years of age, although this cut-off has been empirically adopted without a modern evidence-base. It is clear that chronological age is not a barrier to safe delivery of HD-MTX (>3 g/m2) if physiological fitness (particularly cardiac and renal function) is deemed adequate (e.g. left ventricular ejection fraction >45% and creatinine clearance >50 ml/min) (Zhu et al, 2009; Roth et al, 2012; Welch et al, 2012). A number of phase 2 studies have focused on less myelotoxic induction regimens in patients aged ≥60 years, typically combining HD-MTX with orally administered alkylating agents, without WBRT consolidation (Hoang-Xuan et al, 2003; Illerhaus et al, 2009; Fritsch et al, 2011). Recently, the PRIMAIN study assessed HD-MTX with rituximab, procarbazine and lomustine, followed by procarbazine maintenance in patients aged ≥65 years (median age 73 years). The study was amended to omit lomustine due to excessive toxicity, without any clear loss of efficacy. The 2-year OS with and without lomustine was 47·9% (95% CI 30·4–65·3) and 46% (95% CI 34·1–57·8), respectively (Fritsch et al, 2017). A Nordic study used a modified Bonn protocol (Pels et al, 2003) in patients aged 65–75 years (median age 70 years), incorporating reduced dose HD-MTX-based polychemotherapy with temozolomide, followed by maintenance temozolomide for 1 year. Estimated 2-year OS was 55·6% (95% CI 35·3–71·8), which was similar to a younger cohort receiving a more intensive version of the same regimen (2-year OS 60·7%, 95% CI 43·3–74·2) (Pulczynski et al, 2015). The only randomised phase 2 study in patients aged >60 years (median age 72 years) compared HD-MTX, procarbazine, vincristine and cytarabine (MPV-A) to HD-MTX and temozolomide (MT), without rituximab. No significant difference between the arms could be demonstrated with a 1-year PFS of 36% in both arms (Omuro et al, 2015a). A meta-analysis evaluated individual patient data from 20 prospective and retrospective studies in PCNSL patients aged >60 years. HD-MTX-based therapy was associated with improved OS, although there was no discernible survival benefit to using intensive intravenous treatment protocols over HD-MTX combined with oral alkylating agents (Kasenda et al, 2015). For patients considered unfit for MTX-based therapy there remains a paucity of good quality data to inform treatment decisions. WBRT, corticosteroids and oral chemotherapy are common approaches. A small retrospective study of single agent temozolomide in elderly patients with co-morbidities (n = 19) showed a CR rate of 47%, with prolonged responses (>12 months) in 29·4%, a median PFS of 5 months and median OS of 21 months (Kurzwelly et al, 2010). A number of novel agents have shown promise as single agent therapy in relapsed/refractory PCNSL (see below) but there are currently insufficient data to recommend the use of these, as yet unlicensed, agents for those unfit for MTX-based therapy. In patients aged ≥60 years, WBRT alone (40 Gy + 20 Gy boost) gave a median survival of only 7·6 months (Nelson et al, 1992). Lower doses and shorter treatment durations (20–30 Gy in 1·8–4 Gy fractions) may be a more pragmatic approach as long term neurotoxicity is not the primary clinical concern. Fatigue is the most common side effect of palliative WBRT and may take several months to improve. Performance status, life expectancy and quality of life are important factors to take into consideration when discussing treatment options, including best supportive care. Key in this context are good communication and input from palliative care specialists if required; expectations of patients and their families should be carefully managed. Response should be assessed by contrast-enhanced MRI with neuroradiology review according to international guidelines (Abrey et al, 2005), as summarised in Table 2. The role of interim MRI has not been clearly defined, although early achievement of CR after two cycles of chemotherapy (of 6) appears to be associated with improved OS (Pels et al, 2010). Early response assessment after one to two cycles may also facilitate decisions regarding PBSC harvest and allow detection of early disease progression in the 8–29% who progress during first-line chemotherapy (Ferreri et al, 2016; Langner-Lemercier et al, 2016). Patients with non-progressive disease (stable disease or better) following induction therapy should be considered for consolidation therapy (Soussain et al, 2008; Illerhaus et al, 2016). Given the inherent difficulty in delivering optimal dose intensity across the blood–brain barrier, the risk of early relapse, and the important role of consolidation in the management of PCNSL, we recommend that clinicians plan to commence consolidation therapy within 6–8 weeks of the first day of the final induction chemotherapy cycle. This is a judgement for individual clinicians and assumes that the patient's performance status is reasonable and any significant toxicities have sufficiently resolved. The diffuse multifocal nature of PCNSL necessitates WBRT rather than targeted radiotherapy. Most