Before discussing further our results we believe that three possible questions, that may arise reading this paper, need to be clarified. Firstly, the rational to divide our patients into hematological and non-hematological group. Would it not have been more appropriate to classify them into those with tumor disease (hematological and solid tumors) and those with chronic benign state (chronic heart failure)? Would it not have been more appropriate to compare patients in the three separate groups i.e. hematological, solid tumors and heart failure group? There are several explanations for our decision. Our institution is a referral center for hematological diseases and most of the patients admitted to our institution indeed suffer from hematological diseases. Fewer patients are admitted for solid tumors, heart failure and other diseases related to cachexia. By dividing our patients in hematological and non-hematological ones we have achieved optimal balance of consecutive patients in respective groups. However, the answer is not just of technical or statistical matter. If one would be really interested, and it was indeed our intention, to investigate differences between hematological and patients with other diseases related to cachexia how is one supposed to divide patients? Dividing patients into three groups, apart from creating numerically unbalanced groups would not answer this question. In fact, in order to get the correct answer to this question a comparison between hematological patients and all other groups of patients with diseases that can cause cachexia would be necessary. Unfortunately, such a study could not have been conducted in our institution although we eagerly await a study by others that would address this question.
Secondly, we decided to divide patients into cachectic and non-cachectic just according to the main criterion that is loss of at least 5% of body weight during the previous 12 months or less, not further looking at other criteria proposed by new definition that we refer to . We think that such, at first glance, arbitrary division is founded on the majority of the existing experimental data that defined cachexia just as s state with a certain amount of weight loss during variable time frame. Although we do not question the new definition we still think it can be subjected to various tests. One of them could be determining relations between its major and minor criteria. Comparing results of minor criteria upon establishing groups of patients after using major criteria is just that. In accordance with our work also goes the new proposed definition for cancer cachexia which states that only the loss of weight of 5% for 6 months is a criterion enough to diagnose cancer cachexia .
And lastly we decided, as is obvious from our patient characteristics, to enroll patients that had weight loss for more than 12 months. We think that it is justifiable, although the definition that we refer to  states the weight loss should be in 12 months or less. But the authors also state that “… .time frame may be disease specific and is likely to be shorter in cancer (3–6 months) and longer in chronic heart failure and kidney disease (12 months)… . .”. Although this sentence may imply the upper limit of 12 months we believe that it also may imply periods longer than 12 months. Some of the published data refers to periods exceeding 12 months, as was the case in one of the pivotal studies from Anker et al. with duration of weight loss at inclusion up to 13 months. After inclusion patients were followed for a mean of 686 days which is almost two years . In that study, patients that exceeded 12 months of body weight loss after inclusion were not excluded from follow up and we have followed the same reasoning in our investigation. It is especially important in patients with durable survival in spite of their disease (hematological patients with lymphomas, some cancer patients, heart failure patients) so periods of loss of body weight can be longer.
Given the scarcity of data regarding clinical and laboratory features comparing patients with various diseases that are related to cachexia we conducted our study to address some of these issues. We were specifically focused on differences between hematological and non-hematological patients. We found, as others did, that cachectic patients as a group, had significantly higher levels of CRP and lower levels of both albumin and hemoglobin. Secondly, we divided patients into hematological and non-hematological groups and found that hematological patients had significantly lower levels of CRP and higher levels of albumin. There was no difference in levels of hemoglobin. When we further subdivided patients into cachectic and non-cachectic groups we found that the differences observed for CRP were mainly due to differences found in non-cachectic group. Observed differences in albumin levels were present regardless of this subdivision. Hematological cachectic patients had lower levels of hemoglobin than non-cachectic ones. The same was not true for non-hematological patients. Observed differences could imply possible divergences in pathophysiology between the groups of patients as we divided them. Unfortunately, we were not able to conduct investigations regarding possible differences in for example expression of inflammatory cytokines that could explain observed differences. Because of that our question “is there only one cachexia?”, although provocative, remains purely speculative and awaits further trials with larger numbers of patients and more balanced groups exploring other features of cachexia, probably primarily those incorporated into proposed definition . New proposed definition orientating just on cancer cachexia is in the line with our work i.e. it already recognizes the existence of “different” types of cachexia .
We then evaluated proposed cut-offs for CRP (≥ 5mg/L), hemoglobin (≤ 12 g/dl) and albumin (≤ 3.2 g/dl) and found that CRP had highest sensitivity, followed by hemoglobin and albumin. Inverse relationship was found for specificity. Cut-offs were found to have relatively good negative predictive value whereas positive predictive value was poor. Combining all three cut-offs just emphasized above mentioned relations. When we divided patients to the hematological and non-hematological groups, proposed cut-offs followed same trends as in the whole group. Proposed cut-off for CRP was not able to distinct cachectic from non-cachectic patients in the whole group nor was it able to distinct them when we further subdivided them into hematological and non-hematological groups. On the other hand, proposed cut-off for albumin was able to distinct patients according to their nutritional status regardless of further subdivisions. Hemoglobin cut-off showed its predictive power in both whole group as well as in hematological group but not in non-hematological group. Although such results would imply that CRP cut-off is least effective and albumin cut-off is the most effective one such conclusions would need larger studies powered enough even to give more precise values. Until then, our results serve as ones that arouse suspicion in proposed cut-offs.