In recent years, survival in multiple myeloma (MM) has improved significantly. However, no significant decrease has occurred in early mortality. Infection is a common early complication of newly diagnosed multiple myeloma (NDMM) and a major cause of early death. One study showed that 45% of early deaths within 6 months were due to infection (Augustson et al., 2005). Some studies have shown that infection was listed as a contributing factor to death (McQuilten et al., 2022). Furthermore, even if the early infection is not fatal, it frequently causes significant delays and dose reductions in subsequent treatment, raising the risk of treatment failure (Zahid et al., 2019). Consequently, detecting early infection is critical to reducing early mortality.

The scoring system can help determine a patient’s risk of infection during MM treatment, allowing risk-adaptive strategies to be implemented to prevent early infection. However, an accurate infection prediction necessitates time and massive data. If the current infection prediction model can categorize infection in the real world, constructing a new prediction model seems unnecessary. At present, three infection risk prediction models have been proposed. In 2018, the FIRST score (Dumontet et al., 2018) was proposed first to stratify the risk of early grade ≥ 3 infections in patients with NDMM. This score is based on Eastern Cooperative Oncology Group (ECOG) performance status and serum β2-microglobulin, lactate dehydrogenase, and hemoglobin levels to define high-risk and low-risk groups showing significantly different rates of grade ≥ 3 infection (24.0% vs. 7.0%, respectively; p < 0.001) in the first 4 months. In 2022, the GEM-PETHEMA score (Encinas et al., 2022) comprising serum albumin, ECOG, male sex, and MM type was established to facilitate the identification of three risk groups with different probabilities of severe infection within the first 4 months, using the data from GEM2005 > 65, GEM2005 < 65, and GEM2012 < 65. The authors found that the infection rates of low-risk, intermediate-risk, and high-risk patients were 8.2, 19.2, and 28.3%, respectively. The two models showed good effects in the prediction of NDMM ≥3 infections. However, in addition to uniform clinical trials, a robust infection risk stratification system should perform well in the real-world setting. Since all patients were recruited as part of the clinical trial, they all met the inclusion and exclusion criteria. In the real world, more patients have poor physical fitness. In addition, in China, more patients receive a bortezomib-based regimen. However, unlike, GEM-PETHEMA score, participants in the FIRST score mostly accepted a non-bortezomib regimen. Consequently, determining whether the FIRST score and GEM-PETHEMA score can be used in the real world is essential.

IRMM is an infection prediction model established in China using real-world data (Shang et al., 2022). This model uses performance status, hemoglobin, β2-microglobulin, and globulin to categorize patients into high-risk, moderate-risk, and low-risk groups, which has shown significantly different rates of early infection in the three cohorts (46.5% vs. 22.1% vs. 8.8%; p < 0.001). The results indicated that the IRMM model could predict early grade ≥ 3 infection. However, it has not been validated in the real world by other centers’ data. Validating IRMM scores with data from other facilities would also be logical.

Therefore, the purpose of this study was to determine which of these three prediction models can clearly recognize grade ≥ 3 infection in NDMM patients in real-world clinical practice, enabling clinicians to determine appropriate infection prevention strategies.

In the present study, we first validated three grade ≥ 3 infection prediction models (FIRST score, GEM-PETHEMA score, and IRMM score) in a real-world population of patients with NDMM (mostly frail). According to our studies, the FIRST score and IRMM score could accurately predict grade ≥ 3 infection. However, the FIRST score, which comprises ECOG, β2-microglobulin, hemoglobin, and lactate dehydrogenase, is a simple and valuable infection classificatory system for patients with NDMM that could be employed in regular clinical use. The FIRST score is also suitable for frail patients and those taking bortezomib-based therapy.

Our results show that 66.01% (202/306) of NDMM patients reported infections within the first 4 months from the initiation of treatment, and 40.20% (123/306) ≥ 3 TE infections. The total number of any grade infections in this study was higher than the results of other series (37.3–61.8%; Rosiñol et al., 2019; Voorhees et al., 2020). It was also higher when compared with grade ≥ 3 infection in prior trials (11.9–20.37%; Dumontet et al., 2018; Encinas et al., 2022; Shang et al., 2022). Most of the participants in our research were frail patients, and we do not perform antibacterial prophylaxis for patients, so that may be the reason for the high rate of infection, especially for ≥ 3 TE infections. It makes sense to develop infection risk models. The most common pathogens in infections were Gram-negative bacteria, as in previous studies (Teh et al., 2017). As shown in previous similar studies, the most common site of infection was the respiratory tract (Teh et al., 2015; Zahid et al., 2019). The mortality associated with infection in the first 4 months was low (2.61%), which is consistent with the results of other series (1.34–5.86%; Augustson et al., 2005; Jung et al., 2016; Huang et al., 2017; Terebelo et al., 2017).

