Version July 2025
INTRODUCTION —
Diffuse large B cell lymphoma (DLBCL) is the most common histologic subtype of non-Hodgkin lymphoma (NHL), accounting for approximately 25 percent of NHL cases [1,2]. (See "Classification of hematopoietic neoplasms".)
DLBCL describes a morphologically, genetically, and biologically heterogeneous group of NHLs. Outcomes are associated with pathologic features and clinical factors.
This topic discusses prognostic features and predictive models for DLBCL.
Details of the classification systems and the numerous categories of B cell NHL are presented separately. (See "Classification of hematopoietic neoplasms".)
The prognosis of DLBCL that arises from transformation of a previously undiagnosed indolent lymphoid neoplasm is discussed separately. (See "Histologic transformation of follicular lymphoma", section on 'Prognosis and prognostic factors' and "Richter transformation in chronic lymphocytic leukemia/small lymphocytic lymphoma".)
OVERVIEW —
DLBCL is cured in two-thirds of cases with current therapy, particularly in patients who achieve a complete response with first-line treatment.
In addition to pathologic features and disease stage, other factors that contribute to treatment outcomes include age, socioeconomic conditions, comorbid conditions, and performance status [3,4].
Prognostic models and the pathologic features that are associated with outcomes in DLBCL are discussed in the following sections.
INTERNATIONAL PROGNOSTIC INDEX —
The International Prognostic Index (IPI) and its variants are the main prognostic tools used in patients with DLBCL (table 1) and (figure 1).
Original IPI — The IPI is our preferred instrument for assessing prognosis in patients with DLBCL.
The IPI was developed to evaluate pretreatment features that predict outcomes in patients with aggressive non-Hodgkin lymphoma, including DLBCL. Overall survival (OS) and relapse-free survival (RFS) in patients receiving doxorubicin-containing chemotherapy are associated with age, serum lactate dehydrogenase (LDH), performance status, clinical stage, and extranodal disease [5].
Following the introduction of rituximab, the IPI model was validated in patients treated with R-CHOP (rituximab, cyclophosphamide, doxorubicin, vincristine, prednisone) and R-CHOP-like regimens [6,7].
The following factors correlated with inferior OS and progression-free survival (PFS):
●Age >60 years
●Serum LDH above the upper limit of normal (ULN)
●Eastern Cooperative Oncology Group (ECOG) performance status ≥2 (table 2)
●Clinical stage III or IV (table 3)
●>1 extranodal disease site
One point is given for each of the above characteristics, for a total score ranging from zero to five (table 1 and figure 1).
The prognostic categories and survival outcomes for 1063 patients with aggressive B cell lymphomas who received R-CHOP or CHOP-like chemotherapy in three prospective studies reported the following three-year estimates for OS, PFS, and event-free survival (EFS) [6]:
●Low risk (IPI score zero to one) – 91, 87, 81 percent
●Low-intermediate risk (IPI score two) – 81, 75, 69 percent
●High-intermediate risk (IPI score three) – 65, 59, 53 percent
●High risk (IPI score four to five) – 59, 56, 50 percent
NCCN-IPI — The National Comprehensive Cancer Network (NCCN)-IPI incorporates more detailed information about the clinical variables used in the original IPI.
The NCCN-IPI differs from the original IPI by considering age and LDH as continuous variables, and by replacing the number of extranodal sites with the location of extranodal disease. When applied to the same population of patients, the NCCN-IPI provides greater discrimination between risk groups than the original IPI [8] but is more cumbersome to use at the bedside.
Clinical variables and outcomes of 1650 patients with de novo DLBCL treated at seven NCCN cancer centers from 2000 to 2010 were used to create the NCCN-IPI, which awards points for each of the following variables [8]. Note that the LDH variable in the NCCN-IPI is the ratio of the patient's LDH level to the institution's ULN.
●Age 41 to 60 years – one point
●Age >60 to 75 years – two points
●Age >75 years – three points
●LDH ratio >1 to 3 – one point
●LDH ratio >3 – two points
●ECOG performance status ≥2 – one point
●Ann Arbor stage III to IV – one point
●Extranodal disease involving the bone marrow, central nervous system, liver/gastrointestinal tract, or lung – one point
The NCCN-IPI score is the sum of points from the above variables.
