Fluorodeoxyglucose‐Positron Emission Tomography Predicts Bone Marrow Involvement in the Staging of Follicular Lymphoma

Abstract

Background

Standard bone marrow biopsy (BMB) and bone involvement with follicular lymphoma (FL) on positron emission tomography (PET)/computed tomography (CT) both predict early clinical failure in FL. The key clinical question is whether PET/CT findings can obviate the need for BMB. The goal of this study was to determine the value of PET/CT in determining bone involvement in FL, using posterior iliac crest BMB as the gold standard.

Materials and Methods

A total of 548 patients with newly diagnosed grade 1–3A FL were included. The presence, pattern, and location of bone involvement, spleen involvement, and standardized uptake values (SUVs) in the L3 vertebral body were recorded for all patients and compared with the BMB report.

Results

Excluding patients with focal bone lesions on PET/CT, the sensitivity and specificity of PET/CT in detecting bone or marrow involvement, compared with BMB, were 53% and 88%, respectively. The sensitivity and specificity of spleen involvement on PET/CT in predicting a positive BMB were 55% and 86%, respectively. An L3 SUV_max of less than 2.0 resulted in a negative predictive value (NPV) of 96% for marrow involvement on BMB; an L3 SUV_mean below 1.4 resulted in an NPV of 100%.

Conclusion

In newly diagnosed FL, PET/CT‐detected bone and splenic involvement is highly specific for a positive BMB, and very low SUV values (<2.0 SUV_max and < 1.4 SUV_mean) in the lumbar spine have a high NPV for a negative BMB. Routine BMB may be obviated in these patients. BMB remains necessary to definitively exclude bone marrow involvement in a large majority of patients with a negative PET.

Implications for Practice

Predicting early clinical failure in follicular lymphoma (FL) is important but difficult. Bone marrow involvement by FL is associated with early clinical failure, and determining this involvement is a key component of the initial staging. This study highlights that in certain patients, positron emission tomography/computed tomography is sufficient in determining bone or marrow involvement, without the need for a confirmatory bone marrow biopsy (BMB). An algorithm is provided based on these findings to help clinicians determine which patients would benefit from BMB and when it can be avoided.

Short abstract

To determine the value of PET/CT scans to confirm bone involvement in follicular lymphoma, this study compared different patterns of osseous FDG uptake detected by PET/CT with results of a standard posterior iliac crest bone marrow biopsy.

Introduction

Follicular lymphoma (FL) is recognized as an indolent lymphoma, with an estimated 10‐year overall survival (OS) of 80% 1. Although most patients with FL treated with standard therapy have a good prognosis, approximately 20% of patients will experience early clinical failure, defined as relapse within the first 2 years of immunochemotherapy 2, 3, 4. The ability to determine at diagnosis which patients will fall into this category remains limited 5, 6; however, certain factors can help predict early clinical failure. Bone marrow (BM) involvement with FL using a standard posterior iliac crest bone marrow biopsy (BMB) confers an estimated 10%–15% decrease in 5‐ to 10‐year OS in the Follicular Lymphoma International Prognostic Index 7. Positive BMB is also one of the components of the Follicular Lymphoma International Prognostic Index‐2 8 and is one of two factors in the simplified PRIMA‐prognostic index 9. We recently reported that the incidence and pattern of bone involvement detected by positron emission tomography (PET)/computed tomography (CT) is also associated with early clinical failure in untreated FL 10. Unlike in diffuse large B‐cell lymphoma or Hodgkin lymphoma, where PET/CT has been demonstrated to be accurate enough to obviate the routine need for BMB, there is no such evidence in FL 11, 12, 13, 14, 15, 16.

The goal of this study was to compare different patterns of osseous fluorodeoxyglucose (FDG) uptake detected by PET/CT with results of a standard posterior iliac crest BMB. We evaluated the correlation first using the BMB as the gold standard and then, because PET/CT is used for staging in all patients, we examined the need for a routine BMB if PET/CT results were considered first. These results have important implications for staging and prognosis.

Materials and Methods

Patients

Patients were identified using the Mayo Clinic Lymphoma Database and the University of Iowa/Mayo Clinic Lymphoma SPORE Molecular Epidemiology Resource database. Inclusion criteria consisted of patients with untreated FL, grades 1–3A, and ≥ 18 years of age in the PET/CT era (2003–2016). The year 2003 was selected to reflect the introduction of PET/CT within our institution. Patients were excluded if they had FL grade 3B, absent or inadequate pretreatment PET/CT imaging, or a subsequent biopsy within 1 week of diagnosis that revealed transformation. Patients were also excluded if they did not have a BMB reviewed by our institutional pathologists (Table 1). Patients provided permission to allow review of their medical record for research purposes, and the study was approved by our institutional review board.

