THYMIDINE KINASE (TK-1) IN PROGNOSTIC INDICES FOR DLBCL
20210199659 · 2021-07-01
Inventors
- Vinzent Rolny (Muenchen, DE)
- Sandra Rutz (Muenchen, DE)
- David Morgenstern (Indianapolis, IN, US)
- Boris Pinchuk (Weilheim in Oberbayern, DE)
- Christina Zimmermann (Huglfing, DE)
- Martin KLAMMER (Muenchen, DE)
- Franziska Sonner (Benediktbeuern, DE)
Cpc classification
G01N33/54313
PHYSICS
International classification
Abstract
The present disclosure relates to the finding that thymidine kinase 1 (TK-1) represents a valuable biomarker in a method for determining a Prognostic Index (PI) for risk stratification of a patient with aggressive B-cell lymphoma, especially diffuse large B-cell lymphoma (DLBCL), to the use of TK-1 in such PI and to a PI comprising the marker TK-1.
Claims
1. A method of determining a prognostic index (PI) (also called risk score or PI-score) for a diffuse large B-cell lymphoma (DLBCL) patient the method comprising a) determining at least the following parameters: i) extranodal disease status; ii) Ann Arbor stage; and iii) the level of thymidine kinase 1, wherein absence of extranodal disease values 0 points and presence of extranodal disease values 1 point, wherein an Ann Arbor stage of I or II values 0 points and an Ann Arbor stage of III or IV values 1 point, and wherein a level of the thymidine kinase up to and including the cut-off values 0 points and wherein a level of the thymidine kinase above cut-off values at least 1 point, and b) summing up the values for i), ii) and iii) in the determination of the PI.
2. The method of claim 1, wherein the thymidine kinase cut-off level is set to 4 times the mean value of the TK-1 level as determined in healthy individuals, wherein a thymidine kinase level up to and including the cut-off values 0 points and wherein any TK-1 level above cut-off values 1 point.
3. The method of claim 1, wherein the thymidine kinase cut-off level is set to 4 times the mean value of the TK-1 level as determined in healthy individuals, wherein a low thymidine kinase level, i.e. a level of TK-1 up to and including the cut-off values 0 points, an intermediate thymidine kinase level, i.e. a level of TK-1 above the cut-off and up to and including the value corresponding to 20 times the mean value as determined in healthy individuals values 1 point, and a high thymidine kinase level, i.e. a level of TK-1 of more than 20 times the mean value as determined in healthy individuals values 2 points.
4. The method of claim 1, wherein the cut-off on the protein level for TK-1 is equivalent to and includes 18 U/l in TK-1 enzymatic activity.
5. The method of claim 1, wherein the cut-off on the protein level is equivalent to and includes 18 U/Iland wherein a level of TK-1 above 18 U/l and up to and including 88 U/l values 1 point and a level of TK-1 above 88 U/l values 2 points.
6. The method of claim 1, further comprising the parameter age, wherein if the age is below cut-off, the value is 0 points and if the age is above cut-off, the value is 1 point in the determination of the PI.
7. The method according to claim 6, wherein if the age is below and including 60, the value is 0 points, wherein if the age is above 60, the value is 1 point in the determination of the PI.
8. The method according to claim 1, further comprising the parameter ECOG performance status, wherein, in case the ECOG performance status 1 or below, the value is 0 points and wherein, in case the ECOG performance status is 2 or above, the value is 1 point in the determination of the PI.
9. The method according to claim 1, further comprising the parameter age and the parameter ECOG performance status, wherein if the age is below and including 60, the value is 0 points, wherein if the age is above 60, the value is 1 point in the determination of the PI, wherein, in case the ECOG performance status is 1 or below the value is 0 points, and wherein, in case the ECOG performance status is 2 or above, the value is 1 point in the determination of the PI.
10. The method according to claim 1, further comprising the parameter age and the parameter ECOG performance status, wherein if the age is below and including 40, the value is 0 points, wherein if the age is above 40 and up to and including 60, the value is 1 point, wherein if the age is above 60 and up to and including 75, the value is 2 points and wherein if the age is above 75 the value is 3 points in the determination of the PI, and wherein, in case the ECOG performance status is 1 or below the value is 0 points, and wherein, in case the ECOG performance status is 2 or above, the value is 1 point in the determination of the PI.
11. Use of a value for the level of TK-1 in the determination of a prognostic index for patients with DLBCL.
12. The use according to claim 11, wherein the value for the level of TK-1 is combined with the other parameters as used in IPI, age adjusted IPI, R-IPI, or NCCN-IPI.
13. The use according to claim 12 with the proviso that the value for the level of TK-1 is used to substitute the value for the level of LDH in such PI.
