Gene expression profiling impacts treatment decision making in newly diagnosed multiple myeloma patients in the prospective PROMMIS trial

Abstract Multiple myeloma (MM) is a heterogeneous hematologic malignancy associated with several risk factors including genetic aberrations which impact disease response and survival. Thorough risk classification is essential to select the best clinical strategy to optimize outcomes. The SKY92 molecular signature classifies patients as standard‐ or high‐risk for progression. The PRospective Observational Multiple Myeloma Impact Study (PROMMIS; NCT02911571) measures impact of SKY92 on risk classification and treatment plan. Newly diagnosed MM patients had bone marrow aspirates analyzed for SKY92. Physicians completed a questionnaire for each patient capturing risk classification, hypothetical treatment plan, and physician confidence in the treatment plan, before and after unblinding SKY92. One hundred forty seven MM patients were enrolled. Before unblinding SKY92, physicians regarded 74 (50%) patients as clinical standard‐risk. After unblinding SKY92, 16 patients were re‐assigned as high‐risk by the physician, and for 15 of them treatment strategy was impacted, resulting in an escalated treatment plan. For the 73 (50%) clinical high‐risk patients, SKY92 indicated 46 patients to be standard‐risk; for 31 of these patients the treatment strategy was impacted consistent with a de‐escalation of risk. Overall, SKY92 impacted treatment decisions in 37% of patients (p < 0.001). For clinical decision‐making, physicians incorporated SKY92, and the final assigned clinical risk was in line with SKY92 for 89% of patients. Furthermore, SKY92 significantly increased the confidence of the physicians’ treatment decisions (p < 0.001). This study shows potential added value of SKY92 in MM for treatment decision making.


INTRODUCTION
Multiple myeloma (MM) is a hematological cancer, characterized by accumulation of clonal plasma cells in the bone marrow, leading to impairment of hematopoiesis and the production of monoclonal immunoglobulin. It accounts for 1.8% of all cancers with an estimated 32, 270 newly diagnosed cases and 12, 830 deaths for 2020 in the US, and in Europe >48,000 new cases and 31,000 deaths [1,2]. MM is a very complex heterogenous disease that changes genetically at each relapse in an individual patient. The median overall survival has improved toward 4-10 years [3].
To combat the heterogeneity of the disease, a rapidly increasing number of new therapies have been introduced to the clinical landscape [4]. These drugs are used in single-agent, doublet, triplet, and quadruplet regimens, and physicians often incorporate autologous or allogeneic stem cell transplant in eligible patients [5]. With the development of novel agents and various combinations, treatment decisions have become increasingly complex.
In order to navigate this complex clinical landscape, guidelines such as the International Myeloma Working Group (IMWG) and Stratification for Myeloma and Risk-Adopted Therapy (mSMART) recognize patient risk classification as an important tool [5,6]. Risk classification provides information based on disease biology to report on prognosis that is often used to guide therapeutic decisions. Some risk classifications are based solely on clinical parameters (e.g., International Staging System) [7], while others include cytogenetic aberrations (Revised-International Staging System [R-ISS], mSMART) [6,8]. The National Comprehensive Cancer Network (NCCN) panel agrees that gene expression profiling (GEP) is "a useful tool, that may be helpful to estimate aggressiveness of the disease, helping to make rational therapeutic decisions and individualize treatment" [9]. Although the recently published European Hematology Association-European Society of Medical Oncology (EHA-ESMO) clinical practice guidelines remain more conservative and state that "no prognostic factor or staging system, including R-ISS or GEP, is used routinely to define risk-adapted strategy" [10].
The SKY92 algorithm is a prognostic, 92-gene expression signature, also known as EMC-92. It provides a binary classification, standard risk or high-risk for disease relapse, [11] and improves the prediction of prognosis for MM patients in clinical practice [12].
SKY92 has been analytically and clinically validated and demonstrated both prognostic accuracy for overall and progression free survival in newly diagnosed and relapsed setting as well as independence of other risk stratification markers in multivariate analyses (Table 1) [ [11][12][13][14][15][16][17][18][19][20][21][22][23]. This prospective clinical utility study assessed the impact of SKY92 for newly diagnosed MM on risk stratification and treatment decisions and their confidence in the recommended treatment plan.