Table 1.

Patient characteristics

Characteristics	Total (n = 548), n (%)
Age, years
Median	61.0
Q1, Q3	52.0, 68.5
Range	19.0–91.0
Gender
Female	262 (47.8)
Male	286 (52.2)
Lymphoma category 1
Follicular grade 3A	82 (15.0)
Follicular grade 3A, follicular grade 1	2 (0.4)
Follicular grade 3A, follicular grade 2	1 (0.2)
Follicular grade 3A, follicular grade 1	1 (0.2)
Follicular grade 3A, grade 2	1 (0.2)
Follicular grade 1	370 (67.5)
Follicular grade 1, CLL	1 (0.2)
Follicular grade 1–2	1 (0.2)
Follicular grade 2	89 (16.2)
Stage
Missing	2 (0.4)
I	1 (0.2)
IA	43 (7.8)
IAE	53 (9.7)
IBE	1 (0.2)
IE	1 (0.2)
II AE	8 (1.5)
II BE	1 (0.2)
IIA	59 (10.8)
IIAE	2 (0.4)
IIIA	98 (17.9)
IIIB	4 (0.7)
IVA	232 (42.3)
IVB	43 (7.8)
FLIPI Score
Missing	1 (0.2)
0	71 (13.0)
1	151 (27.6)
2	161 (29.4)
3	112 (20.4)
4	46 (8.4)
5	6 (1.1)
FLIPI‐2 Score
Missing	2 (0.4)
0	112 (20.4)
1	188 (34.3)
2	148 (27.0)
3	72 (13.1)
4	24 (4.4)
5	2 (0.4)
GELF Score
Missing	4 (0.7)
0	230 (42.0)
1	281 (51.3)
2	31 (5.7)
3	2 (0.4)
B_symptoms
Missing	2 (0.4)
No	496 (90.5)
Yes	50 (9.1)
Extranodal sites
No	271 (49.5)
Yes	277 (50.5)
No. of extranodal sites
None	271 (49.5)
1	212 (38.7)
2	54 (9.9)
3+	11 (2.0)
Bone only
Missing	419 (76.5)
Yes	129 (23.5)
Bone + other
Missing	488 (89.1)
Yes	60 (10.9)
No bone only other
Missing	460 (83.9)
Yes	88 (16.1)
Bone pain
Missing	3 (0.5)
No	534 (97.4)
Yes	11 (2.0)
Transformed
Missing	5 (0.9)
No	491 (89.6)
Yes	52 (9.5)
Follow‐up status
Alive	484 (88.3)
Deceased	64 (11.7)
EFS status
No event	332 (60.6)
Event	216 (39.4)
Years of follow‐up
Median	3.8
Range	0.0–14.2

Abbreviations: CLL, chronic lymphocytic leukemia; EFS, event‐free survival; FLIPI, Follicular Lymphoma International Prognostic Index; FLIPI‐2, Follicular Lymphoma International Prognostic Index‐2; GELF, Groupe d'Etude des Lymphomes Folliculaires.

FDG PET/CT Imaging

PET/CT was performed according to the standard institutional protocol, which required accurate height and weight and a blood glucose level of less than 200 mg/dL. The patients received 10–15 mCi of ¹⁸F‐FDG followed by a 60‐minute incubation period. PET images were acquired with two‐dimensional or three‐dimensional ordered‐subset expectation maximum, 128 × 128 matrix, and 3–5 minutes per bed position depending on patient body mass index. Low‐dose helical CT images were obtained for attenuation correction and anatomic localization. PET/CT image reviews were conducted blinded to clinical and histopathological data by a trained medical resident (F.S.‐P.) and nuclear radiologist (S.M.B.). MIM software (MIM Software Inc., Cleveland, OH) was used for PET/CT analysis. Data collected included presence of bone involvement, pattern and location of bone involvement, presence of spleen involvement, and a measurement of the mean and maximum standardized uptake value (SUV) in the L3 vertebral body. L3 was selected as a control site as it is near the posterior iliac crest where the BMB is performed and is an easily recognizable region for baseline collection of SUV in all PET scans. If L3 was involved by a focal lesion, the closest uninvolved vertebra above or below was used for background bone SUV analysis.

