A meta-analysis of the effects and relationship between guanidinoacetic acid and the growth performance of broiler chickens

Highlights

• •A meta-analysis of guanidinoacetic acid (GAA) in broilers was conducted using 53 studies.
• •GAA supplementation improved average daily gain and feed conversion ratio in the standard diet.
• •GAA had a minimal effect on arginine-deficient, low-protein, and low-energy diets.
• •No difference between low vs. high inclusion of GAA on the production performance of broilers.
• •Predictive simulation using Ross 308 broilers resulted in 2.11% (60 grams) higher final BW with GAA inclusion.

Abstract

Regulating dietary Arginine (Arg) levels in the diet of broilers has been a focus of researchers due to the potential economic benefit to the industry. This meta-analysis evaluated the effects of dietary guanidinoacetic acid (GAA) supplementation on the production performance of broiler chickens fed diets that were either adequate or deficient in Arg or energy. Through studies, the basal diet generally met standard requirements for broilers, with deficiencies in Arg or energy being experientially induced. The meta-analysis using weighted random-effects models revealed that dietary GAA improved average daily gain (ADG; P < 0.001) by 2.17% and 1.08% during starter and finisher periods, respectively. Concomitantly, the feed conversion ratio (FCR; P < 0.001) also improved by 2.37% and 2.41% in the starter and finisher periods, respectively, with minimal evidence of publication bias for these key traits. However, significant heterogeneity (I² > 90%) indicated strong between-study variances. In broilers fed Arg-deficient diets, GAA supplementation only showed minimal effect to reduce the detrimental effect of an Arg-deficient diet (-4.75 vs. -3.64% ADG reduction), suggesting its limited Arg-sparing function. Under low energy diet, GAA restored ADG to levels equivalent to those of energy-sufficient diets. In the finisher phase, comparable but less pronounced effects were observed, suggesting reduced responsiveness with age. Dose-response evaluation suggested no difference in ADG or FCR between low and high GAA inclusion, suggesting low GAA level is more effective. Meta-regression models identified feed intake and arginine intake as the dominant predictors of ADG, and including GAA consistently improved model accuracy and increased ADG compared to the control diet. The meta-refgression suggested that 0.06% GAA supplementation increased final body weight by approximately 2.11% (60 g) in Ross 308 broilers. Collectively, the results confirm that GAA exerts a robust, dose-limited benefit of growth efficiency in broiler chickens, thereby supporting its incorporation into Arg or energy deficient diets that would have potential economic benefit.

Introduction

Modern broiler chicken has become the world's most efficient source of protein due to advances in genetics, nutrition, and environmental management strategies. However, volatility in feed ingredient prices and the inherent variability in the nutrient composition of alternative feedstuffs remain the main challenges to maintaining profitability, requiring precise formulation strategies. With the growing use of non-traditional protein sources and reduced crude protein (CP) diets to lower costs and environmental impacts, supplementation with crystalline amino acids (AA) has gained special interest to support optimal broiler growth (DeGroot et al., 2018). Guanidinoacetic acid (GAA), also known as N-(aminoimino-methyl)-glycine, serves as the immediate biochemical precursor of creatine and plays a central role in maintaining cellular energy homeostasis via the creatine–phosphocreatine system (Asiriwardhana and Bertolo, 2022). In poultry, GAA is synthesized endogenously from arginine (Arg) and glycine (Gly) through the enzymatic activity of L-Arg:Gly amidinotransferase, providing the substrate necessary for creatine formation and subsequent energy buffering in rapidly growing tissues of broilers (Portocarero and Braun, 2021). This biosynthetic reaction imposes a considerable demand on Arg, an essential AA in broilers, owing to its limited capacity for de novo synthesis. Consequently, dietary supplementation with GAA can reduce dietary Arg requirements for other critical physiological functions, increasing its availability for protein deposition, growth, and immune system support (Khajali et al., 2020; Portocarero and Braun, 2021).