When it was applied to a separate cohort of 1138 patients in the British Columbia Cancer Agency registry treated with R-CHOP, the NCCN-IPI stratified patients into four risk groups with significantly different OS and PFS at five years:
●Low risk (zero to one point; 12 percent of patients); 96 percent OS; 94 percent PFS
●Low-intermediate risk (two to three points; 37 percent of patients); 77 percent OS; 72 percent PFS
●High-intermediate risk (four to five points; 37 percent of patients); 56 percent OS; 54 percent PFS
●High risk (≥6 points; 14 percent of patients); 38 percent OS; 35 percent PFS
Other IPI variants
Age-adjusted IPI — An age-adjusted IPI was evaluated for the 1274 patients ≤60 years of age in the original study group [5]. To generate this score, all the prognostic factors listed above, with the exception of age and number of extranodal sites, were given one point, for a score ranging from zero to three (table 1):
●Low risk – Age-adjusted IPI score of zero
●Low-intermediate risk – Age-adjusted IPI score of one
●High-intermediate risk – Age-adjusted IPI score of two
●High risk – Age-adjusted IPI score of three
Five-year OS rates for patients ≤60 with age-adjusted scores of zero, one, two, and three were 83, 69, 46, and 32 percent, respectively. Five-year OS rates for those >60 with the same scores were 56, 44, 37, and 21 percent, respectively.
Stage-adjusted IPI — Modifications of the IPI have been made for patients with stage I or II aggressive non-Hodgkin lymphoma ("stage-adjusted" or "stage-modified" IPI) [9,10] since there are marked differences in prognosis among such patients. In one proposed scoring system, one point was given for each of the following pretreatment variables:
●Age >60
●Increased serum LDH levels
●Stage II or stage IIE disease (table 3)
●Performance status ≥2 (table 1)
As an example of the variability in prognosis among patients with stage I or II aggressive non-Hodgkin lymphoma undergoing similar treatment (ie, three courses of doxorubicin-containing chemotherapy plus involved region radiation therapy), patients with scores on the stage-modified IPI from zero to four had the following 10-year OS rates [11]:
●Score zero – 90 percent
●Score one or two – 56 percent
●Score three or four – 48 percent
The prognostic utility of this stage-modified IPI has also been shown for primary gastric and intestinal DLBCL [12,13]. It has been further noted that patients with bulky stage II aggressive non-Hodgkin lymphoma have an estimated five-year OS of 49 percent, similar to that of patients with stage III or IV disease [14]. (See "Clinical presentation and diagnosis of primary gastrointestinal lymphomas", section on 'Gastric lymphoma' and "Treatment of extranodal marginal zone lymphoma of mucosa-associated lymphoid tissue (MALT lymphoma)".)
PATHOLOGIC FEATURES
Cell of origin status — DLBCL can be classified into two categories according to the cell of origin (COO) status. The COO categories have different molecular abnormalities and prognoses (figure 2).
Gene expression profiling (GEP) is the gold standard for determining the COO status, but methods that correlate with GEP are often applied in clinical practice. We determine the COO status of all cases of DLBCL.
●Germinal center B (GCB) cell type – GCB cases manifest a GEP that resembles a normal GCB cell.
GCB tumors demonstrate t(14;18) translocations in 30 to 40 percent of cases and have better survival outcomes with standard chemotherapy regimens.
●Activated B cell (ABC) type – ABC cases have a GEP that resembles an ABC.
These tumors are likely derived from a post-GCB cell that commenced early stages of plasmacytic differentiation. ABC tumors frequently demonstrate trisomy 3, deletion of CDKN2A, which encodes INK4A/ARF, and constitutive activation of the antiapoptotic nuclear factor kappa B (NF-kB) pathway; they only rarely have t(14;18) translocations. When compared with GCB tumors, ABC tumors are associated with inferior rates of five-year survival following standard R-CHOP (rituximab, cyclophosphamide, doxorubicin, vincristine, prednisone).
In clinical practice, COO status is most often determined by an immunohistochemistry algorithm (eg, Tally and Hans) [15,16] (figure 3) because of the ease of use. The concordance with GEP is approximately 80 percent. These methods are discussed below. (See 'Hans and Tally methods' below.)