protocols empirically use a WBRT dose of 30–45 Gy although it should be emphasised that the optimal total dose and role of a ‘boost’ remains uncertain, particularly in the era of more intensive remission induction protocols. WBRT has traditionally been used as consolidation following response to first-line chemotherapy (DeAngelis et al, 2002), but its role in the context of modern immuno-chemotherapy is less clear. A randomised phase 3 trial compared WBRT (45 Gy) with no further treatment in patients achieving CR following MTX-based chemotherapy (Thiel et al, 2010; Korfel et al, 2015). The design and conduct of this study have been criticised; only 58% patients received protocol treatment and the trial failed to meet its non-inferiority end-point. Therefore, it remains unclear whether WBRT consolidation can be safely omitted for patients in CR (Thiel et al, 2010). Irreversible, and sometimes disabling, neurocognitive dysfunction is a well-recognised consequence of WBRT, particularly in those aged >60 years. Indeed, a systematic meta-analysis found no clear overall benefit, in terms of quality-adjusted life years, for WBRT in first remission in patients >60 years of age (Prica et al, 2012). In an attempt to mitigate this risk, trial strategies have investigated hyper-fractionated or lower radiation doses (Correa et al, 2009; Doolittle et al, 2013; Ferreri et al, 2016; Glass et al, 2016). Use of reduced dose radiotherapy (23·4 Gy) as consolidation for patients in CR has been explored by Morris et al (2013) in a non-randomised phase 2 study. In this subset of patients (n = 12), survival and neurocognitive outcomes were encouraging, but the number of evaluable patients was small and definitive conclusions cannot be drawn (Morris et al, 2013). Randomised studies are needed to validate this approach. Any decision to employ WBRT as consolidation should be based on high quality MRI imaging reviewed by a neuroradiologist or, ideally, following neuroradiology consensus opinion through MDT structures. Encouraging results from early studies with HDT-ASCT consolidation in PCNSL have challenged the role of WBRT as the favoured first-line consolidation strategy (Soussain et al, 2001; Illerhaus et al, 2006). On an intention-to-treat basis, prospective trials of thiotepa/carmustine-conditioned ASCT after intensive induction chemotherapy have reported 3- to 5-year OS rates of 70–81%, where 79–92% received planned HDT-ASCT (Kasenda et al, 2012; Illerhaus et al, 2016). In the study by Illerhaus et al (2016) only a minority (14%) received additional WBRT. Broadly similar results have also been achieved with thiotepa, busulfan and cyclophosphamide conditioning (Soussain et al, 2012; Omuro et al, 2015b; Houillier et al, 2016), although no formal comparison of conditioning regimens has been conducted. Historical results with the BEAM regimen (carmustine, etoposide, cytarabine, melphalan) were disappointing (Abrey et al, 2003). In the IELSG32 study, 24 of 28 patients with partial response or stable disease after induction therapy achieved CR following HDT-ASCT (Ferreri et al, 2017). Long-term survival rates, however, are lower in chemorefractory patients who fail to achieve partial remission to induction therapy (Soussain et al, 2012; Schorb et al, 2013). Survival outcomes from the prospective studies have been mirrored by a recently published ‘real world’ UK retrospective study (Kassam et al, 2017). Importantly, results of the first randomised trials comparing WBRT consolidation with HDT-ASCT have recently been reported. Interim analysis of the PRECIS trial reported a 2-year PFS of 63·2% (95% CI 49·5–80·5) for WBRT consolidation versus 86·8% (95% CI 76·6–98·3) after HDT-ASCT, with identical 2-year OS between the two arms (Houillier et al, 2016). The IELSG32 trial reported no significant difference in PFS or OS between WBRT and HDT-ASCT consolidation on intention-to-treat analysis (2-year OS 80% [95% CI 70–90] vs. 69%, respectively [95% CI 59–79]) (Ferreri et al, 2017). Although longer follow-up and neurocognitive results are awaited, a reduced incidence of delayed neurotoxicity is likely to favour HDT-ASCT as consolidation therapy. Notably, feasibility of consolidation with HDT-ASCT has also been demonstrated in selected older patients (>65 years old) by a recent European collaborative study (Schorb et al, 2017a). A non-myeloablative chemotherapy consolidation approach has been investigated as an alternative to HDT-ASCT. A single arm trial of intensive EA consolidation (etoposide/cytarabine) after MT-R induction (HD-MTX/temozolomide/rituximab) reported a 4-year OS of 65%. Notwithstanding a relatively high rate of early disease progression, long-term disease control using this strategy appears broadly comparable to chemo-radiation protocols (Rubenstein et al, 2013). Randomised trials are ongoing to ascertain whether non-myeloablative chemotherapy approaches, such as EA or DeVIC (dexamethasone, etoposide, ifosfamide and carboplatin) (Motomura et al, 2011) are comparable to HDT-ASCT consolidation (NCT01511562). Around 