LAUSR

We have read carefully the systematic review conducted by Wei et al. [1], that estimated the magnitude of association between low calf circumference (CC) and mortality. In this letter, we highlight some concerns about its methodology. First, the search strategy adopted by authors was not sufficiently sensitive since it failed to identify potentially eligible studies. At least three more studies fulfill the inclusion criteria of their systematic review: (a) the study conducted by Santos et al. associated low CC (≤ 33 cm for females and ≤ 34 cm for males) with in-hospital death and mortality six months post-discharge in a hospitalized sample of 600 adults and older patients [2], (b) the study conducted by Mello et al. involving 548 older persons also associated low CC (< 31 cm) with death [3], c) the study conducted by Yin L et al., including 12,138 patients with cancer at five institutions in China, defined the optimal thresholds of CC and demonstrated a higher risk of death in those with low CC [4]. Second, to perform the Newcastle–Ottawa Scale (NCO) [5], it is necessary to clarify the definitions adopted in each domain, but it is unclear in that systematic review. According to the NCO manual, the authors should decide an acceptable length of time before the quality assessment begins. The results presented in Table 2 suggest that it was not standardized since reference 19 received 1 point (the study had 12 months of follow-up) while the references 18 and 26 received 0 points (studies had three years of followup). Also, reference 33 received 0 points in the item “adequacy of follow up cohorts” (this item aims to ensure that losses are not related to either the exposure or the outcome). The study had only six losses (4% of the total sample) due to participants being discharged home in the first six months of institutionalization. Third, in the Statistical Analysis Section, it is described that the authors extracted the hazard ratio (HR) from each included study or calculated it based on raw event count data. However, it is not clear how they calculated the HR since the time for event occurrence in each participant is usually unavailable in primary studies. In addition, the authors described in the Results Section that four studies provided the odds ratio (OR) for the association between low CC and mortality. However, they pooled these four studies with the others that provided the HR measure, with one study (reference 31) that did not provide any association measure (the authors did not report if they calculated it and how), and with other two studies that reported the relative risk (references 19 and 24). This approach for pooling the data is not appropriate as outlined in Cochrane Handbook. The most appropriate statistical plan would be to perform two separate meta-analyses, one for studies that provided HR and another for studies with OR. Furthermore, the authors described that all results of subgroup analysis pointed to a significant association between low CC and death without showing the P for the difference between subgroups. It is a mistake to compare within-subgroup inferences such as P values, so authors can perform a simple significance test to investigate differences between two or more subgroups, as described in Cochrane Handbook [6]. Another problem was found in data extraction since the authors included the references 23, 26 and 27 in the subgroup of hospitalized patients; This comment refers to the article available online at https:// doi. org/ 10. 1007/ s4199902100603-3.