Study design
This study is an observational, prospective, multi-center study to assess the impact of SKY92 (SkylineDx, the Netherlands) test results on physician decision making regarding risk classification and treatment plan in newly diagnosed MM patients combined with the physician's confidence in the chosen treatment plan (study design: Figure 1).

SKY92
Fresh bone marrow aspirates were collected in heparin or ethylenediaminetetraacetic acid (EDTA) containing tubes and processed using a Ficoll density gradient, followed by immunomagnetic separation of CD138 positive plasma cells. After RNA extraction, quantity, purity, and integrity were measured. cDNA was prepared, and fragmented cRNA was combined with hybridization reagents to produce hybridization cocktails. These cocktails were hybridized to a SKY92 microarray (U133 Plus 2.0 GeneChip, Thermo Fisher) and scanned on a microarray platform, GCS3000Dx2. If all 10 quality control acceptance criteria were met, the SKY92 score was calculated [11].

End points
The primary end point was the percentage of patients for whom SKY92 led to an alteration in the treatment plan (defined as the hypothetical treatment if the physician could use the SKY92 result). A secondary end point was the physician's confidence in their treatment decisions.

Statistical analysis
We analyzed the proportion of patients whose treatment plan changed after unblinding SKY92, using a two-sided Exact Binomial test. The principal investigators were aligned considering 15% to be the accept-

Patients
Two-hundred fifty patients signed informed consent between February 2018 and April 2020 in nine participating institutions. One hundred three patients were screen failures because of no IMWG "active" MM diagnosis (n = 32), bone marrow sample quality not sufficient for SKY92 analysis (n = 54) and other reasons, for instance bone marrow sample could not be collected or patients withdrew consent (n = 17).
One hundred forty-seven patients were enrolled, and 30 physicians

Physician's change in risk classification
Prior to unblinding the SKY92 result, physicians assessed the patient's risk classification according to their own clinical routinely used methods. On the basis of that assessment, the physician is asked to summarize the patient's risk classification into either standard or high-

Physician's change in hypothetical treatment plan
The physician's change in risk classification after unblinding the SKY92 result resulted in a change in the proposed hypothetical treatment plan in several patient cases. For 31 of 46 (67%) patients that were regarded clinically high-risk by the physician prior to unblinding SKY92, the treatment plan was impacted, consistent with a downgrading of risk, and resulted in a de-escalated treatment plan when SKY92 reported a standard-risk result (Figure 3). Furthermore, for 15 of 16 (94%) patients assessed as clinically standard-risk by the physician prior to unblinding SKY92, the treatment plan was impacted resulting in an escalated treatment plan when SKY92 reported a high-risk result.
Finally, there were eight clinically high-risk patients prior to unblinding SKY92, where the physician still escalated the treatment plan because SKY92 was also high-risk. For 19 concordant standard-risk patients and five concordant high-risk patients, the physician indicated the SKY92 test to be helpful because it confirmed their treatment plan. In summary, treatment plan decisions were impacted by SKY92 in 37% (54 of 147) patient cases, which is significantly F I G U R E 4 Heatmap of cytogenetic abnormalities and risk classification. Figure depicting comparison of cytogenetic abnormalities and risk classification for each of the n = 147 patients included in this analysis. Each column represents a patient, each row represents a cytogenetic abnormality (either present or not present), and risk classification system. For each patient, risk according to IMWG and R-ISS is depicted. The seventh row is the physician assessed risk prior to unblinding SKY92 ("prior risk phys"); the eighth row is the risk as determined by SKY92. The last row is the physician assessed risk after unblinding SKY92 ("post risk phys"). The n = 46 patients for which physicians changed their risk assignment are highlighted by the arrow
Overall, utilizing SKY92 led to physicians having significantly more confidence in their treatment plan (p < 0.001).

