As shown in forest plot of final weigh, the overall ESEFW under random effect model showed no significant difference from 0 (P > 0.05). Studies in Subgroup Ragworm were all performed with the same species (S. solea), so the conclusion of poor performance of formulated feed in comparison with ragworm (ESEFW=-13.87, P < 0.05, ESEFCR=3.66, P < 0.05) could only be reached with this individual species. In the study in Subgroup NF > FF, the ration of formulated feed was less than that of natural feed in wet weight because of the relatively poor acceptability of the subject (hybrid mandarin fish, Siniperca spp.) to formulated feed, and thus, negative ESEFW turned out (ESEFW=-2.17, P < 0.01). However, it could be seen through meta regression that PLN had a significant effect on ESEFW (ßPL(N)=-0.32, P < 0.05), and the intercept was significantly higher than 0 (P < 0.05). This indicated that fish fed on formulated feed had significantly lower FW than the counterpart fed on natural feed when the effect of PLN was eliminated.
The overall ESEFCR had no significant difference than 0 (P > 0.05), but as shown in forest plot of FCR, contradictory results were seen among subgroups defined by ration control. In Subgroup Equal, FCR of natural feed was significantly higher than that of formulated feed. This result was very likely to be accounted for by overfeeding of fish in natural feed group, which tended to result in elevated FCR (Andrews 1979; Wootton et al., 1988; Marimuthu et al., 2011). Overfeeding was shown by the fact that in the assays, the rations of formulated feed were 3%~8% of fish biomass, and in order to reach equal dry mass, 7 ~ 12 times as much Chironomidae spp. larvae had to be fed (Kamiński et al., 2016, 2010), while feeding to satiation twice daily needed only about 4 times as much Chironomidae spp. larvae as formulated feed (Meyer H. A. et al., 2016). Furthermore, it was noted that fish displayed obvious symptoms of overfeeding when 20.7% of fish biomass Chironomidae spp. larvae in wet weight (that is, 3.9% in dry weight) was fed (Kamler et al., 2006). Although it was also reasonable to suspect that the result was accounted for by the natural food itself instead of overfeeding as the same natural feed (Chironomidae spp. larvae) was used in this subgroup, this possibility, to a large extend, seemed to be ruled out by the findings of high digestibility (De La Noüe et al., 1985) as well as high conversion efficiency of Chironomidae spp. larvae (fed ad libitum) (Meyer et al., 2016). As for Subgroup Isocaloric, in one assay, the natural feed ration was twice as less as the formulated feed ration (Klett et al., 2022); in the other, the rations of the two types of feed were equivalent in dry mass because of the approximately equal energy levels but were both at relatively low levels (about 1.7% of fish biomass on dry matter basis; calculated according to the average fish biomass and daily food intake presented by the author) (Michael et al., 2010). We hold the opinion that the data from Subgroup Satiation reflected the comprehensive effect of nutrient compositions, palatability and sensory attractivity of the feed, and thus, from these data (ESEFCR=2.18, 95%CI [0.78, 3.59]) conclusion should be drawn, which was that formulated feed had significantly higher FCR than natural feed.
Results from subgroups defined by natural feed and from meta regression also supported this conclusion. In subgroups consisted of only one assay (k = 1), the influence of other factors such as species, ration control method could not be eliminated; while in Subgroup Chironimidae larvae, although more than one assay was included (k = 4), as mentioned above, their methods of ration control were identical. Therefore, only the combined ESE of Subgroup Trash fish (k = 18) was credible which indicated that fish fed formulated feed had significantly higher FCR than those fed on trash fish (ESEFCR=2.84, P < 0.05), which is, in a sense, in consistent with the meta-analysis conclusion that ESEFCR was positively correlated with fishmeal replacement level (Galkanda-Arachchige et al, 2019). As shown in Table 2., effect of trophic level was detected with intercept significantly higher than 0 (ß0 was 25.01 with 95% CI [0.62, 49.40] when data from Subgroup Equal and Subgroup Isocaloric were left out, and ß0 was 70.55 with 95% CI [16.90, 124.21] when conduced within Subgroup Trash fish). These results further confirmed the conclusion that formulated feed had significantly higher FCR than natural feed.