Accumulating evidence indicates that dietary supplementation with GAA improves broiler growth performance, feed conversion efficiency (FCR), and carcass yield across a range of feeding conditions, including both nutrient-adequate and protein or energy-restricted (low metabolizable or ME) diets (Dilger et al., 2013; Mousavi et al., 2013; Sharma et al., 2022). Under Arg-deficient feeding conditions, GAA supplementation has been shown to restore growth performance to levels observed in Arg-adequate diets, highlighting its Arg-sparing capacity (Fosoul et al., 2019). Similarly, in broilers fed low CP (Sharma et al., 2022) or low ME diets (Abudabos et al., 2014), dietary GAA mitigated the reduction in body weight (BW) gain and feed efficiency, primarily by enhancing creatine-mediated energy metabolism and increasing intramuscular creatine stores. Beyond supporting growth performance, GAA supplementation has been associated with improved muscle energy homeostasis, enhanced antioxidant defenses, and greater resilience to thermal and hypoxic stress, underscoring its multifaceted, promising physiological roles in broilers (Majdeddin et al., 2020; Marques and Wyse, 2019). Despite extensive evidence supporting the benefits of GAA, the magnitude of responses has been inconsistent across studies. These variations might be influenced by factors such as experimental design (Li et al., 2023), supplementation levels (Ahmadipour et al., 2018a), dietary composition (Esser et al., 2018), and the genetic background of broiler strains (Córdova-Noboa et al., 2018a; Mohebbifar et al., 2022).

Meta-analysis can provide a robust approach for integrating heterogeneous experimental findings from multiple studies (Sauvant et al., 2008), enabling generalization through the systematic quantitative identification of critical dietary and physiological factors that modulate broiler growth responses to GAA supplementation. Furthermore, meta-regression enables the evaluation of variables that correspond to the changes in growth performance and estimates the growth of broilers fed GAA-supplemented diets, providing a foundation for developing growth prediction models relevant to modern broiler genotypes. GAA is a strong candidate since it is a direct creatine precursor. Thus, the hypothesis is that GAA supplementation will improve growth performance, especially when Arg-deficiency is most pronounced in Arg-limiting diets. Therefore, the objectives of this study were: (1) to provide a quantitative evaluation of GAA supplementation on broiler growth performance, namely ADG, FCR, and feed intake (FI), across a number of nutritional scenarios, including low CP, low ME and Arg-deficient diets; and (2) to conduct a meta-regression analysis to estimate any interactive effects of dietary GAA and other predictors to estimate interactive contribution of dietary GAA among other measures with respect to ADG variability.

Materials and methods

Articles search

A systematic search for peer-reviewed publications were conducted in August 2025 using Scopus (https://www.scopus.com), PubMed (http://www.ncbi.nlm.nih.gov/pubmed), and Web of Science (https://www.webofscience.com/) in August 2025. A manual search was also conducted across animal and veterinary-related peer-reviewed journals to identify suitable articles that might not have been indexed in the above databases (e.g., in press or accepted articles). The search strategy utilized “guanidinoacetic acid” and “broiler chicken” as the primary keywords to identify relevant publications. The search results from each database were were imported into Microsoft Excel to facilitate the screening and selection process.

Inclusion criteria and study selection

To qualify for the meta-analysis, this study met: (1) published in peer-reviewed journals indexed in Scopus, Web of Science, or PubMed; (2) consist of original research articles; (3) involve in vivo randomized control trials of fast growing broiler strains; (4) use GAA in dietary treatments with control groups; (5) present mean production data (average daily gain or ADG and feed intake) along with variance (SEM or SD); (6) detail replicates and animal specifics (strain, number, sex, age), experimental settings, diets, methods, and statistical analysis; and (7) confirm ethical approval for animal research (IACUC/ACUP). Fig. 1 illustrates the study selection process adhering to the PRISMA protocol (Page et al., 2021). Initially, all titles from searches were screened in Microsoft Excel to eliminate duplicates. Relevant titles were retained while non-relevant ones were removed. Full papers corresponding to the selected titles were then downloaded and imported into Mendeley reference manager for a thorough review. This process resulted in 47 studies selected for data extraction.

