gms | German Medical Science

Background: A prognostic impact of the value of serum free light chains (sFLC) at diagnosis and its corresponding ratio of kappa and lambda chains (rFLC) in patients with multiple myeloma (MM) is variously discussed in existing literature [1], [2], [3], [4] with main focus on the prediction of progression free and overall survival. However, rFLC-values change during the therapy of MM and a time-dependent modeling can lead to improved prediction. Furthermore, interactions with other prognostic factors such as treatment response, immune reconstitution and cytogenic high-risk factors prove to be a challenge for statistical modelling and the transfer of results into clinical practice.

The aim of this study is to provide a comprehensive statistical model for a time-to-event scenario within a complex setting of time-dependent and fixed covariates, which can be easily interpreted and transferred to clinical practice.

Methods: The continuous sFLC values were measured at successive treatment phases. Subsequently, the achievement of rFLC-normalisation [5] was recorded at the end of each treatment phase.

A relevant phase of MM therapy is the maintenance therapy. In a first approach, it was considered whether or not normalisation was achieved until the start of maintenance therapy at the latest. Therefore, the achievement of normalisation is of interest, regardless of whether the status of normalisation was lost again. In this approach, therefore, the time from randomisation to the first achievement of rFLC normalisation was considered.

A second approach models the course of rFLC normalisation over successive treatment phases. Thus, patient-specific follow-up times are divided into start and stop intervals. Further, response to treatment at specific phases as well as fixed baseline variables have been included in the multivariate model. Both approaches can be treated as multi-state models consisting of several time-dependent and fixed covariates.

Results: To analyse the prognostic impact of rFLC-normalisation, we considered the time to progression (PFS), defined as time from randomization to progression or death, whichever occured first. This yields the application of a multivariable time-dependent Cox regression model to analyse PFS [6], [7], dividing the individual follow-up time with respect to changes within the time-dependent covariates. This has been done for both approaches mentioned. We show, that the results of the first approach are consistent with those of the second approach, namely, a highly significant benefit for patients having achieved rFLC-normalisation until the start of maintenance at the latest. This is a surprising finding, since one may assume, that the consolidation of information within the first approach may weaken the prognostic impact caused by rFLC-normalisation over time.

Conclusion: We show the prognostic value of rFLC normalisation for MM patients using a complex time-dependent model approach, as so far has not been presented in existing literature.

Thereby we can show, that a consolidation of time-dependent information as performed by the first approach simplifies the transfer to the clinical practise without weakening the validness and significance. This enables the clinicians to make a patient-individual risk-classification with respect to the maintenance phase given the information wether a patient has achieved rFLC-normalisation until the start of maintenance therapy.

The authors declare that they have no competing interests.

The authors declare that an ethics committee vote is not required.