14. A prognostic index (PI) for DLBCL comprising a value for the level of thymidine kinase 1.
15. The prognostic index (PI) of claim 14, wherein said PI is selected from IPI, age adjusted IPI, R-IPI, or NCCN-IPI.
Description
BRIEF DESCRIPTION OF THE FIGURES
[0151] The patent or patent application file contains at least one figure executed in color. Copies of this patent or patent application publication with color figure(s) will be provided by the Office upon request and payment of the necessary fee.
[0152]
[0153]
[0154]
[0156] The following Examples illustrate the invention:
EXAMPLE 1: MATERIALS & GENERAL METHODS
[0157] Protein Chemistry and Labeling Techniques
[0158] Standard protein chemistry and labeling techniques are provided e.g. in Hermanson, G. “Bioconjugate Techniques” 3rd Edition (2013) Academic Press.
[0159] Bioinformatics
[0160] Bioinformatics methods are provided in e.g. Keith J. M. (ed.) “Bioinformatics” Vol. I and Vol. II, Methods in Molecular Biology Vol. 1525 and Vol. 1526 (2017) Springer, and in Martin, A. C. R. & Allen, J. “Bioinformatics Tools for Analysis of Antibodies” in: Dübel S. & Reichert J. M. (eds.) “Handbook of Therapeutic Antibodies” Wiley-VCH (2014).
[0161] Electrochemiluminescent Immunoassays
[0162] Immunoassays and related methods are provided in e.g. Wild D. (ed.) “The Immunoassay Handbook” 4th Edition (2013) Elsevier. Ruthenium complexes as electrochemiluminescent labels are provided in e.g. Staffilani M. et al. Inorg. Chem. 42 (2003) 7789-7798. Typically, for the performance of electrochemiluminescence (ECL) based immunoassays an Elecsys 2010 analyzer or a successor system was used, e.g. a Roche analyzer (Roche Diagnostics GmbH, Mannheim Germany) such as E170, cobas e 601 module, cobas e 602 module, cobas e 801 module, and cobas e 411, and Roche Elecsys assays designed for these analyzers, each used under standard conditions, if not indicated otherwise.
EXAMPLE 2: ANTI-hTK-1 ANTIBODIES
[0163]
TABLE-US-00003 hTK-1, clone 6C6, heavy chain: (SEQ ID NO: 3) METGLRWLLLVAVLKGVQCQEQLEESGGDLVKPEGS LTLTCTASRFSFSSSYWICWVRQAPGKGLEWIACIY AGDSGSSYYASWAKGRFTVSKTSSTTVTLQTTSLTA ADTATYFCARASVGAAYDYFALWGPGTLVTVSSGQP KAPSVFPLAPCCGDTPSSTVTLGCLVKGYLPEPVTV TWNSG hTK-1, clone 6C6, light chain: (SEQ ID NO: 4) MDTRAPTQLLGLLLLWLPGARCALVMTQTPASVEAA MGGTVTIKCQASEDVSSHLAWYQQRPGQPPKLLIYG ASDLASGVPSRFTGSGSGTQFTLAISDLECADAATY YCQGYYYISDSPYVFGGGTEVVVKGDPVAPTVLIFP PAADQVATGTVTIVCVANKYFPDVTVTWEVDGTTQT TGIENSKTPQNSADCTYNLSSTLTLTSTQYNSHKEY TCKVTQGTTSVVQSFNRGDC hTK-1, clone 4H4, heavy chain: (SEQ ID NO: 7) METGLRWLLLVAVLKGVQCQSLEESGGGLVQPEGSL TLTCTASGFSFSSGYDMCWVRQTPGKGLEWIACEVD SDGVTYYASWAKGRFTISKTSSTTVTLQMTSLTAAD TATYFCARGYESSSGVYIPYFTLWGPGTLVTVSSGQ PKAPSVFPLAPCCGDTPSSTVTLGCLVKGYLPEPVT VTWNSG hTK-1, clone 4H4, light chain: (SEQ ID NO: 8) MDMRAPTQLLGLLLLWLPGARCADIVLTQTPASVEA AVGGTVTIKCQASQSIYSYLAWYQHKPGQPPKLLIY KASTLASGVPSRFKGSGSGTEYTLTISDLECADAAT YYCQHYYYSSTSGGGVFGGGTEVVVKGDPVAPTVLI FPPAADQVATGTVTIVCVANKYFPDVTVTWEVDGTT QTTGIENSKTPQNSADCTYNLSSTLTLTSTQYNSHK EYTCKVTQGTTSVVQSFNRGDC hTK-1, clone 23C11, heavy chain: (SEQ ID NO: 9) METGLRWLLLVAVLKGVQCQSLEESGGRLVTPGTPL TLTCTASGFSLSNYYMSWVRQAPGKGLEWIGIIYGD DNTYCANWTKGRFTISKTSTTVDLTITSPTTEDTAT YFCARGPDYIAAKMDIWGPGTLVTVSLGQPKAPSVF PLAPCCGDTPSSTVTLGCLVKGYLPEPVTVTWNSG hTK-1, clone 23C11, light chain: (SEQ ID NO: 10) MDTRAPTQLLGLLLLWLPGARCDVVMTQTPASVEAA VGGTVTIKCQASQSISGYLSWYQQKPGQRPKLLIYR ASTLESGVPSRFKGSGSGTEFTLTISDLECADAATY YCQCTYGSSTFSSYGNAFGGGTEVVVKGDPVAPTVL IFPPAADQVATGTVTIVCVANKYFPDVTVTWEVDGT TQTTGIENSKTPQNSADCTYNLSSTLTLTSTQYNSH KEYTCKVTQGTTSVVQSFNRGDC
[0164] As obvious full length immunoglobulins or any binding fragments thereof—if desired/required—can be easily construed by any person of skill in the art based on sequences disclosed above.