Fig. 1.

Forest plot of subgroup meta-analysis showing the effects of Guanidinoacetic acid (GAA) supplementation on average daily gain of broiler chickens under various dietary conditions during starter (day 1-21) and finisher (day 22-42) phase. The effect size is expressed at 95% confidence intervals (lower – upper) of weighted relative means difference (RMD) to the standard - control diet. The x-axis shows the RMD; the central-blue line represents the zero effect (RMD = 0); the black-diamonds represent represents the SMD of subgroup effect. Asterisk symbol is provided to represent the level of significance (* <0.05; ** <0.01; *** <0.001). arg_def = Arg-deficient diet, arg_def+gaa = arg_def + GAA supplementation, arg+ = basal diet supplemented with Arg, gaa+ = basal diet supplemented with GAA, low_cp = low protein diet, low_cp+arg = low_cp diet supplemented with Arg, me_def = metabolizable energy-deficient diet, me_def_gaa = me_def diet supplemented with GAA.

Database development

The data included detailed information from all eligible publications including general author information, publication year, country of publication, number of replicates (n) and birds per replicate, broiler strain, sex, and time of the trial. Data also included diet-related information such as GAA product information (manufacturer, inclusion level %), diet characteristics (crude protein [CP], metabolizable energy [ME], lysine, methionine, and arginine levels), and other treatments introduced in the studies. The mean and variance (standard deviation or standard error of the mean) of key outcomes such as average daily gain (ADG), feed intake, and feed conversion ratio (FCR) were calculated. ADG was derived from final body weight divided by the experimental period in days, while FCR was calculated as body weight divided by feed intake. The standard error was obtained from the standard deviation using the formula SE = SD/sqrt(n), where n represents the number of replicates. Web-Plot-Digitizer (https://apps.automeris.io/wpd/) was employed to extract graphical data.

Where relevant, each separate experiment within a single study was considered an independent data entry. For each response variable, an additional column was included to compute the normalized inverse of the standard error of the mean (SEM), serving as the weighting factor (WF). This WF served as the weight assigned to each response variable and was determined using the formula WF = W1/W2, where W1 represents the reciprocal of the experiment’s SEM (1/SEM), and W2 is the average of all W1 values across the dataset (Brisson et al., 2022). Several categorical variables were coded to support model evaluation, along with the continuous data. The feeding period was categorized into starter (1-21 days) and finisher (22 – 42 days). Since the experimental design of the studies involved different dietary conditions (i.e., deficient in CP, ME, and Arginine), their levels were expressed as percentages relative to the control-standard diet (standard diet is referred to as 100%) to minimize biases introduced by different dietary formulas across studies. The GAA levels were also categorized into low (≤0.06%) or high (≥0.06%). Groups were separated based on dietary treatments, including con = standard diet, Arg_def = arginine-deficient diet, Arg_def+gaa = Arg_def + GAA supplementation, Arg = basal diet supplemented with arginine, GAA = basal diet supplemented with GAA, Low_CP = low protein diet, Low_cp+Arg = low_cp diet supplemented with arginine, ME_def = metabolizable energy-deficient diet, and ME_def_GAA = ME_def diet supplemented with GAA. Table 1 summarizes the study characteristics included in this meta-analysis.

Table 1.

Summary of the references used for the meta-analysis of the effect of dietary GAA supplementation on the performance of broiler chickens.