DISCUSSION
The definition for high-risk MM is recognized to be constantly evolving with advances in diagnostics and therapeutics [24]. Many factors are being used to determine whether a patient should be regarded highrisk for disease progression and hence be treated more rigorously. The current study is not intended to show SKY92 improves outcome, since this has already been validated in previous studies [11][12][13][14][15][16][17][18][19][20][21][22][23]. The study is aimed to assess whether SKY92 could provide additional prognostic guidance for physicians. The results from this first prospective multicenter study show the potential utility of the robust gene expressionbased risk classifier SKY92 in providing additional guidance for risk assessment.
Unblinding SKY92 results led to a renewed clinical risk assessment by the investigators and an optimized, adapted treatment plan in 37% of patients. In 46 MM patients, initially deemed to be high-risk by the physician, treatment was de-escalated in 30 cases (65%) after patients were reclassified to standard risk by the physician based on the SKY92 result. Conversely, there were 16 (22%) patients who were initially classified as standard risk but were reclassified to high-risk based on the SKY92 high-risk result, and in 15 cases (94%), treatment was intensified. The final clinical risk assessment was in line with SKY92 in 89% of patients. Furthermore, knowing the SKY92 result led to an increased confidence of the physician in their proposed treatment plan. This increased confidence is of paramount importance to the complex heterogeneous clinical practice of MM in which physicians are confronted with a large, evolving body of data, and its significance in how to treat a patient.
Overall, the PROMMIS study population seems to be fairly representative as far as risk classification when compared with the pooled dataset used for R-ISS development, [8] where 62% of patients were in the intermediate-risk group (52% in PROMMIS), 28% were in the lowrisk group (30% in PROMMIS), and 10% were in the high-risk group (8% in PROMMIS). The higher number of high-risk patients by means of SKY92 (29%) compared to clinical risk classification guidelines such as R-ISS, is a known phenomenon, [16,23] suggesting that a substantial number of high-risk patients cannot be diagnosed by current clinical methods alone.
With 29% of patients SKY92 high-risk in this cohort (compared to 15-25% of high-risk patients in historical, newly diagnosed cohorts [11][12][13][14][15][16][17][18][19][20][21][22][23]), it seems reasonable to surmise that physicians selected higher risk patients to be included in the PROMMIS study. This notion is reflected by the observation that 50% of patients were thought to have high-risk myeloma by their physicians based on their routine clinical practice before unblinding of the SKY92 result, this may include but is not restricted to standard risk-stratification methods like R-ISS or cytogenetic aberrations such as amplification or duplication of chromosome 1q or other translocations/copy-number abnormalities that have adverse prognostic implications [25,26]. Other parameters could have been derived from modern imaging techniques such as MRI and CT-PET scanning; [27][28][29] patient-specific high-risk features such as old age, poor performance status, and comorbidities; or clinical features such as primary plasma cell leukemia and extramedullary disease [30]. It is interesting to note that abnormalities of 1q occurred in 70% of SKY92 high-risk patients compared to 33% in SKY92 standard-risk patients. The SKY92 signature is enriched for genes located on 1q [11], and similar findings have been found for other risk scores based on GEP such as UAMS70 [31]. Despite some overlap in 1q gene enrichment, SKY92 seems to identify a larger group of patients as high-risk (15%-25%) compared with the UAMS70 risk score, which identifies 12% of patients as high-risk [31]. Two papers [12,23] are available with comparative data for both risk scores: in a pooled dataset both GEP profiles were analyzed and showed that UAMS70 identified 9% of patients as high-risk whereas SKY92 identified 18% of patients as high-risk [12]. In the Myeloma XI trial, a comparative analysis for quantitative risk scores score was performed showing a significant correlation (79%) between the two profiles. SKY92 high-risk patients had significantly shorter PFS and OS compared with their SKY92 standard-risk counterparts. Similar performance of prognostication was shown for UAMS70 on OS, but not PFS [23]. Other techniques for molecular profiling of MM such as next generation sequencing, DNAseq or RNAseq, have not been widely adopted in MM (yet) and are currently mostly used for detection of measurable residual disease [32].
One limitation in this study is the screen failure rate of 103 patients. There was improved physician confidence in treatment decisions after SKY92 results in a disease where selecting therapy is of paramount importance.

ACKNOWLEDGEMENTS
We are grateful to all patients, to all the investigators, and study personnel involved in patient recruitment, data management, and SKY92 sample processing. The PROMMIS study (ClinicalTrials.gov Identifier: NCT02911571) was sponsored by SkylineDx. Snoo analyzed and interpreted the data and wrote the manuscript. All authors participated in data interpretation and critical appraisal of the manuscript and approved the final manuscript.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.