EXAMPLE 3: EPITOPE CHARACTERIZATION
[0165] As described in Example 2, four different monoclonal antibodies could be generated that exhibit the required binding properties needed for them to be of utility in an immunoassay development.
[0166] In a first attempt to characterize the epitope bound by the newly generated antibodies a PepScan analysis was performed. For this analysis synthetic peptides, consisting of 15 amino acids each, each shifted by 1 amino acid (1-15: 2-16, etc.) and spanning the entire sequence of hTK-1 (SEQ ID NO:1) were synthesized. These PepScan peptides were spotted onto microscope slides. After blocking for non-specific binding cell culture supernatant of the various MABs was incubated on the microscope slides. Unbound MAB was washed off and bound MAB was detected by use of HRP-labeled goat anti-rabbit IgG according to a routine method.
[0167] Only one of the four MABs (antibody 4H11) obtained by the method of Example 2 did react with a linear epitope. As could be shown the epitope bound by this antibody is comprised in the sequence spanning amino acid residues 211 through 230 of hTK-1 (SEQ ID NO: 6).
[0168] Polyclonal as well as monoclonal antibodies reacting with a synthetic peptide corresponding to a polypeptide consisting of amino acids 194 through 225 of hTK-1 (SEQ ID NO:2) are known in the prior art, see e.g. WO 2015/094106. The three MABs 4H4; 6C6; and 23C11, respectively, did not bind to a polypeptide consisting of amino acids 194 through 225 of hTK-1 (SEQ ID NO:2) nor did they show significant binding to any of the PepScan peptides tested. This indicates that these three MABs all bind to a conformation dependent epitope on hTK-1. As these three MABs from the very beginning looked quite promising for further assay development, additional efforts were made in order to gain more knowledge on the epitope bound by these three MABs.
[0169] The epitope characterization by competition experiments was performed on a GE Healthcare Biacore 4000 instrument at 25° C. A Biacore Biotin Capture Kit, Series S sensor (Cat.-No. 28-9202-34) was mounted into the instrument and was hydrodynamically addressed and preconditioned according to the manufacturer's instructions. The system buffer was HBS-N (10 mM HEPES pH 7.4, 150 mM NaCl). The sample buffer was the system buffer. The biotin capture reagent, as provided by the manufacturer GE Healthcare, was diluted 1:50 in system buffer and was injected at 10 μl/min for 60 sec over flow cells 1, 2, 3 and 4 to address the spots 1, 2 and 4, 5. Spot 3 served as a reference. 10 nM biotinylated primary antibody was injected at 30 μl/min for 120 sec contact time to address the spots 1 and 5 in all four flow cells. Spots 2 and 4 served as controls. 10 nM human recombinant thymidine kinase-1 (hTK-1, Roche, 114 kDa, tetramer) were injected at 30 μl/min into all flow cells for 180 sec contact time to address the spots 1, 2 and 4, 5. 100 nM of non-biotinylated primary antibody were again injected at 30 μl/min to address the spots 1, 2 and 4, 5 on all flow cells for 180 sec contact time in order to block remaining accessible epitopes of the primary antibody. 100 nM of secondary antibody was injected at 30 μl/min for 180 sec contact time into all flow cells to address the spots 1, 2 and 4, 5. Finally the complexes formed on the sensor surface were completely removed by a 120 sec contact time regeneration step over all flow cells and all spots using the regeneration solution as provided by the manufacturer GE Healthcare.
