Donor Age and Non-Relapse Mortality: Study of Their Association after HLA-Matched Allogeneic Hematopoietic Cell Transplantation for Acute Myeloid Leukemia and Myelodysplastic Syndrome

Round 1

Reviewer 1 Report

The impact of donor age remains an important question. It seems that younger donors can improve transplant outcomes from a number of imperfect studies. In this study, the single centre nature allows for some homogeneity in transplantation aspects but is limited by modest numbers. The strength of the observations is limited by small numbers. It highlights an important issue, however, and is reflective of current practice and is relevant to practice.

Minor comments:

1) I would suggest a "softer" title and conclusion given that the HR for DA in association with NRM is only 1.04. The same is true for the association of DA with cGVHD (HR 1.03). Just another event here or there could alter the statistical significance of this result. How about "Assessing the impact of donor age on NRM in .....".

2) In the Discuission under "Limitations", I think the authors need to acknowledge that the older donors were predominantly related (as opposed to unrelated). They mention that related donors did not receive ATG. Is it possible that the omission of ATG is association with increased cGVHD, rather than donor age? It is unclear whether ATG usage was included in the model. At a minimum, this should be mentioned as a possibility. 

3) Inclusion of variables in multivariate analysis should include some assurance that the variables or more or less evenly distributed within the cohort. If there is a high association with other variables, even MVA cannot overcome this intrinsic bias. The authors should provide some comment on this aspect. Even if they continue to present the results of MVA, it would help the reader understand in statistical terms whether these results have some inherent bias that cannot be totally overcome. A comment in the limitations on this issue seems warranted.

Author Response

We thank you for your very relevant comments. Here are our answers:

1)The HRs of 1.04 and 1.03, respectively for the association of donor age (DA) with NRM and cGVHD are for DA taken as a continuous variable, which means that for each year of age, the HR increases by 0.04 for NRM and 0.03 for cGVHD. With DA dichotomised at 50 years, the HR 3.35. We have inserted a brief precision in the Results to specify DA as continuous variable.

2) This is a crucial point indeed. The HR (0.70) and p-value (0.41) for the association of ATG with cGVHD moderate-severe are reported below in a model with DA (continuous) and aGVHD (the 2 variables independently associated with NRM):

Hazard ratio Lower 95%CI Upper 95%CI     p.value

agedon        1.0230      0.9996       1.048 0.054000000

agvhd24       6.2110      3.1900      12.090 0.000000078

atglap        0.7043      0.3049       1.627 0.410000000

All we can conclude is that there is probably an association of ATG (HR) with cGVHD, but the sample size may be to low (therefore the large confidence interval and high p-value). We have added a mention of this non-significance in the Results section and have included this notion in the Discussion.

3) We agree with this comment. We have included this limitation in the Discussion, we acknowledge that our cohort, even treated in a single centre using uniform protocols suffers from heterogeneity because of changes in practice over time (penultimate paragraph of Discussion).

Reviewer 2 Report

It's a very important clinical question, and the study methodology is sound.  The statistical analysis appears well done.  The paper itself is well written.

It's surprising you found such a large difference in outcomes based on donor age, while other studies have shown a less certain relationship.  I'd encourage you to take this further with a larger national sample size.

A few comments:

1) Older recipients are more likely to have older donors - so donor age and recipient age are likely correlated.  Could you confirm this?  Did you look for interaction between these two variables in your multivariable analysis?

2) Do you have any data on cell dose?  If older donors mobilized less effectively that might be a possible mechanism.  Similarly, any other details on the graft (T cell dose, etc) might be informative.

3) It looks like you found no relationship between acute/chronic GVHD and NRM (line 158) - which is a bit surprising.  Any comment on why this might be the case.

Author Response

We thank you for your very relevant comments. Here are our answers:

1)This is indeed a very important point. We have added a limitation pointing to this issue in the penultimate paragraph of the Discussion. Interaction terms in a competing-risk regression are not as straightforward as in traditional Cox regression. Therefore, we have tested various models including and excluding the potentially collinear variables: age/donor (MUDs are younger) and ATG/donor (ATG was given mostly to MUD recipients, 49% in MUD transplants, only 10% in MSD). The limitation inherent to multivariable analyses with high collinearity among variables is now mentioned in the Discussion.

2) We have not found a significant difference between older and younger donors in terms of CD34+ cell mobilisation: (in million cells / recipient weight in kgs) 6.1 for donors 40 or older vs. 6.3 for <40, and 6.0 for donors 60 or older vs. 6.1 for donors <60. We therefore did not mention this point for the sake of space. Unfortunately, we have CD3+ data in only in small minority of transplants.

3) This is point deserving further explanation, and in order not to induce confusion we have removed the assertion from the Results section. In fact, the multivariable models for NRM were devised only with variable that are modifiable before transplant, which is the objective of our work: indirectly (because of retrospective nature) show that younger donors are better, GVH being non modifiable. But result from multivariable analyses (not shown in the paper) do show that time-depending aGVHD (and cGVHD) have an influence on NRM (the worst type of GVHD being refractory aGVHD). As we do not wish to overload the article with these lengthy methodological explanations, we prefer to simply remove the assertion from the Results section.