The PET imaging detects any involvement of the bone structure and the marrow contained in the medullary cavity, and we refer to this as “bone involvement.” The criteria for PET‐defined bone involvement, derived from Deauville criteria, 4, 17 were diffuse bone FDG uptake ≥ liver and/or the presence of focal osseous lesions with uptake ≥ liver. PET bone involvement was divided into three patterns: focal or multifocal, defined as discrete, separate areas of bone involvement; diffuse, defined as vertebral and pelvic FDG uptake equal to or higher than the liver; and focal or multifocal on diffuse, a mixed pattern with areas of focal involvement on a background of diffuse increased activity (Fig. 1). The spleen was considered positive on PET when its FDG uptake was ≥ that of the liver.

Figure 1.

Patterns of bone involvement with follicular lymphoma on positron emission tomography/computed tomography. (A): Negative for bone involvement; axial skeleton fluorodeoxyglucose uptake < liver; no focal lesions. (B): Positive for focal lesions on a background of diffuse bone involvement; a focal lesion with uptake ≥ liver is present in the thoracic spine, superimposed on diffusely increased activity throughout the pelvis and spine (including at L3) ≥ liver (focal/multifocal on diffuse pattern). (C): Positive for multifocal bone involvement; multiple lesions are present in the spine, pelvis, and scapulae with uptake ≥ liver. The background marrow activity, however, is normal (including at L3), with uptake < liver (focal/multifocal pattern). (D): Positive for diffuse bone marrow involvement; no focal lesions, but diffusely increase activity throughout the pelvis and spine (including at L3) > liver (diffuse pattern).

Bone Marrow Biopsy

All bone marrows were reviewed by a Mayo Clinic hematopathologist. A nontangential/nonsubcortical core of at least 1 cm in length was required to be considered adequate. Results from immunohistochemistry and flow cytometry testing were considered along with bone marrow histology to determine bone marrow positivity for lymphoma.

Statistical Analysis

The primary aim of our study was to determine the sensitivity and specificity of PET/CT in determining bone involvement by FL as compared with BMB as the “gold standard.” Secondary aims included determining the sensitivity and specificity of spleen involvement on PET/CT in predicting bone involvement on BMB as well as analyzing the correlation between L3 vertebral body SUV and bone involvement on BMB.

Results

Of the 613 patients initially included in the study, 89% (548/613) had a posterior iliac crest BMB result available as part of the staging evaluation. These 548 patients were included in our analysis.

Bone FDG Uptake on PET Versus Conventional Bone Marrow Biopsy

A total of 36% (197/548) of cases were positive for FL involvement by BMB pathology, and 34% (189/548) had FDG uptake in the bone detected by PET/CT. Assuming BMB as the gold standard for marrow involvement by FL, PET/CT analysis yielded 119 “true positives,” 70 “false positives,” 281 “true negatives,” and 78 “false negatives.” Based on these results, the sensitivity of PET/CT in detecting bone involvement, as compared with BMB, was 60%, with a specificity of 80%.

Of the 70 PET/CT scans that showed positive bone FDG uptake but were negative on BMB, 44% (31/70) had focal involvement without any diffuse component on PET/CT. Of these, 74% (23/31) had no focal involvement in the pelvic bones. Given that BMB is performed in the posterior iliac crest, a focal lesion outside of the pelvis would understandably be missed (Fig. 2). As such, we performed a secondary analysis including only patients with a diffuse component of bone involvement or no bone involvement on PET/CT (i.e., excluding Pattern C on Fig. 1). Based on these results, the sensitivity of PET/CT in detecting bone involvement, as compared with BMB, was lower at 53% (88/166) but with a higher specificity of 88% (281/320).

Figure 2.

Examples of focally positive bone involvement on the positron emission tomography/computed tomography with negative posterior iliac crest bone marrow. (A): Coronal images demonstrating a biopsy‐proven focal follicular lymphoma (FL) lesion in the left tibia in a patient with a negative posterior iliac crest bone marrow biopsy (BMB). (B): Biopsy‐proven focal FL lesion in the left scapula in a patient with a negative posterior iliac crest BMB.

We evaluated whether the accuracy of PET/CT in detecting bone marrow involvement changed based on extent of involvement on BMB. Patients with only focal bone involvement and without a diffuse component on PET/CT were excluded from this subanalysis for reasons stated previously. One hundred sixty‐two patients had both a quantified, positive BMB and either a negative PET/CT or a diffuse component of bone involvement on PET/CT. Our results (Table 2) showed that when there was less than 20% FL involvement on BMB, PET/CT detected only 43% (50/117) of cases. However, when there was 20% or more involvement of FL on BMB, PET/CT detected 80% (36/45) of cases.

Table 2.