[0170] Four recombinant monoclonal rabbit IgG antibodies were this way investigated for their hTK-1 epitope accessibility properties: Antibody 4H11 (an antibody binding to an epitope comprised in amino acids 211 to 230 (SEQ ID NO:6) of hTK-1 and rabbit MABs 23C11, 6C6 and 4H4.
[0171] Before and after each sample injection, report points were set. The read out of the report points in Response Units [RU] was done by using the Biacore Evaluation V.1.1 software.
[0172] To the initial biotinylated primary antibody capturing signal (bi-Ab1, [RU]) the second binding response signal of the non-biotinylated primary antibody (block Ab1 [RU]) was added. The Molar Ratio Epitope Accessibility MR.sub.EA=Ab2 [RU]/(bi-Ab1 [RU]+block Ab1 [RU]) was calculated and was used as an estimate for the epitope accessibility of the respective antibodies used in the assay.
[0173] In order to verify the tetrameric state of the hTK analyte, a second Molar Ratio was calculated from the hTK binding signal versus the capture level of the biotinylated primary antibody by using the formula MR=hTK [RU]/bi-Ab1 [RU]*Molecular Weight bi-Ab1 (150 kDa)/Molecular Weight hTK (114 kDa).
[0174] For example, antibody 4H11 showed a binding stoichiometry (=Molar Ratio) antibody 4H11/hTK-1 MR of 1:1. In this biosensor assay a single, tetrameric hTK-1 molecule binds to a single antibody 4H11 molecule. Within the described antibodies, only antibody 4H11 shows a homologous hTK-1 complex formation when being used as block Ab. Therefore, a sandwich assay would be possible by using a sequential assay protocol using antibody 4H11 twice. No homologous complex formation could be detected for the rabbit monoclonal antibodies MABs 23C11, 6C6 and 4H4. This means that MABs 23C11, 6C6 and 4H4 bind to the same epitope region. Antibodies 23C11, 6C6 and 4H4 form a sandwich with antibody 4H11 as secondary antibody. According to this assay the best performing sandwich pair is 6C6 as biotinylated primary antibody, which forms a complex with antibody 4H11 showing a Molar Ratio MR.sub.EA=0.4, which means 40% epitope accessibility on the hTK analyte.
[0175] As obvious from the table shown below, the antibody 4H11 (binding to a C-terminal linear epitope comprised in SEQ ID NO:6) is able to form immuno complexes with 23C11, 6C6 and 4H4. On the other hand it is clear that the rabbit monoclonal antibodies MABs 23C11, 6C6 and 4H4 share the same epitope (see below Tab. 3).
TABLE-US-00004 TABLE 3 Epitope Accessibility Matrix biotinylated primary secondary antibodies antibodies 4H11-Ig6 23C11-IgG 6C6-IgG 4H4-IgG 4H11-IgG 0.1 0.1 0.1 0.1 23C11-IgG 0.2 0.0 0.0 0.0 6C6-IgG 0.4 0.0 0.0 0.0 4H4-IgG 0.2 0.0 0.0 0.0
[0176] Shown is the sandwich formation of four anti-hTK-1 antibodies using recombinant h-TK-1 as an analyte in solution. A value of 0.0 in the table indicates that the first and the second antibody used bind to the same epitope. A value of 0.1 or higher indicates sandwich formation, despite the intermediate blocking step, i.e. the two antibodies investigated bind to different epitopes.
EXAMPLE 4: PRODUCTION OF THE MAB-CONJUGATES FOR USE IN ELECSYS IMMUNOASSAY EXPERIMENTS
[0177] The procedures used are familiar to the skilled artisan. It therefore is considered redundant to give experimental details.
[0178] In brief, the following procedures/steps were carried out in order to obtain the antibody conjugates for the capture and detection sides of the immunological assay.
[0179] Cell culture supernatant (the recombinant antibodies comprised therein) as obtained from the by B-cell PCR generated cells (see above) was used as a starting material.
[0180] The recombinant antibody comprised in the tissue culture supernatant was purified by affinity chromatography to protein A.
[0181] An antibody used as a capture antibody was cleaved to the F(ab′)2 fragment with pepsin and the F(ab′)2 fragment further purified by affinity chromatography and size-exclusion chromatography. The F(ab′)2 fragment was then reduced to Fab′ and site-specific biotinylated via thiol-chemistry, thereby obtaining a mono-biotinylated Fab′-fragment.
[0182] An antibody used as a detection antibody was chemically conjugated to sulfo-Ruthenium (WO 2003/002974) by use of a sulfo-BPRu NHS Ester (=CAS Reg. Number 482618-42-8 also known in the art as ruthenate(2-), bis[[2,2′-bipyridine]-4, 4′-dimethanesulfonato(2-)-.sup.1, 1′][1-[4-(4′-methyl[2,2′-bipyridin]-4-yl-.sup.1, 1′)-1-oxobutoxy]-2,5-pyrrolidinedione]-, sodium (1:2), (OC-6-31)) and unbound label was removed by size-exclusion chromatography.