Study	Authors	EP1 (d)	N/IT2	B/T3	N/RP4	Country	Strains	Sex	GAA5		Deficiency level
									Levels	Source	Arg (%)	ME (kcal)
1	(Verhelle and Saremi, 2025)	10	1200	240	16	Belgium	Ross 308	Male	0-0.12	CreAMINO (> 96%)	91	50-100
2	(Westreicher-Kristen et al., 2025)	10	504	168	6	Netherlands	Ross 308	Male	0-0.06	CreAMINO (> 96%)	93-97	-
3	(Barekatain et al., 2025)	11	640	80	8	Australia	Ross 308	Male	0-0.08	Alzchem Trostberg GmbH (99%)	93	-
4	(Li et al., 2024)	21	640	160	8	China	Cobb	Male	0-0.18	Tiancheng Pharmaceutical (99%)	-	-
5	(Al-Abdullatif et al., 2024)	14	120	20	4	Saudi Arabia	Ross 308	Mixed	0-0.12	CreAMINO (> 96%)	-	-
6	(Alaa et al., 2024)	10	364	70	7	Egypt	Arbor Acres	Male	0-0.06	CreAMINO (> 96%)	-	-
7	(Peng et al., 2023)	7	192	48	6	China	Arbor Acres	Male	0-0.06	Tiancheng Pharmaceutical (99%)	-	-
9	(Maynard et al., 2023)	14	750	250	10	Germany	Ross 708	Male	0-0.12	CreAMINO (> 96%)	-	-
10	(Salgado et al., 2023)	21	1280	160	8	Brazil	Cobb 500	Male	0-0.06	CreAMINO (> 96%)	-	150-225
11	(Sharma et al., 2022)	15	720	80	8	Australia	Ross 308	Male	0-0.192	CreAMINO (> 96%)	-	-
12	(Pirgozliev et al., 2022)	14	1280	320	16	Germany	Ross 308	Mixed	0-0.06	Evonik (99%)	-	50-100
13	(Mohebbifar et al., 2022)	10	640	160	8	Iran	Cobb	Male	0-0.18	CreAMINO (> 96%)	-	-
14	(de Souza et al., 2021)	10	600	200	8	Brazil	Cobb 500	Male	0-0.06	CreAMINO (> 96%)	-	-
15	(Ceylan et al., 2021)	10	11400	1900	10	Turkey	Ross 308	Male	0-0.06	CreAMINO (> 96%)	-	50-100
16	(Khalil et al., 2021b)	10	546	182	14	Thailand	Ross 308	Male	0-0.12	CreAMINO (> 96%)	-	-
17	(Boroumandnia et al., 2021)	9	144	24	4	Iran	Ross 308	Male	0-0.3	Evonik (99%)	-	-
18	(Khalil et al., 2021a)	11	192	48	8	Egypt	Ross 308	Mixed	0-0.06	CreAMINO (> 96%)	-	-
19	(Zhao et al., 2021)	18	192	48	6	China	Ross 308	Male	0-0.06	Junde Tongchuang Biology (98%)	-	-
20	(Majdeddin et al., 2020)	10	720	240	12	Belgium	Ross 308	Male	0-0.12	CreAMINO (> 96%)	-	-
21	(Boney et al., 2020)	21	1728	432	12	USA	Hubbard´Cobb	Mixed	0-0.06	CreAMINO (> 96%)	-	-
22	(Zarghi et al., 2020)	10	450	50	5	Iran	Ross 308	Male	0-0.12	CreAMINO (> 96%)	-	-
23	(Çenesiz et al., 2020)	17	792	128	8	Turkey	Ross 308	Male	0-0.06	CreAMINO (> 96%)	-	50-100
24	(Zhang et al., 2019)	15	320	160	8	China	Arbor Acres	Male	0-0.12	Tiancheng Pharmaceutical (99%)	-	-
25	(DeGroot et al., 2019)	14	360	72	12	USA	Ross 708	Male	0-0.18	CreAMINO (> 96%)	-	-
26	(Majdeddin et al., 2019)	10	540	90	9	Iran	Ross 308	Male	0-0.12	CreAMINO (> 96%)	-	-
27	(Faraji et al., 2019)	19	420	60	5	Iran	Cobb 500	Male	0-0.225	CreAMINO (> 96%)	-	-
28	(Degroot et al., 2018)	14	280	40	8	USA	Ross 708	Male	0-0.12	CreAMINO (> 96%)	-	-
29	(Córdova-Noboa et al., 2018a)	14	800	200	10	USA	Ross 708	Male	0-0.06	CreAMINO (> 96%)	93	-
30	(Córdova-Noboa et al., 2018b)	14	1280	320	16	USA	Ross 708	Male	0-0.06	CreAMINO (> 96%)	-	-
31	(Ahmadipour et al., 2018a)	n.a	300	60	4	Iran	Ross 308	Male	0-0.2	Evonik (99%)	-	-
32	(Esser et al., 2018)	21	1260	315	9	Brazil	Cobb	Male	0-0.08	CreAMINO (> 96%)	-	-
33	(Ahmadipour et al., 2018b)	21	300	60	4	Iran	Ross 308	Male	0-0.2	Evonik (99%)	-	-
34	(Ale Saheb Fosoul et al., 2018)	15	390	65	5	Iran	Ross 308	Male	0-0.12	CreAMINO (> 96%)	-	150
35	(Majdeddin et al., 2018)	10	540	60	6	Iran	Ross 308	Male	0-0.12	CreAMINO (> 96%)	-	-
36	(Tabatabaei Yazdi et al., 2017)	10	450	50	5	Iran	Ross 308	Male	0-0.12	CreAMINO (> 96%)	-	-
37	(Kodambashi Emami et al., 2017)	11	900	180	12	USA	Ross 308	Male	0-0.12	CreAMINO (> 96%)	-	-
38	(Tossenberger et al., 2016)	21	768	256	8	Germany	Ross 308	Male	0-0.6	CreAMINO (> 96%)	-	-
39	(Abudabos et al., 2014)	10	200	20	5	Saudi Arabia	Ross 308	Male	0-0.06	CreAMINO (> 96%)	-	25-75
40	(Mousavi et al., 2013)	10	1536	256	4	Iran	Cobb 500	Mixed	0-0.06	CreAMINO (> 96%)	-	159-318
41	(Michiels et al., 2012)	13	768	192	6	Belgium	Ross 308	Male	0-0.12	CreAMINO (> 96%)	-	-