6–25% of all relapses are asymptomatic and detected on follow-up imaging (Langner-Lemercier et al, 2016; Fossard et al, 2017; Mylam et al, 2017). In a large population-based retrospective study, asymptomatic patients had better performance status, which may have facilitated delivery of intensive salvage therapy, and was associated with improved outcomes. On multivariable analysis, performance status at relapse was one of the factors most strongly associated with OS (Langner-Lemercier et al, 2016). Smaller studies have not shown a clear benefit for surveillance imaging, but this may relate to the frequency and timing of surveillance (Fossard et al, 2017; Mylam et al, 2017). Prospective data are lacking. A recommended schedule for disease monitoring has been outlined in international consensus guidelines (Abrey et al, 2005). Primary intraocular lymphoma is classified as a variant of PCNSL with its own therapeutic considerations. The optimal treatment for intraocular disease remains controversial as evidence is largely limited to retrospective and small prospective case series. Given that CNS lymphoma is the principal cause of death in PIOL (Grimm et al, 2007), the potential coexistence of CNS disease should be proactively addressed within PIOL treatment algorithms. Chemotherapy agents that penetrate the blood-brain barrier, particularly HD-MTX and cytarabine, also cross the blood-ocular barrier (Baumann et al, 1986; Siegel et al, 1989; Batchelor et al, 2003), resulting in ocular responses and persistent remissions (Grimm et al, 2007; Riemens et al, 2015). Single agent ifosfamide and trofosfamide have also been shown to have some efficacy in a small prospective study (n = 10) (Jahnke et al, 2009). The optimal combination of these drugs is yet to be determined. Intravitreal chemotherapy alone, usually MTX +/− rituximab (Frenkel et al, 2008; Larkin et al, 2014) can achieve remission in a proportion of patients. However, CNS relapse occurs in 33–58% patients, and ocular toxicity is reported in 26–36% (Grimm et al, 2007; Riemens et al, 2015), hence, systemic PCNSL regimens should be employed. A limited number of small retrospective and single arm studies have reported promising outcomes with a combined intravitreal and systemic approach (Ma et al, 2016). However, randomised data are absent and, given the potential for additional ocular toxicity, this approach cannot be routinely recommended. Irrespective of treatment modality used, relapse rates remain high (Grimm et al, 2007; Riemens et al, 2015; Nguyen et al, 2016) therefore consolidation therapy is recommended. Both ocular radiotherapy (usually incorporating both globes +/− WBRT) (Ferreri et al, 2002b) and HDT-ASCT (Soussain et al, 2001) can induce durable remissions in PIOL. At present there are insufficient data to recommend one consolidation strategy over another. Treatment of relapsed and refractory (R/R) PCNSL remains a major area of unmet clinical need. The prognosis of R/R PCNSL is very poor with a median OS of 3·5 months (Langner-Lemercier et al, 2016). Published studies of salvage therapy for R/R PCNSL comprise small, non-randomised and often retrospective analyses, with low quality evidence and thus no established standard of care. Studies of non-myeloablative chemotherapy, either single-agent or combination approaches, typically report dismal outcomes with response rates of 30–55% and median PFS of 2–11 months (Grommes & DeAngelis, 2017). Consequently, patients should be offered clinical trial entry wherever available. Ifosfamide-based salvage regimens, usually in combination with etoposide +/− carboplatin and rituximab (R-IE and R-ICE), have resulted in overall response rates of 41–95% in predominantly chemorefractory or heavily treated patient cohorts (Motomura et al, 2011; Choi et al, 2013; Mappa et al, 2013; Choquet et al, 2015; Langner-Lemercier et al, 2016). For patients who experience durable first remissions following HD-MTX-based protocols, re-treatment with HD-MTX-based regimens may be effective, although the effectiveness of this approach is much more uncertain in the modern-era of intensified MTX-containing regimens. Two retrospective studies described a median PFS of 16 and 25·8 months after re-treatment with HD-MTX at relapse, usually as part of multi-agent salvage regimens, in patients who had experienced as median duration of response to first-line HD-MTX of 24·4 and 26 months, respectively (Plotkin et al, 2004; Pentsova et al, 2014). This may be particularly relevant for MTX-experienced, but rituximab- and thiotepa-naïve, patients for whom the more intensive MATRix protocol may be an option, although it should be noted that currently no published data exist for MATRix in R/R PCNSL. Encouraging results have been achieved with a number of novel agents, with response rates >50% observed with single agent ibrutinib (Chamoun et al, 2017; Grommes et al, 2017; Lionakis et al, 2017) and nivolumab (Nayak et al, 2017). Lenalidomide also crosses the blood–brain barrier and appears to have activity in PCNSL when combined