A focused gene expression profiling test (Lymph2Cx) can be performed on formalin-fixed paraffin-embedded tissue and is highly concordant (>95 percent) with conventional GEP using ribonucleic acid (RNA) from unfixed cells (eg, Affymetrix gene chip) [17]. (See 'Lymph2Cx platform' below.)
While COO status is useful for a prognosis, it does not currently influence the initial treatment of DLBCL. (See "Initial treatment of advanced-stage diffuse large B cell lymphoma".)
Gene expression profiling — GEP for DLBCL can be performed by complementary deoxyribonucleic acid (DNA) microarrays ("Lymphochip" microarrays).
GEP evaluates the expression of 14 or 27 genes and uses algorithms to produce a linear predictor score. The score has an assigned probability of identifying the case as ABC type. If the probability of being ABC type is >90 percent or <10 percent, cases are labeled ABC or GCB, respectively. Approximately 10 to 15 percent of cases do not meet these criteria and are identified as "unclassifiable."
GEP studies suggest that signatures of the surrounding stromal cells also influence outcomes [18,19], but assessing and quantifying the host immune response in the DLBCL microenvironment is not currently used to determine treatment strategies.
Hans and Tally methods — Panels of immunohistochemical markers can be used to identify COO status [20]. No single immunohistochemical marker can differentiate between GCB and ABC subtypes; therefore, immunohistochemistry-based methods rely on the differential expression of multiple protein markers in the GCB and ABC subtypes.
The two most popular methods are the Hans algorithm and the Tally method (figure 3) [21]:
●Hans algorithm – The Hans classifier uses CD10, BCL6, and MUM1 in a stepwise progression [15]. Cases that express CD10 are classified as GCB. Those that do not express CD10 are evaluated for BCL6 and, if negative, are classified as non-GCB. For those that express BCL6, MUM1 is used as a "tie-breaker." If positive, the cases are classified as non-GCB. If negative, they are classified as GCB. In the initial report, the Hans algorithm correctly assigned 112 of 142 tumors (79 percent concordance). Subsequent reports have estimated the sensitivity (85 to 90 percent), specificity (52 to 82 percent), positive predictive value (55 to 82 percent), and negative predictive value (83 to 90 percent) as compared with GEP [21].
●Tally method – The Tally method uses immunohistochemical markers to calculate a GCB score (CD10 and GCET1) and ABC score (MUM1 and FOXP1). Cases are classified as GCB or ABC if their GCB or ABC scores are greater, respectively. If the scores are equal, LMO2 is used as a "tie-breaker." If LMO2 is <30 percent, the case is classified as ABC. Otherwise, the case is GCB. An initial concordance of 93 percent was reported. Subsequent reports have estimated the sensitivity (80 to 99 percent), specificity (54 to 86 percent), positive predictive value (55 to 87 percent), and negative predictive value (79 to 99 percent) as compared with GEP [21].
Limitations of these methods include reliance on markers that are not simply called positive or negative but require staining thresholds (eg, >30 percent of cells positive). Because immunohistochemistry methods are not standardized across pathology departments, it may be difficult to use algorithms that rely on staining cutoffs in routine clinical practice, particularly at referral centers where cases and tissue blocks from multiple hospitals are seen in consultation.
Lymph2Cx platform — An alternative approach is the use of technologies that precisely quantify RNA transcript levels in formalin-fixed paraffin-embedded tissue sections. These technologies, which essentially permit transcripts to be counted, appear to be less sensitive to variations in tissue fixation and processing than immunohistochemistry and correlate well with results obtained by GEP on gene chips [22]. The Lymph2Cx is an example of a digital gene expression test that is highly concordant with other methods of GEP.
An analysis of 344 patients with de novo DLBCL treated with R-CHOP reported the following estimated outcomes at five years according to molecular subtype as defined by Lymph2Cx [17]:
●GCB-type DLBCL – Freedom from progression (FFP) 76 percent; progression-free survival (PFS) 73 percent; disease-specific survival (DSS) 82 percent; overall survival (OS) 78 percent.
●ABC-type DLBCL – FFP 51 percent; PFS 48 percent; DSS 61 percent; OS 56 percent.