EXAMPLE 5: SAMPLES AND hTK-1 MEASUREMENTS
[0183] 5.1 Samples
[0184] A “black-and-white” panel was investigated. On the one hand serum samples from 50 (for some experiments only 49 were still available) healthy donors have been used for quantification of hTK-1 in various assays. On the other hand hTK-1 was measured in 48 (for some experiments only 47 were still available) samples from patients with diffuse large B-cell lymphoma (DLBCL).
[0185] 5.2 Prototype Electrochemiluminescence Immunoassays
[0186] Based on the monoclonal rabbit antibodies obtained as described in Example 3, purified and conjugated as described in Example 4, respectively, several prototype immunoassays have been established.
[0187] The typical set-up of a prototype electrochemiluminescent immunoassay makes use of biotinylated capture antibody (or an antigen binding fragment thereof) and a detection antibody (or an antigen binding fragment thereof) which is labeled with a ruthenium complex.
[0188] Immunoassay data in most cases were generated using a conventional Fab′ fragment that was biotinylated according to conventional procedures. Measurements were carried out in a sandwich assay format on a Cobas® E170 analyzer from Roche. Signal detection in the Cobas® E170 analyzer is based on electrochemiluminescence. In this sandwich assay the biotin-conjugate (i.e. the capture antibody) is immobilized on the surface of a streptavidin-coated magnetic bead. The detection-antibody bears a complexed ruthenium cation as the signaling moiety. In the presence of analyte, the chromogenic ruthenium complex is bridged to the solid phase and emits light at 620 nm after excitation at the platinum electrode comprised in the measuring cell of the Cobas® E170 analyzer. The signal output is in arbitrary light units.
[0189] Measurements were, e.g., performed with calibrators spiked with recombinant hTK-1 from HEK-cells as well as with the human serum samples mentioned above.
[0190] The experimental hTK-1 assay was conducted as follows. 25 μl of human serum sample or of spiked calibrator, 25 μl of pretreatment reagent (comprising 10 mM DTBA) were mixed an incubated for 9 minutes; thereafter 60 μl of capture antibody-biotin conjugate and 60 μl of detection antibody ruthenium label conjugate were incubated together for another 9 minutes followed by the addition of 30 μl streptavidin-coated paramagnetic microparticles. The final mixture was incubated for further 9 minutes. Afterwards, the hTK-1 was detected as usual (i.e. via the electrochemiluminescent signal generated in these experiments).
[0191] Below given are the results obtained for the combination of biotinylated 6C6 (used as Fab′-Bi) with ruthenylated 4H11 (used as IgG).
[0192] Calibration results are given in Table 4 below.
TABLE-US-00005 TABLE 4 Measurement of hTK-1 with the immunoassay prototype TK-1: Prototype Assay 6C6-Fab′-Bi: 4H11-IgG-suBPRu [hTK-1] in ng/mL Counts Conc MW.sub.conc 0.0000 1837 0.0000 0.0000 1826 0.0000 1.00 30193 1.30 1.30 30103 1.30 5.00 105843 4.92 4.91 105452 4.90 10.0 198859 9.47 9.60 204252 9.73 100 1960303 99.5 100 1994291 101
[0193] In Table 4 signals obtained in the prototype immunoassay using various amounts of recombinant hTK-1 as analyte are given. Double measurements have been performed.
[0194] Box-Whisker-Plots have been calculated, the receiver-operator-characteristic (ROC) has been analyzed and the area under the curve (AUC) was determined. For the prototype immunoassay the AUC was 0.965. Both Box-Whisker-Plots (BoxPlots) with the determined concentrations of hTK-1 and the AUCs are shown in
[0195] 5.3 DiaSorin TK-1 Activity Assay
[0196] The LIAISON® Thymidine Kinase assay, manufactured by DiaSorin, is an indirect, modified two-step, competitive chemiluminescence immunoassay (CLIA) for the quantitative determination of TK in human serum and EDTA plasma. The LIAISON® Thymidine Kinase assay was performed according to the instructions given by the manufacturer with 50 control samples and 48 samples from patients with DLBCL. It utilizes an initial enzymatic reaction in which TK in the sample converts AZT (3′-azido-3′-deoxythymidine) to AZTMP (3′-azido-3′-deoxythymidine mono phosphate), this is followed by a competitive immunoassay for the quantitative determination of AZTMP. The amount of AZT converted to AZTMP is a measure of the amount of TK present in the sample In the assay, 50 μL of sample is incubated with 100 μL of Assay Buffer 1, 20 μL of Assay Buffer 2, and 20 μL of paramagnetic particles coated with anti-AZTMP polyclonal antibody. Rabbit anti-goat IgG, then anti-AZTMP goat polyclonal is coated to the solid phase. This is incubated for 40 minutes and then 100 μL of tracer, an AZTMP analogue conjugated to an isoluminol derivative is added. During the first incubation, AZTMP binds to the solid phase. In the second incubation, the tracer conjugate competes for binding with the AZTMP in the solution. After a 20 minute incubation, the unbound material is removed with a wash cycle. The starter reagents are then added and a flash chemiluminescent reaction is initiated. The light signal is measured by a photomultiplier as relative light units (RLU) and is proportional to the concentration of TK present in calibrators, controls, or samples.