²Duration of the experimental period

³Total number of birds

⁴Number of birds per treatment

⁵Number of replicate pen per treatment

⁶Guanidinoacetic acid; n.a. = not available

Publication bias assessment

Biases of publications were assessed using funnel plots (Fig. 2), Egger’s test and Begg’s test (Egger et al., 1997), with the resulting P-values reported for each outcome. Furthermore, a sensitivity analysis was performed to assess the robustness of the treatment effects by identifying influential studies, outliers, and sources of substantial heterogeneity. This was implemented using a leave-one-out approach (Viechtbauer, 2010).

Fig. 2.

Forest plot of subgroup meta-analysis showing the effects of Guanidinoacetic acid (GAA) supplementation on feed conversion ratio of broiler chickens under various dietary conditions during starter (day 1-21) and finisher (day 22-42) phase. The effect size is expressed at 95% confidence intervals (lower – upper) of weighted relative means difference (RMD) to the standard - control diet. The x-axis shows the RMD; the central-blue line represents the zero effect (RMD = 0); the black-diamonds represent represents the SMD of subgroup effect. Asterisk symbol is provided to represent the level of significance (* <0.05; ** <0.01; *** <0.001). arg_def = Arg-deficient diet, arg_def+gaa = arg_def + GAA supplementation, arg+ = basal diet supplemented with Arg, gaa+ = basal diet supplemented with GAA, low_cp = low protein diet, low_cp+arg = low_cp diet supplemented with Arg, me_def = metabolizable energy-deficient diet, me_def_gaa = me_def diet supplemented with GAA.

Meta-analysis and meta-regression

In this study, we performed conventional meta-analysis in the RStudio environment (RStudio version 2024.4.2 +764) using the “metafor” package (Schwarzer et al., 2015) and meta-regression modelling using “lme4” and “lmerTest” programs. Prior to formal analysis, both visual and statistical assessments were performed to detect potential outliers. Cook’s distance and studentized residuals were computed using the “influence()” function. Data points with Cook’s distance values greater than 4/N (where N represents the total number of observations) were considered potential outliers or influential observations (Naomi Altman and Martin Krzywinski, 2016). However, this approach led to excessive data removal from our dataset. Therefore, a data-driven approach, as proposed by Kebreab et al. (2025), was used: data point with Cook’s distance >3× the mean value or a studentized residual >3 was considered an outlier and were removed from the dataset.

