The nomogram has shown its superiority over the TNM system in many cancers. Based on the 8th TNM system, we developed and validated advanced incorporating the hematological biomarkers for both OS and PFS in NPC patients at diagnosis, it is able to generate individualized prognostication and it is proved to be more sufficient than the current 8th TNM system, yielding advantage in facilitating individualized risk stratification and decision making. Moreover, it is simple, low cost and easy to apply in clinical practice. We also included EBV DNA in the study and evaluated the role of it in different multivariable models, and to our knowledge, this is a first attempt to evaluate EBV DNA in different multivariable models in the IMRT era for the prognostication of NPC.
There were nomograms for NPC in the recent years, and they demonstrated consistent superiority over the TNM system. However, it has been a pity that most of them were based on the outdated 7th TNM system(9, 20), as this makes prognostication difficult under the condition that the clinician must refer back to the 7th edition. Based on the 8th edition, few nomograms were set up, unfortunately, there were significant drawbacks. One common drawback was categorizing all continuous variables. Since it causes information loss and reduces accuracy, it has been long regarded as problematic, highly inefficient and unnecessary, and was strongly advised against by specialists(12, 21, 22). In a nomogram suggested by Xu et.al(10), for example, the variables were all dichotomized, the uncertainty of best cut-off values and problem of overfitting should be under carefully evaluation. Another major drawback was the inappropriate consideration about potential variables. In the study of Pan et.al, they neglected the importance of EBV DNA and inflammation risk factors which were widely accepted as crucial predictors nowadays, and they failed to develop nomogram for PFS(23); Sun et.al built up the model with merely TNM system and EBV DNA(24), but as the present study indicated, this combination might not be effective enough, more variables should be contained.
The nomograms proposed in this study was based on the current 8th TNM system and they kept variables continuous as far as possible to avoid unnecessary information loss, moreover, they included crucial potential predictors which is widely accepted and routinely available in most of the hospitals. Regarding the special role of EBV DNA, they developed nomogram with or without it. Due to their simple yet cost-effective character, they provided more probabilities for a further large-scale validation and a widely acceptance and application, benefiting individualized risk stratification, clinical decision making and proper surveillance.
Numerous studies indicated that the hematological biomarkers are effective in predicting survival outcomes for NPC patients. NLR, PLR, LMR, as the representation of inflammation in vivo which plays important role in tumor development, they are closely linked to the prognosis(25, 26). Low level of Hgb implies poor cancer oxygenation status and are associated with the poorer survival(27, 28). SF and ALP is proved to be an adverse prognostic factor(29, 30), as a high Alb indicates enough nutrition in human body and is of positive prognostic importance(31). LDH were reported to be of great value in measuring the tumor burden in the NPC and were efficient in estimating outcomes(32). Baseline information in the present study (sex, age) and the TNM system (T category, N category) had been long used in predicting survival outcomes. Other potential biomarkers, such as epidermal growth factor receptor (EGFR), microRNAs, are expensive, technically difficult to some extent, and not routinely available in some hospitals, so they were ruled out in this study, as our aim is to develop nomograms that are simple, cost-effective and easy to apply.
Meanwhile, the special role of the plasma EBV DNA is noteworthy. On the one hand, the EBV infection is common causal agent of NPC and the EBV DNA are used to screening and diagnosis(5), in fact, it has become the first and the only nonanatomic factors to be included in the 8th edition of TNM system. There were also many studies indicating that adding EBV DNA to the TNM system(4), or even it alone(20) could result in more accurate prognostic prediction than the TNM system. But how important it is indeed and whether it is better than the combination of other prognostic biomarkers remained unknown. On the other hand, the test for EBV DNA has not been available in many hospitals, and in the hospital that could perform such test, interlaboratory variability occurs for lacking of globally methodology standardization(33). Taken these together, we developed nomograms with or without EBV DNA, the model of TNM system with EBV DNA, and compared them with each other and with the TNM system in order to evaluate the importance of EBV DNA and also to provide a more available nomogram without EBV DNA. The comparison indicated that all three models incorporating nonanatomic factors with the TNM system are more accurate and of better clinical usefulness than the TNM system alone, which is in accordance with the previous studies(20) in NPC in the past era of two-dimensional conformal radiotherapy. And the result also demonstrated that while the nomogram with EBV DNA rank first in the prognostication for patients as we supposed, the nomogram without EBV DNA outperformed the combination of TNM system and EBV DNA. This is meaningful considering the fact that some hospitals are not able to perform tests for EBV DNA and the methodology of testing it has not been globally standardized yet. The result indicated an outstanding advantage and a promising future for nomogram without EBV DNA. It also indicated that EBV DNA was a stable prognostic predictor for NPC, incorporating it with other variables to TNM system is of value, however it was not suggested to incorporate EBV DNA alone.
The nomograms confirmed their superiority in the validation cohort. A higher C-index and a good agreement in the calibration plot demonstrated their robustness and generalizability in application. To better facilitate clinical decision making and proper surveillance, we further stratified patients into different risk groups. The stratification should be valuable in initial decision making before individualization and in the design for future clinical trial.
Despite that the nomograms have many advantages over the TNM system and benefit the clinical practice, there were some limitations. First, as a retrospective study, there might be potential selection bias. Second, it was from a single center in high risk area. Third, the nomograms did not include C-reactive protein as a potential biomarker, however it could be argued that C-reactive protein is not routinely available in most of the hospitals now and is not always consistently associated with the survival. Four, validated in the temporal cohort indicated the robustness and generalizability of nomograms, inevitably, they needs more validation in other areas and at other time to further prove their transportability before widely application(12). The first step of validation is taken here and more will be carried out in the near future.