Improving accuracy of prognosis in patients with myelodysplastic syndromes using self-reported quality of life data. Opportunities for a new research agenda in developing prognostic models

KEYWORDS:

• Myelodysplasia
• quality of life
• patient-reported outcomes
• survival
• prognosis

Myelodysplastic syndromes (MDS) encompass a group of myeloid stem cell clonal disorders characterized by a defect in hematopoietic stem cell maturation, resulting in peripheral cytopenias as well as by a wide variation in illness trajectory. For example, there is large variability in the risk of progression to acute myeloid leukemia (AML) [1]. Possible treatment options might range from supportive care alone to possible disease-modifying treatments, such as hypomethylating agents or stem cell transplantation (in few selected cases) [1]. The management of MDS is challenging for several reasons, including the fact that, in many cases, patients are elderly at the time of diagnosis with potential-associated comorbid conditions [2]. In addition, there is evidence showing that physical, functional, emotional, and social well-being of MDS patients can be affected by disease- and treatment-related factors thereby substantially impacting quality of life (QoL) [3–5]. In these patients, QoL may be compromised for several reasons, including severe anemia, the frequent occurrence of infections, and the need of blood transfusions [6,7].

In health-care research and practice, QoL is typically collected through the use of standardized patient’s self-reported questionnaires and it is basically the only way to obtain the unique patients’ view on the impact of the disease and therapy on his/her own life. Routine assessment of QoL in MDS patients is important and has been advocated in international consensus guidelines [8]. Owing to the variability of the disease course, accurate identification of life expectancy and likelihood of progression to AML has historically been a critical issue for clinical decision-making. Therefore, given the vital role of outcome prediction, major efforts have been put by the scientific community in this area and various prognostic models have been developed [9]. Over the years, a number of single prognostic factors have been identified and prognostic models have been developed by combining these factors in different ways, with the ultimate goal of better discriminating between different risk group categories [9]. The main prognostic index, and probably the most commonly used in clinical practice, was developed by Greenberg and colleagues in 1997 [10], that is, the International Prognostic Scoring System (IPSS). This scoring system is based on percentage of bone marrow (BM) blasts, number of peripheral, and cytogenetic abnormalities and, based on these factors, patients can be classified into four distinct groups with variable risk of progression to AML and life expectancy. Later on, further research efforts have led to the improvement of this prognostic index, mainly by re-classifying cytogenetic abnormalities into more refined groupings and identifying different thresholds of cytopenias [11]. Still in a remarkable endeavor to better classify these patients and offer them personalized treatments, other prognostic models were also developed. The WHO-based Prognostic Scoring System (WPSS) [12] or the MD Anderson Comprehensive Scoring System (MDA-CSS) [13] are excellent examples of this effort to keep improving our ability to predict outcomes in this population. Interestingly, some other new prognostic variables are considered in these two latter indices compared to the IPSS classification. For example, the MDA-CSS includes performance status (physician-reported), age, platelets, hemoglobin, white blood cell counts, BM blasts, cytogenetic abnormalities, and possible prior transfusions. The prognostic value of comorbidities was also shown in the context of MDS. In particular, in a large retrospective cohort study [14], a prognostic model incorporating baseline comorbidities with age and IPSS was developed to predict survival. In this study, it was found, for example, that patients with severe comorbidities had a 50% decrease in survival, regardless of age and IPSS risk group [14].

However, while all these prognostic indices have greatly contributed in improving patient care, it is noteworthy that they do not include patient’s self-reported health status information, so as typically obtained by the use of QoL questionnaires. Indeed, routinely used prognostic models developed so far have been exclusively based on clinical, laboratory, and/or physician reported data. The big question is whether we can go further than this and whether patients themselves can also contribute in providing meaningful prognostic information. Of course, one might argue that hemoglobin, platelets, BM blasts, or other types of traditional clinical/laboratory information are ‘objective’ measurements, therefore the only ones we should keep considering in our continued effort to improve outcome prediction. Interestingly, whether patient’s self-reported information, obtained through the use of simple questionnaires, can significantly contribute to the science of prognostication has been recently well clarified in a seminal study by Ganna and Ingelsson published in the Lancet [15]. This was a large population-based study involving about 500,000 persons included in the UK Biobank project [16] with the aim of investigating predictors of mortality. Remarkably, out of all the numerous potential predictors investigated, they found that self-reported health and self-reported walking pace were the strongest predictors in both sexes and across different causes of deaths. The authors concluded that measures that can simply be obtained by questionnaires were the strongest predictors of all-cause mortality. The extent to which these findings are applicable to the MDS population has been investigated in a study recently published in the Lancet Oncology by our research group [17]. In this observational study enrolling higher-risk patients (IPSS intermediate-2 and high risk), the prognostic value for survival of patient’s self-reported fatigue, obtained by using a well-validated QoL questionnaire (EORTC QLQ-C30) [18], was examined. It was found that pretreatment self-reported fatigue independently predicted overall survival beyond well-established prognostic indices (i.e. IPSS, IPSS-R, and WPSS). Patients reporting higher baseline fatigue severity had a median survival of 14 months (95% CI, 11–17) compared to patients reporting lower fatigue severity who had a survival of 19 months (95% CI, 17–26) (p = 0·0012). This finding was also independent of therapy received after baseline assessment and of progression to AML. Also, when analyzing the prognostic power of the considered prognostic indices used alone, or in conjunction with self-reported fatigue severity, a statistically significant increase in the likelihood ratio test (p < 0·0001) was found when self-reported fatigue was added to each prognostic index (i.e. IPSS, IPSS-R, and WPSS). The prognostic value of patient’s self-reported QoL data found in MDS patients in our study [17] is not a standalone finding in oncology. Rather, it is fully in line with the already consolidated evidence found in a number of other cancer populations [19]. Several studies have in fact shown that self-reported QoL frequently replaces important clinical data, such as performance status or tumor size, in final multivariate prognostic models [20,21].

Overall, these findings now challenge the MDS community to also consider patients’ self-reported information when devising and/or refining prognostic models, and several research questions will have to be addressed in the near future. For example, it will be important to also investigate whether patient’s self-reported QoL information can predict not only overall survival, but also progression to AML. Also, it would be interesting to examine the prognostic value of QoL in patients with lower risk disease.

We advocate this is an important area of research as, eventually using prognostic indices that are also constructed based on patients’ self-reported information will greatly contribute moving toward a more personalized treatment approach. Putting patients at the heart of health-care research [22] is a critical issue in medicine in general, and the MDS community has now opportunities to make further progress in this direction.

Declaration of interest

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.