These results suggest that the Lymph2Cx can distinguish two prognostically distinct groups of DLBCL. When widely available, Lymph2Cx will likely replace immunohistochemistry algorithms and GEP for COO assignment in DLBCL since it is more reliable than immunohistochemistry and more practical than GEP methods that require fresh, unfixed tissue.
MYC, BCL2, and BCL6 abnormalities — The most common cytogenetic abnormalities in DLBCL involve MYC, BCL6, and BCL2.
●MYC – Translocations involving MYC are seen in 5 to 15 percent of cases of DLBCL and can occur alone or in the context of additional rearrangements.
When treated with conventional immunochemotherapy, DLBCL with MYC rearrangement alone (ie, without concurrent rearrangements of BCL2 and/or BCL6) does not confer an adverse prognosis when using doxorubicin-based combination chemotherapy, such as R-CHOP [23].
The adverse prognostic impact of MYC rearrangement in "double-hit" lymphoma (eg, those with MYC and BCL2 rearrangements) appears to be restricted to tumors in which the rearrangement partner of MYC is an immunoglobulin gene, rather than another (ie, nonimmunoglobulin) partner [23].
MYC amplification occurs in 2 percent of DLBCL and may also be associated with worse outcomes [24].
●BCL6 – BCL6 translocations are found in up to one-third of DLBCL cases. There is a trend toward association with the ABC subtype, but the presence of this genetic aberration does not appear to have independent prognostic value [25].
●BCL2 – BCL2 translocations are found in approximately one-third of DLBCL cases, mostly in the GCB molecular subtype.
BCL2 translocations do not appear to influence survival when they are the sole genetic abnormality (eg, no concurrent MYC translocation) [26,27]. Expression of BCL2 protein in DLBCL does not correlate with the t(14;18) chromosomal rearrangement.
While the ABC subtype of DLBCL only rarely has the t(14;18), amplifications of 18q21 are seen in up to two-thirds of cases, providing a possible mechanism for BCL2 overexpression in these tumors. The impact of BCL2 protein expression on clinical outcomes is controversial [27,28], and additional studies are needed to clarify the prognostic impact of BCL2 expression in different subtypes of DLBCL.
Double-hit lymphoma — The term double-hit lymphoma is sometimes used colloquially to refer to cases of lymphoma that morphologically resemble DLBCL or high-grade B cell lymphoma, not otherwise specified (HGBCL, NOS), but they have translocations of MYC in combination with a BCL2 rearrangement (according to the 5th edition of the World Health Organization [2]) or in combination with BCL2 and/or BCL6 rearrangement (according to the International Consensus Classification [29]). (See "Epidemiology, clinical manifestations, pathologic features, and diagnosis of Burkitt lymphoma".)
Double-hit lymphomas have a worse prognosis than DLBCL when treated with R-CHOP (figure 4) [30-33]. Tumors with MYC, BCL2, and BCL6, so-called "triple-hit lymphomas" are not a discrete entity from double-hit lymphoma [23,30].
Double-expressor lymphoma — While all DLBCL with MYC translocation demonstrate increased MYC expression, some tumors have overexpression of MYC due to other mechanisms (eg, gene amplification). Overexpression of MYC and BCL2 can be identified with immunohistochemistry.
In several independent studies, coexpression of MYC and BCL2 was seen in 20 to 30 percent of DLBCL cases and predicted a lower complete response rate and shorter PFS and OS [31,32,34]. Furthermore, in an analysis of 893 patients with de novo DLBCL treated with R-CHOP, multivariate analysis demonstrated that coexpression of MYC/BCL2 was associated with inferior OS (hazard ratio [HR] 2.52; 95% CI 1.73-3.67) and PFS (HR 2.45; 95% CI 1.71-3.51) [33]. However, while the presence of an MYC translocation conferred a poor prognosis independent of BCL2, MYC protein expression was only associated with worse outcomes when accompanied by BCL2 expression.
Further studies are needed to clarify the prognosis of lymphomas that overexpress BCL2 and MYC based on immunohistochemistry. Cases with MYC overexpression identified by immunohistochemistry should undergo genetic testing for MYC translocations. Furthermore, since double-hit DLBCLs have a particularly poor prognosis, testing for BCL2 and BCL6 rearrangements should also be considered in cases exhibiting high levels of MYC protein expression. In general, double-expressor lymphomas without rearrangements of MYC and BCL2 are not considered a discrete diagnostic entity (ie, they are not distinguished from DLBCL, NOS).