with rituximab (Ghesquieres et al, 2016; Rubenstein et al, 2016). However, these agents are not yet licensed for this indication and survival data are immature. For eligible, chemosensitive patients, available data support the use of thiotepa-based HDT-ASCT in R/R PCNSL, with a median PFS of 24–41 months reported (Soussain et al, 2008, 2012; Kasenda et al, 2017). Either HDT-ASCT or WBRT should be considered in second remission if not undertaken as first-line consolidation therapy. WBRT alone (median dose 36–40 Gy) can result in CR rates of 37–58% and a median survival of 10–16 months in radiotherapy-naïve patients who are refractory to, or relapse after, HD-MTX therapy (Nguyen et al, 2005; Hottinger et al, 2007; Khimani et al, 2011). Cognitive impairment is a feature of PCNSL, with inevitable heterogeneity in neuropsychological deficits being associated with different tumour locations. In some patients there may also be unilateral weakness affecting motor function and/or behavioural syndromes, such as disinhibition or apathy, which may influence test selection or approach during assessments. Elucidating the cognitive effects of disease versus treatment is difficult due to methodological limitations of published studies, including lack of baseline cognitive assessments and control groups, and variable disease status. The type of treatment and age are important considerations. In a review of 17 studies, cognitive impairment was found in most PCNSL patients treated with WBRT plus chemotherapy whereas patients treated with chemotherapy alone had either stable or improved cognitive performance (Correa et al, 2007). In the longest observational study of 80 PCNSL survivors in CR, patients receiving WBRT had lower mean scores in attention, executive function, motor skills and overall neuropsychological composite score compared with those treated without WBRT (Doolittle et al, 2013). Treatment-related cognitive morbidity is also associated with a lower quality of life and poor prognosis (Prica et al, 2012; Doolittle et al, 2013). There is no consensus on the optimal neurocognitive test battery for PCNSL patients, or on the interval at which follow-up neurocognitive assessments should be performed. Baseline and serial mini mental state examination have been suggested as a minimum requirement (Abrey et al, 2005). Table 3 shows the proposed minimum core battery for the assessment of neuropsychological functions and quality of life in PCNSL patients recommended by Correa et al (2007) and proposed alternative measures for UK patients. The battery can be completed within 30–40 min and has been incorporated into prospective trials (Correa et al, 2007). EORTC-QLQ-C30 (Aaronson et al, 1993) BCM 20 (Osoba et al, 1996) EORTC-QLQ-C30 BCM 20 Given the cognitive, psychological and physical effects of both CNS lymphoma and its treatment, it is important that holistic needs are addressed during treatment and throughout the recovery period. Early referral to support services and specialist therapies should be considered according to individual patient need. While the advice and information in this guidance is believed to be true and accurate at the time of going to press, neither the authors, the BSH nor the publishers accept any legal responsibility for the content of this guidance. The BSH Haemato-oncology task force members at the time of writing this guidance were Dr Gail Jones (Chair), Dr Guy Pratt (Secretary), Dr Simon Stern, Dr Jonathan Lambert, Dr Nilima Parry-Jones, Dr Pam McKay and Dr Alastair Whiteway. The authors would like to thank them, the BSH sounding board and the BSH guidelines committee for their support in preparing this guideline. The authors are also grateful to Lymphoma Action for their review and input. The BSH paid the expenses incurred during the writing of this guidance. All authors have made a declaration of interests to the BSH and Task Force Chairs which may be viewed on request. CPF, KL and KC have received advisory board and/or speakers’ fees from Adienne. CPF, EHP, EGE, CH, PM, AD and KC have received financial reimbursement, advisory board fees and/or research funding from F. Hoffman-La Roche. The following members of the writing have no conflicts of interest to declare: JS, DPA and CFu. Members of the writing group will inform the writing group Chair if any new pertinent evidence becomes available that would alter the strength of the recommendations made in this document or render it obsolete. The document will be archived and removed from the BSH current guidelines website if it becomes obsolete. If new recommendations are made an addendum will be published on the BSH guidelines website (https://b-s-h.org.uk/guidelines/). All authors reviewed the literature and contributed to the drafting and editing of this manuscript. Appendix S1. Prognostic scores in primary CNS lymphoma. Appendix S2. MATRix regimen-methotrexate, cytarabine, thiotepa and rituximab. Appendix S3. R-MP regimen-rituximab, methotrexate and procarbazine. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.