PET/CT detection of diffuse BM involvement in 162 cases with a positive BMB for FL by % involvement on BMB

PET/CT results	% Marrow involvement with FL on BMB
	≤5 (n = 43)	5%–9% (n = 35)	10%–19% (n = 39)	20%–49% (n = 25)	50%–79% (n = 15)	≥80% (n = 5)
Positive for diffuse component	20 (43%)	12 (34%)	18 (46%)	19 (64%)	12 (80%)	5 (100%)
Negative for BM involvement	23	23	21	6	3	0

Abbreviations: BM, bone marrow; BMB, bone marrow biopsy; FL, follicular lymphoma; PET/CT, positron emission tomography/computed tomography.

Splenic PET/CT Involvement and Bone Marrow Involvement on BMB

In all, 29% (157/548) of patients had evidence of splenic involvement on PET/CT. Of these, 69% (109/157) also had a positive BMB for FL involvement. Of the 391 patients with negative spleen on PET/CT, 23% (88/391) had a positive BMB. Using BMB as the gold standard for determination of bone involvement, spleen involvement on PET/CT had a sensitivity of 55% (109/197) and a specificity of 86% (303/351) in predicting bone involvement.

Axial Skeleton SUV as a Predictor of BMB Results

SUV data (SUV_max and SUV_mean) were collected at L3 in the 448 patients who had either a negative PET/CT or a diffuse pattern of bone involvement. This was done to determine whether there were one or more PET‐derived SUV cutoff points at L3 that could be used as a surrogate for predicting bone involvement. Patients with focal bone lesions were not included in this analysis because they were already considered positive.

We analyzed the positive and negative predictive values (PPVs, NPVs) of all SUV values collected at L3. An NPV or PPV of ≥95% was considered significant. The NPV for an SUV_max of less than 2.0 was 96% (n = 25, or 5% patients classified as negative), and the NPV for an SUV_mean of less than 1.4 was 100% (n = 14, or 3% patients classified as negative). There was no SUV cutoff point with a PPV ≥95%.

In a receiver operating characteristic (ROC) analysis, the SUV_max value that resulted in the highest combined sensitivity and specificity was an SUV_max of 3.3, with a sensitivity of 46% and a specificity of 75%. The area under the curve (AUC) of this ROC analysis was 0.64.

Discussion

Existing evidence has been insufficient in determining whether PET/CT can replace BMB in FL. Only a few studies have investigated PET/CT use for determining bone marrow involvement in FL, most of which had a sample size of less than 150 patients 13, 18, 19, 20, 21, 22, 23.

Our analysis shows that the specificity of PET/CT for detection of bone marrow involvement is 88% when correcting for focal lesions (which may be missed on BMB) without diffuse FDG uptake. If PET/CT is done at staging and appears positive with a diffuse bone component, it is likely that the posterior iliac crest BMB will be positive and a BMB may not be necessary. In addition, spleen involvement on PET/CT also had a specificity of 86% for bone marrow involvement on BMB; therefore, patients with bone and/or spleen involvement by PET/CT can potentially be spared a BMB.

Other findings from this patient cohort, reported separately 10, further suggest that BMB does not add significant prognostic information to these patient subsets. In a multivariate analysis, spleen involvement on PET/CT predicted for lower EFS, irrespective of the BMB results. Similar results were found for certain patterns of bone involvement 10. In addition, BMB is likely unnecessary in patients with focal bone involvement on PET/CT outside the posterior iliac crest, as BMB performed in that location may miss bone involvement, and the BMB results would be unlikely to change clinical management. Based on these results, at least 16% of patients, and up to 39% of patients, would fit within a group where BMB is not necessary to confirm the presence or absence of bone or marrow involvement.

Although the specificity for predicting bone involvement is high, the sensitivity of PET/CT is only approximately 50%. Thus, in PET/CT‐negative cases one cannot assume a negative BMB. As expected, our results show that more extensive FL involvement on BMB correlates with improved detection by PET/CT. Namely, if FL involves 20% or more of the bone marrow, PET/CT detects 80% of cases; if the involvement is 80% or more, PET/CT detects involvement 100% of the time. We also found the SUV value at L3 to be helpful in predicting a negative BMB. If there is very low FDG uptake (<2.0 for SUV_max or < 1.4 for SUV_mean) at L3, which occurred in 5% of patients, there is >95% likelihood that the posterior iliac crest BMB will be negative. Although SUV values did provide useful information in terms of NPV, it should be noted that SUV values were neither sensitive nor specific in detecting bone involvement on BMB, as demonstrated by our ROC analysis (AUC = 0.64). Factors that can result in a false‐positive PET/CT include conditions that lead to increased marrow activity such as systemic infections, severe anemia, inflammatory disorders, recent growth factor use, and chemotherapy in post‐treatment scans.