[0197] Box-Whisker-Plots have been calculated, the receiver-operator-characteristic (ROC) has been analyzed and the area under the curve (AUC) was determined. For LIAISON® Thymidine Kinase assay the AUC was found to be 0.958. Both Box-Whisker-Plot (Boxplot) and the AUC are shown in
EXAMPLE 6: COMPARISON OF TK-1 VALUES AS DETERMINED WITH ACTIVITY ASSAY/IMMUNOASSAY
[0198] The values obtained with LIAISON® Thymidine Kinase assay on the one hand and the prototype immunoassay on the other hand were compared to each other. Taking into account that one assay measure thymidine kinase activity while the other measures the amount of immunoreactive hTK-1 a surprisingly high correlation (in the range of 0.95 or even above—dependent on the statistical method used) between the two different assays was found. The good correlation between these different assays for hTK-1 is also quite obvious from
EXAMPLE 7: USE OF TK-1 VALUES AS KEY ELEMENT OF A PROGNOSTIC INDEX
[0199] 7.1 Analytical Approach
[0200] The prognostic ability of thymidine kinase 1 (TK-1) was retrospectively assessed with samples from the pharmacological MAIN study, which was conducted with patients suffering from diffuse large B-cell lymphoma (DLBCL) [Seymour et al., 2014]. TK-1 concentrations in serum were measured with the assay termed prototype B) in Example 5.2 for each of the 407 baseline samples that were still available. The progression-free survival (PFS) event rate within this subset was highly similar to the PFS of the total study population (both approximately 31%).
[0201] In order to have a fair comparison of all multivariate models, we created a data subset, such that only samples with all relevant data points available (TK-1 measurement and all index components) were included. This resulted in the final analysis data set containing 370 samples, of which 116 stemmed from patients who exhibited a PFS event within three years after R-CHOP treatment start. Due to the small number of samples with an event-free follow-up time longer than three years, samples were censored at the 3-year mark.
[0202] In order to assess whether the prognostic value of TK-1 is independent from known clinical and demographic risk factors, Cox proportional hazard models for each of the three described prognostic indices including the respective index's risk components plus log 2-transformed TK-1 values were calculated.
[0203] 7.2 Contribution of TK-1 to R-IPI and NCCN-IPI
[0204] To assess the ability of TK-1 to improve the existing risk scores for prognosis of PFS, the index components were either extended with log 2-transformed TK-1 values, or the LDH component was replaced with the log 2 TK-1 values. Both extension and substitution was performed by creating a Cox portioned hazard model including the respective risk components as independent variables, and comparing the c-indices of the original models (R-IPI and NCCN-IPI) with the c-indices of TK-1-extended models (with and without the LDH component, respectively). The significance in terms of p-values of the respective differences were assessed with a bootstrap resampling approach.
[0205] a) TK-1 as Covariate in the R-IPI
[0206] Results of the statistical analysis of the contribution provided by TK-1 to the R-IPI are given below in Table 5.
TABLE-US-00006 TABLE 5 Multivariate Cox proportional hazard model including R-IPI's risk components binarized according to the thresholds in Table 1 plus TK-1 (log2-transformed). The hazard ratios are given along with their respective 95% confidence interval and the p-value Hazard ratio 95%-CI Covariate (HR) HR p-value IPI.Age (bin) 1.083 0.750-1.564 0.671 IPI.AnnArborStage (bin) 1.577 0.971-2.561 0.066 IPLECOG (bin) 1.829 1.226-2.731 0.003 IPI.LDH (bin) 0.912 0.582-1.429 0.688 IPI.ExtranodalSites (bin) 1.183 0.785-1.784 0.422 TK-1 (log2) 1.307 1.127-1.516 <0.001
[0207] The Cox proportional hazard model combining the binary R-IPI risk components with TK-1 (log 2-transformed) clearly shows that TK-1 substantially adds prognostic information to the model. The hazard ratio (HR) is approximately 1.3 with the smallest p-value of all model components of <0.001 (see Table 5). The hazard ratio can be interpreted in a way that a 2-fold increase of TK-1 is associated with a 1.3-fold increase of risk for a PFS event (i.e. disease progression or death).