Identifying molecular subtype — In order to subtype DLBCL and avoid misclassifying HGBCL with MYC and BCL2 and/or BCL6 gene rearrangements, we use the following tests to assess molecular risk in all lymphomas that have the morphologic features of DLBCL:
●Gene rearrangements – Evaluation of MYC, BCL2, and BCL6 gene status by fluorescence in situ hybridization (FISH) or cytogenetics.
An alternative is to perform immunohistochemistry for MYC and BCL2; cases with MYC overexpression on immunohistochemistry should have further testing for MYC, BCL2, and BCL6 rearrangements by FISH. Our approach is to perform MYC FISH studies only when MYC staining is positive in >40 percent of tumor cell nuclei. However, it should be noted that rare mutations of MYC can affect the recognition of MYC protein by anti-MYC antibodies.
●COO status – Evaluation of COO status by GEP analysis, immunohistochemistry, or Lymph2Cx platform.
Given their ease of use, immunohistochemistry algorithms (eg, Tally and Hans) (figure 3) are the most commonly used; concordance with GEP is approximately 80 percent.
Deep sequencing — Deep sequencing of DLBCL genomic DNA has confirmed that heterogeneity in DLBCL extends to the tumor cell genome [35-37]. Studies that included nearly 2000 cases of DLBCL identified driver mutations that may identify genetically distinct subtypes of GCB and non-GCB DLBCL, with different clinical outcomes following standard therapy [38-40]. These observations must be confirmed before genomic testing is used to stratify patients and select therapy.
Cell-free plasma DNA — Circulating cell-free DNA can be quantified in plasma by next-generation sequencing of IgH gene segments derived from DLBCL [41-43]. In initial studies, tumor DNA load was reported to correlate with imaging-based tumor stage and to be a sensitive predictor of disease relapse. This modality requires further evaluation before being used clinically to assess disease burden and response to therapy.
SUMMARY
●Description – Diffuse large B cell lymphoma (DLBCL) is the most common histologic subtype of non-Hodgkin lymphoma (approximately one-quarter of cases). DLBCL is morphologically, genetically, biologically, and prognostically heterogeneous.
●Prognostic factors – Outcomes are associated with both clinical features and pathologic aspects of DLBCL.
●Clinical features – Any of the following tools can be used to estimate prognosis for patients with DLBCL:
•International Prognostic Index (IPI) – The IPI (table 1) uses age, performance status (table 2), serum lactate dehydrogenase (LDH), disease stage (table 3), and number of extranodal sites to define prognostic categories. The IPI distinguishes two prognostic groups among patients treated with standard therapy (R-CHOP; rituximab, cyclophosphamide, doxorubicin, vincristine, prednisone), although it defined four groups in the prerituximab era. (See 'International Prognostic Index' above.)
•National Comprehensive Cancer Network (NCCN)-IPI—The NCCN-IPI uses the same clinical variables, but age and LDH are considered continuous rather than dichotomous variables, and the location of extranodal disease is used rather than the number of extranodal sites.
●Pathologic features
•Mutations – Evaluation of MYC, BCL2, and BCL6 expression or mutation is used to exclude certain lymphomas that resemble DLBCL (eg, so-called "double-hit" lymphoma, which is no longer considered a subtype of DLBCL) and may affect prognosis or treatment. (See 'MYC, BCL2, and BCL6 abnormalities' above.)
•Cell of origin status – Cell of origin (COO) studies using immunohistochemistry algorithms (eg, Tally and Hans) (figure 3) or molecular methods may predict responsiveness to therapy. (See 'Cell of origin status' above.)
There are two major COO subtypes:
-Germinal center B (GCB) cell type – Malignant cells resemble normal GCB cells and have more favorable outcomes with standard chemoimmunotherapy (figure 2).
-Non-GCB or activated B cell type – Malignant cells resemble activated B cells and are associated with inferior outcomes.
ACKNOWLEDGMENT —
The UpToDate editorial staff acknowledges Arnold S Freedman, MD, who contributed to earlier versions of this topic review.
