In this specific example, Messori and colleagues used the middle point of the enrollment interval as a representative point in time for the whole group. By using this approach, it is easy to fall into ecological fallacy, in which any inference about the individual is estimated based on the result of the group but might not represent the actual individual. As an example, a study enrolling patients between the years 1990 and 1999 will be assigned as year 1995 without having into account the potential differences in therapy that took place during that interval as it is likely that practices changed between 1990 and 1999.Additionally, meta-regression analyses do not address other weaknesses of categorical meta-analyses such as missing data and lack of adjustment for other potential important factors. These two aspects, not surprisingly, go hand in hand. For example, it is likely that the addition of appropriate antibiotic prophylaxis, growth factor support and use of intrathecal chemotherapy to decrease central nervous system involvement, in addition to rituximab, have partially improved the outcome; however, this will be unlikely to be evaluated without patient-level data. A recent study presented in abstract format at the 2012 American Society of Hematology Annual Meeting attempts to address this issue by analyzing patient-level data on 1,546 patients with HIV-associated NHL from 19 studies [3]. Preliminary results show an improvement on response and survival rates in patients treated with rituximab and chemotherapy. A final peer-reviewed publication is eagerly expected. Reassessing an old claim: Natural selection of hemizygotes and heterozygotes for G6PD deficiency in Africa by resistance to severe malaria reporting that a different mutation, G6PD 968 T ! C, had a frequency of 10% in Senegalese Serer males, and a G6PD deficiency prevalence of 12% in that ethnic group [3]; remarkably, the G6PD 968 T ! C mutation had been described in the literature since 1989 [4]. The Senegalese data reported by De Araujo et al. prompted a replication study and eventually G6PD 968 T ! C was proven to be the most frequent A2 mutation in The Gambia, present in 7% of males [5]. Therefore, since G6PD 968 T ! C was overlooked by Ruwende et al., a large proportion of G6PD A2-deficient subjects were certainly misclassified as normal in their study, as pointed out by L. Luzzatto [6]. 2. The G6PD 202 G ! A mutation in The Gambia had already been reported at surprisingly low frequency level (less than 2%) in 2004 [7]. The real frequency of this mutation in Gambian males was subsequently confirmed to be 2-3% [5,8], not 5.9% as reported by Ruwende et al. The reason why they found such an inflated frequency is unknown. 3. The frequency of G6PD A2 deficiency in Gambian males is approximately 10%, not 5.9% as reported by Ruwende et al.; the overall frequency of about 10% is accounted for by the sum of 968 T ! C (7%) plus 202 G ! A (3%). A third, non A2 deleterious mutation, 542 A2>T (G6PD Santamaria) has a frequency of over 2% in males (5), bri...