[0208] For the trichotomized TK-1 scenario, TK-1 concentrations were first stratified into three groups (i.e. low, intermediate and high levels). The optimal cutoffs were demined in a univariate setting by maximizing the log rank statistic (i.e. maximizing the different between the three survival curves in a Kaplan-Meier analysis), while ensuring that at least 10 samples were present in each group. The herewith determined ranges are: [0209] TK-1 low=0%-24% (0-0.7 ng/ml) [0210] TK-1 intermed.=24.1%-86% (0.71-3.5 ng/ml) [0211] TK-1 high=86.1%-100% (above 3.5 ng/ml)
[0212] Based on the method comparison between Roche Elecsys TK-1 concentration and Diasorin TK-1 activity measurements, these ranges can be further expressed in terms of U/L based on the relationship 1 ng/ml TK-1 (Roche)::25 U/L TK-1 (Diasorin). The mean activity in healthy individuals was reported to be 4.3 U/L by Diasorin, which is used as a relative reference for the cutoffs: [0213] TK-1 low=0-4 times the mean (0-18 U/L) [0214] TK-1 intermed.=4-20 times the mean (17.6-88 U/L) [0215] TK-1 high=above 20 times the mean (above 88 U/L)
[0216] In the multivariate Cox model, the trichotomized TK-1 values significantly contribute to the prognostic ability with a p-value of 0.013 and <0.001 for intermediate versus low and high versus low TK-1 levels, respectively (see Table 6). The corresponding hazard ratios are 2.11 and 5.40, meaning that a patient with an intermediate TK-1 level has an approximately 2.1-fold higher risk for a PFS event over low-TK-1 patients; and a patient with a high TK-1 level even 5.4-fold.
TABLE-US-00007 TABLE 6 Multivariate Cox proportional hazard model including R-IPI's risk components binarized according to the thresholds in Table 1 plus TK-1 (trichotomized). The hazard ratios are given along with their respective 95% confidence interval and the p-value. Hazard ratio 95%-CI Covariate (HR) HR p-value IPI.Age (bin) 1.040 0.720-1.503 0.834 IPI.AnnArborStage (bin) 1.582 0.978-2.560 0.062 IPI.ECOG (bin) 1.939 1.302-2.889 0.001 IPI.LDH (bin) 0.813 0.522-1.268 0.362 IPI.ExtranodalSites (bin) 1.163 0.771-1.753 0.471 TK-1.intermed (bin) 2.113 1.171-3.814 0.013 TK-1.high (bin) 5.399 2.659-10.96 <0.001
[0217] b) TK-1 as a Covariate in the NCCN-IPI
[0218] Results of the statistical analysis of the contribution provided by TK-1 to the NCCN-IPI are given below in Table 7.
TABLE-US-00008 TABLE 7 Multivariate Cox proportional hazard model including NCCN-IPI's risk components binarized according to the thresholds in Table 2 plus TK-1 (log2-transformed). The hazard ratios are given along with their respective 95% confidence interval and the p-value Hazard ratio 95%-CI Covariate (HR) HR p-value NCCN.Age40-60 (bin) 1.978 0.962-4.064 0.064 NCCN.Age60-75 (bin) 1.569 0.762-3.279 0.221 NCCN.Age > 75 (bin) 3.377 1.437-7.932 0.005 NCCN.LDH1-3 (bin) 0.967 0.612-1.577 0.886 NCCN.LDH > 3 (bin) 1.557 0.708-3.424 0.271 NCCN.AnnArborStage (bin) 1.593 0.986-2.575 0.057 NCCN.ExtranodalDisease (bin) 1,003 0.563-1.785 0.993 NCCN.ECOG (bin) 1.795 1.201-2.682 0.004 TK-1 (log2) 1.248 1.056-1.476 0.009
[0219] The multivariate model combining the binary NCCN-IPI risk components with TK-1 (log 2-transformed) also shows significant added prognostic value of TK-1 (p-value=0.009, see Table 7). The hazard ratio (HR) is approximately 1.25, which is only slightly smaller than in the R-IPI model.
[0220] Results of the statistical analysis of the contribution provided by trichotomized TK-1 to the NCCN-IPI are given below in Table 8.