As for HSI, under random effect model, the overall ESEHSI had no significant different from 0 (ESEHSI=0.61, P > 0.05), the same results were also seen in subgroups defined by natural feed. In Subgroup Equal, fish fed natural feed had significantly higher HSI than the counterpart fed formulated feed (ESEHSI=-0.34, P < 0.01); when isocaloric ration was fed, opposite result turned out (ESEHSI=1.51, P < 0.01). These results, on the one hand, could also be explained by overfeeding; and on the other hand, might be accounted for by other factors such as species because only one study was included in each of them. However, leaving out the studies of feeding isocaloric ration and equal dry mass, 5 of the remaining 8 species had positive ESEHSI (P < 0.05). Significant effect of experimental duration was detected when it was combined with PL or LL as moderators (Table 2.). Individual study (not included in this meta-analysis) showed that short term treatment may not evoke the response of this indicator (Kim et al., 2002); at the same time, long-term feeding trial suggested that biochemical and histological alterations in liver may not be reflected by HSI, and formulated feed as well as natural feed resulted in liver alteration (Bolla et al., 2001). In recent years, serious hepatomegaly was seen at the end of the growing-out stage (over one-year old) in our pilot-scale culture of American shad (Alosea sapidissima). Therefore, negative effect of formulated feed on liver may exist in many species and long-term studies using histological and biochemical indicators are needed to evaluate the impact of formulated feed on fish liver.
Risk rate (RR) was the recommended effect size measure in evaluation of SR (Cooper et al., 2009). However, in this meta-analysis, most of the original studies only provided the means and SDs (or SEMs) of the replicates, so we could not know accurately the number of deaths which is essential for RR calculation. Therefore, Hedges’g calculated with mean and SD of SR was used. Previous meta-analysis showed that freshwater fish larvae fed on formulated feeds have a 2.5 times higher chance to die than those fed on live feed (Sales 2011). In this meta-analysis, we excluded the original studies which performed with fish larvae as well as those focused on weaning regime. Overall ESESR, as well as ESESR of subgroups defined by natural feed and ration control, had no significant difference from 0 (P > 0.05). This indicated that after weaning, fish fed on formulated feed survive as well as the counterpart fed on natural feed.
In meta-analyses in aquaculture, species often plays an important role in heterogeneity due to its rich diversity (Galkanda-Arachchige et al., 2019). The same situation was also seen in this meta-analysis. However, as shown in Table 3., heterogeneity among studies performed with the same species was high too (In fact, the assays in Subgroup Epinephelus fuscoguttatus were the farm level replicates of a single study (Bunlipatanon et al., 2012), and thus, they were all the same in feed, ration control, initial body weight and stocking density); therefore, there must be some unconsidered factors responsible for the heterogeneity. We suspected that they were related to water quality and the health of the subjects; but further analysis was limited by the absence of related data in the original studies. It was suggested that fish size has significant effect on energy requirement (Ende et al., 2017). In this study, initial body weight was used as moderator in regression analysis, but no effect was detected. The reason might be that the differences in initial body weight do not well describe the differences in growth stages due to the diversity in maturity sizes and growth characteristics of the species.
Meta-analysis on fish protein requirement showed that there is a positive correlation between trophic level and dietary protein requirement; and fish of trophic level 3.0 ~ 3.9 requires 43.8% dietary protein and fish of trophic level 4.0 ~ 4.7 requires 46.4% dietary protein (Teles ea al., 2019). However, as shown in Table 1., in studies included in this meta-analysis, formulated feed with 49.08 ± 9.94% (34.07%~75.66%) protein was provided to fish of trophic level 2.75 ~ 4.45, and no significant correlation was detected between trophic level and PLF (R2 = 0.36, P > 0.05). It could be seen from that that, in these studies, feeds were not formulated or selected in strict accordance with the protein requirements of the subjects, and high protein feeds tended to be used. Natural feed had even higher protein level, but did not result in a higher FCR value. As for LL, there was no significant difference between natural feed and formulated feed. Therefore, it was not protein or lipid level of formulated feed that account for the performance gaps between the two types of feeds. Reasons may lie in the differences between natural and formulated feeds in the profiles of amino acids and fatty acids, palatability and nutrient utilization (Klett et al., 2022; Chen et al., 2021; Ende et al., 2017). Further studies are needed to find out the deficiency of formulated feed and then fill the gaps.
In addition, this meta-analysis focused on the performance of formulated feed in terms of final weight, feed conversion rate and survival of fish; while the impacts of formulated feed on fillet quality and reproduction performance have also drawn attention and many studies on this issue have been conducted. However, there has not been meta-analysis conclusion on it.