The PET/CT‐derived factors of pattern of bone involvement, the presence of spleen involvement, and the SUV value at L3 are summarized in a decision‐making aid for routine clinical practice (Fig. 3). Patients who have no PET/CT findings suggesting marrow involvement can still be positive on BMB. We demonstrated that these cases typically have low levels of FL involvement and a favorable prognosis 10. Indeed, the PET/CT is most useful in detecting higher burden skeletal disease outside the posterior iliac crest 10. There are specific clinical instances where a BMB may be useful beyond this algorithm, such as when there is concern for transformation or the patient has significant cytopenias.

Figure 3.

Decision‐making algorithm for bone marrow biopsy at initial staging of follicular lymphoma using PET/CT data.Abbreviations: BMB, bone marrow biopsy; inv, involvement; NPV, negative predictive value; PET/CT, positron emission tomography; SUV, standard uptake value.

With respect to the spleen, our analysis highlights that there is a correlation between spleen involvement on PET/CT and bone involvement on BMB. Few data exist correlating BM and spleen involvement in lymphoma, and the pathophysiology behind this phenomenon is unclear. Perhaps this is due to both organs being part of the reticuloendothelial system, and clonal lymphocytes may proliferate in a similar fashion and rate in both the spleen and bone marrow.

Strengths of our study include a large data set (n = 548), uniform review of all PET/CTs, and all bone marrow biopsies reviewed by the same institutional hematopathology group.

A limitation to our study is that we did not confirm splenic involvement on PET/CT with spleen biopsy or splenectomy. Elevated FDG uptake in the spleen can be present in conditions such as systemic infections and systemic inflammatory disorders, which were not controlled for in our results. In addition, our SUV analysis is only beneficial for 5% of patients based on our results, given the low number of patients who had SUV values reaching NPV >95%. It should also be noted that SUV values may vary from one institution and from one scan to the other, and these SUV values should be verified using data from other institutions. It will be important to validate our findings in a separate data set, in prospective studies such as those conducted by large National Clinical Trials Network groups. Finally, the clinical significance of determining bone or marrow involvement in FL is currently uncertain, as it does not necessarily change immediate management. However, we expect that with future improvements in prognostication, we will hopefully be able to identify patients who require more aggressive treatment, and our findings may then have higher clinical impact.

Conclusion

In newly diagnosed FL, PET/CT‐detected diffuse pattern bone and splenic involvement is highly specific for a positive BMB. Very low SUV values (<2.0 SUV_max and < 1.4 SUV_mean) in the lumbar spine have a high negative predictive value for marrow involvement on BMB. Routine BMB may be obviated in these patients. However, given low sensitivity, BMB remains necessary to exclude BM involvement in a large majority of patients with a negative PET/CT.

Author Contributions

Conception/design: Frédérique St‐Pierre, Stephen M. Broski, Thomas M. Habermann, Thomas E. Witzig

Provision of study material or patients: Kay Ristow, Thomas M. Habermann, Thomas E. Witzig

Collection and/or assembly of data: Frédérique St‐Pierre, Stephen M. Broski, Betsy R. LaPlant, Matthew J. Maurer, Kay Ristow, William R. Macon, Thomas E. Witzig

Data analysis and interpretation: Frédérique St‐Pierre, Stephen M. Broski, Betsy R. LaPlant, Matthew J. Maurer, Kay Ristow, Gita Thanarajasingam, William R. Macon, Thomas M. Habermann, Thomas E. Witzig

Manuscript writing: Frédérique St‐Pierre, Stephen M. Broski, Betsy R. LaPlant, Matthew J. Maurer, Kay Ristow, Gita Thanarajasingam, William R. Macon, Thomas M. Habermann, Thomas E. Witzig

Final approval of manuscript: Frédérique St‐Pierre, Stephen M. Broski, Betsy R. LaPlant, Matthew J. Maurer, Kay Ristow, Gita Thanarajasingam, William R. Macon, Thomas M. Habermann, Thomas E. Witzig

Disclosures

Matthew Maurer: Celgene/Bristol‐Myers Squibb, Nanostring (RF), Morphosys, Kite (C/A). The other authors indicated no financial relationships.

(C/A) Consulting/advisory relationship; (RF) Research funding; (E) Employment; (ET) Expert testimony; (H) Honoraria received; (OI) Ownership interests; (IP) Intellectual property rights/inventor/patent holder; (SAB) Scientific advisory board