TABLE-US-00009 TABLE 8 Multivariate Cox proportional hazard model including NCCN-IPI's risk components binarized according to the thresholds in Table 2 plus TK-1 (trichotomized). The hazard ratios are given along with their respective 95% confidence interval and the p-values Hazard ratio 95%-CI Covariate (HR) HR p-value NCCN-40-60 (bin) 1.899 0.923-3.908 0.082 NCCN.Age60-75 (bin) 1.504 0.731-3.095 0.268 NCCN.Age > 75 (bin) 2.975 1.261-7.016 0.013 NCCN.LDH1-3 (bin) 0.851 0.542-1.338 0.485 NCCN.LDH > 3 (bin) 1.255 0.597-2.637 0.549 NCCN.AnnArborStage (bin) 1.614 1.001-2.602 0.050 NCCN.ExtranodalDisease (bin) 0.904 0.506-1.615 0.733 NCCN.ECOG (bin) 1.908 1.279-2.845 0.002 TK-1.intermed (bin) 2.148 1.193-3.868 0.011 TK-1.high (bin) 4.716 2.234-9.954 <0.001
[0221] The model combining the binary NCCN-IPI risk components with TK-1 (trichotomized) shows significant added prognostic value of TK-1 (intermed versus low: p-value=0.011, high versus low: p-value<0.001; see Table 8). The corresponding HRs are 2.15 and 4.72, respectively, and are also very similar to the R-IPI counterpart.
[0222] 7.3 TK-1 Improves the Prognostic Ability of R-IPI as Well as of NCCN-IPI
[0223] A widely used measure of a risk model's prognostic ability is the c-index, which can be regarded as the area under the ROC curve (AUC) analog in survival models. Table 9 shows the computed c-indices for the investigated prognostic indices (R-IPI and NCCN-IPI) along with the c-indices of the TK-1-extended models with and without excluding the respective binary LDH component. The inclusion of TK-1 improves these indices by 3.2%-4%, regardless of the presence of LDH (see also Table 9). This improvement can be considered clinically relevant and is also statistically significant in most cases, although the study was not powered to show a significant improvement of the c-index (i.e. the bootstrapping-based confidence intervals do not include the 0, with the exception of NCCN-IPI vs NCCN-IPI+TK-1(log 2) w/o LDH and KPI vs KPI+TK-1(log 2) w/o LDH (see Table 10)). This clearly indicates that TK-1 not only improves the investigated prognostic indices, but even has the potential to replace LDH in these indices.
TABLE-US-00010 TABLE 9 c-indices of multivariate prognostic indices with TK-1 (+/−LDH) Multivariate model c-index R-IPI 0.626 R-IPI + TK-1(log2) 0.661 R-IPI + TK-1(log2) w/o LDH 0.660 R-IPI + TK-1(trichotomized) 0.678 R-IPI + TK-1(trictiotomized) w/o LDH 0.674 NCCN-IPI 0.639 NCCN-IPI + TK-1(log2) 0.671 NCCN-IPI + TK-1(log2) w/o LDH 0.670 NCCN-IPI + TK-1(trichotomized) 0.686 NCCN-IPI + TK-1(trichotomized) w/o 0.686 LDH
TABLE-US-00011 TABLE 10 C-index differences between the reirence models and the models including TK-1 (+/−LDH) Index comparison c-index diff 95%-CI R-IPI vs R-IPI + TK-1(log2) 0.034 0.007-0.073 R-IPI vs R-IPI + TK-1(log2) w/o LDH 0.034 0.003-0.070 R-IPI vs R-IPI + TK-1(trichotomized) 0.052 0.020-0.096 R-IPI vs R-IPI + TK-1(trichotomized) w/o 0.048 0.014-0.091 LDH NCCN-IPI vs NCCN-IPI + TK-1(log2) 0.032 0.002-0.066 NCCN-IPI vs NCCN-IPI + TK-1(log2) w/o 0.031 −0.008-0.065 LDH NCCN-IPI vs R-IPI + TK-1(trichotomized) 0.047 0.015-0.088 NCCN-IPI vs R-IPI + TK-1(trichotomized) 0.047 0.006-0.086 w/o LDH
[0224] The data given in Tables 9 and 10, respectively, clearly show that TK-1 can be used in combination with existing prognostic indices to considerably improve the clinical scores in predicting disease outcome for DLBCL patients. The inclusion of TK-1 did improve each of the prognostic index investigated.
[0225] As also obvious from the Tables 9 and 10, TK-1 has even the potential to replace LDH within these indexes, since for two prognostic indices (R-IPI and KPI) the c-values are even slightly better if LDH is left out and for NCCN-IPI only slightly worse if LDH is not included.
[0226] For a new patient diagnosed to suffer from DLBCL, especially, before treatment is initiated, TK-1 should be measured and compared to a pre-defined cutoff. If the measured value for the new patient is above the predefined cutoff, the patient would be considered to have a higher risk score and be more likely to experience unfavorable disease.