Besides the uniform clinical trials, a robust risk stratification system should work effectively in a wide range of clinical circumstances. This study demonstrates that the FIRST score can predict grade ≥ 3 early infection in the real world and shows reasonable distinction. The infection rate of the high-risk group was significantly higher than that of the low-risk group when divided by FIRST score (50.00% vs. 25.81%, p < 0.001). The GEM-PETHEMA score performed poorly in infection prediction and group discrimination. Based on real-world data, statistical differences existed in the incidence of infection between the high-risk group and the low-risk group (60.00% vs. 35.93%, p = 0.013) by GEM-PETHEMA score, but with no statistical difference between the moderate-risk group and the low-risk group (41.28% vs. 35.93%, p = 0.370) or between the high-risk group and the moderate-risk group (60.00% vs. 41.28%, p = 0.095). Furthermore, the cumulative incidence of infection did not differ statistically between the three GEM-PETHEMA score groups (p = 0.090). This indicates that the GEM-PETHEMA score does not appear to have effective discrimination for grade ≥ 3 infection in the real world. When the GEM-PETHEMA score research group tested the FIRST score using their data, they found no significant association with a higher risk of early severe infection between the high-risk and low-risk groups (p = 0.347; Encinas et al., 2022). However, our study results show that the FIRST score had a relatively high predictive value compared to the GEM-PETHEMA score. Differences in the patients included may also be the reason. The FIRST score’s patients are comprised of transplant-ineligible candidates. In the GEM-PETHEMA score, patients who also includes transplant-eligible candidates. In contrast, the patients in our study were similar to the patients in the FIRST score study, and the majority of them were transplant-ineligible candidates. However, the exact reason remains unclear, and this difference of results should be answered in future studies. The IRMM score is based on real-world data. The findings of this study indicate that the IRMM score had an elevated predictive value. When we inspected the high-risk, moderate-risk, and low-risk groups by IRMM score, we found significant differences between the high-risk and low-risk groups (p < 0.001) and between the moderate-risk and low-risk groups (p < 0.001). However, the differentiation between the high-risk and moderate-risk groups was unclear (p = 0.437).

This study attempted to further analyze these three models as to which may have a better ability to distinguish grade ≥ 3 infection in NDMM patients using ROC curves and the Youden index. However, from the validation results, the three prediction models were not much different. The probability of grade ≥ 3 infection in different infection groups was compared (p < 0.05), as well as the sensitivity and specificity of the three models (p > 0.05). However, the FIRST score model was only slightly better (AUC: 0.650 vs. 0.619 vs. 0.630). In addition, the cumulative infection rate by the FIRST and IRMM scores in the first 4 months was significantly statistically different, whereas the GEM-PETHEMA score showed no significant difference. Additionally, considering the prediction indicators and calculation process, the calculation of the IRMM score is relatively complex, whereas the calculation of FIRST and GEM-PETHEMA scores is relatively simple. Combining validated results and computational processes, we believe that the FIRST score will be more beneficial in clinical practice.

However, notably, induction regimens for developing the FIRST score did include non-proteasome-based regimens. In the real world, especially in China, most patients are treated with bortezomib-based regimens (Chinese Hematology Association, 2020; Zhou et al., 2022). Our results show that the FIRST score can be well-verified in real-world data. In addition, in the real world, the proportion of frail NDMM patients who are not transplanted is high (Antoine-Pepeljugoski and Braunstein, 2019; Goel et al., 2022; Medhekar et al., 2022; Yao et al., 2022). This study further explores the application value in frail patients, and the results show that in frail patients, the FIRST score can still carry out stratification.

Monitoring high-risk patients and taking active relevant preventive measures may be significant. The indications for antimicrobial prophylaxis are controversial. A randomized phase III URCC/ECOG study showed that the use of prophylactic antibiotics during the first 2 months of treatment did not reduce the incidence of severe infection (Vesole et al., 2012). In contrast, recent TEAMM results have shown that prophylactic use of levofloxacin in active myeloma treatment during the first 12 weeks of therapy significantly reduced febrile episodes and deaths compared to placebo (Kim et al., 2022). Nevertheless, prophylactic antibiotics are not recommended for all patients (Girmenia et al., 2019). The IMWG infection consensus guidelines and recommendation suggests that antimicrobial prophylaxis should be targeted at people at high or moderate risk of MM infection (Raje et al., 2022). Following the findings of this study, the FIRST score can better monitor high-risk patients. The FIRST score application facilitates infection prevention or adjustment of the intensity of induction therapy to reduce the incidence of infection.

Our study has several limitations. First, despite multicenter data collection, the sample size was relatively small. The nature of retrospective design narrows the conclusion, so biases were unpreventable. Second, the regimen containing daratumumab has been recommended for the first-line treatment of NDMM (Lonial et al., 2016). However, because this drug is not yet included in medical insurance reimbursement in China, few patients can use the regimen containing daratumumab for first-line treatment. Studies have reported that the incidence of infection is relatively high in the treatment of monoclonal antibody–based regimens, and these three models must be further verified in the regimen containing daratumumab. Third, based on the background of the COVID-19 pandemic, COVID-19 infection in NDMM, infection rates, and risk factors need to be investigated further, which can be done in follow-up studies.

In conclusion, our study confirms that the FIRST score is a simple and robust infection stratification tool for patients with NDMM, including frail patients and those being treated with bortezomib-based therapy. It can discriminate patients with grade ≥ 3 infection, help to guide infection prevention, and improve outcomes of patients with NDMM.

The original contributions presented in the study are included in the article/Supplementary material, further inquiries can be directed to the corresponding authors.

The studies involving human participants were reviewed and approved by Shanxi Bethune hospital ethics committee. Written informed consent for participation was waived for this study.

WT and JW designed and supervised the work, interpreted the results, and approved the final version. XL, WL, LZ, LY, QY, JZ, SH, and XC collected clinical data. XL and XC analyzed data. XL, WL, WT, and JW wrote and revised the manuscript. All authors contributed to the article and approved the submitted version.

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.