For the meta-analysis, the clean dataset (after outlier removal) was used. To ensure consistency and comparability across studies, we used the weighted relative mean difference (WRMD) to estimate the effect size between treatment and control and expressed it as % difference. This approach guarantees a constant and unbiased interpretation regardless of the experimental unit. To account for variation arising from study-specific effects, the weighting factor mentioned above was used. This technique assigns greater weight to studies with lower variance (i.e., higher accuracy). The equations for calculating RMD and its variance are presented as follows:

$$\text{RMD}=\left(\frac{\overline{Y}T-\overline{Y}C}{\overline{Y}C}\right)\times 100$$

$$\text{Var}\left(\text{RMD}\right)\approx {100}^2\times \left[\frac{1}{{\left(\overline{Y}C\right)}^2}\times var\left(\overline{Y}T\right)+{\left(\frac{\overline{Y}T}{{\left(\overline{Y}C\right)}^2}\right)}^2\times var\left(\overline{Y}C\right)\right]$$

$$\text{and}\text{inverse}-\text{variance}\text{weighted}=\frac{\sum Yi\left(\frac{1}{S{E}_i^2}\right)}{\sum \left(\frac{1}{S{E}_i^2}\right)}$$

where $\overline{Y}T$ = mean of the treatment group, $\overline{Y}C$ = mean of the control or baseline, SE₁ = SE of the treatment or intervention, SE₂ = SE of the control group, $var\left(\overline{Y}T\right)$ = SEM² of the control or treatment groups (Schwarzer et al., 2015). RMD was reported as a 95% confidence interval and plotted as a forest plot. Between-study heterogeneity was assessed using Cochran’s Q and I2 statistics (Higgins et al., 2011), with values estimated using the Der Simonian–Laird (DL) estimator. The intercept-free structure of the mixed-effect meta-analysis was used for the model fitting. The subgroup meta-analysis was carried out to estimate covariate-specific effects using “rma.mv()” function, whereas the subgroup was declared as a moderator and study-specific effect was used as a random effect in the model [(i.e., mods = ∼ 0 + covariates + (1 | study_id)]. The I² statistic was reported to indicate the heterogeneity value, which was categorized as no heterogeneity (0% < I² ≤ 25%), low heterogeneity (25% < I² ≤ 50%), moderate heterogeneity (50% < I² ≤ 75%), or high heterogeneity (I² > 75%).

Meta-regression was performed to evaluate the relationship between dietary GAA and growth performance under different dietary scenarios. For this purpose, we initially used the full data set in the model fitting, with ADG as the dependent variable and GAA intake (g/d) as the independent variable. Multiple linear and nonlinear regression models were fitted using lme4 and lmerTest program, treating the above predictors as fixed effects and the experiment_id as a random effect. Prior to model fitting, multicollinearity was examined using the variance inflation factor (VIF), in which predictors with VIF>10 were not tested simultaneously (Dai et al., 2022). The meta-regression was conducted to follow the principle of a backward elimination procedure; we removed the predictor once at a time when it was not statistically significant between the full and reduced model (Irawan et al., 2023). The following full mathematical models were used:

Y_i(j)w_i(j) = β₀ + β₁x_i(j)1 + β₂x_i(j)2 + . . . + β_px_i(j)p + r_i(j) + s_j + e_i(j)

Where Y_i(j) is the predicted effect size for the ith experiment (i = 1 to 99) nested in the jth study (j = 1 to 39), w_i(j) is the weighting factor, β_p are fixed effects corresponding to the intercept and explanatory predictors x_i(j), r_i(j) is the random effect of experiment, s_j is the random effect of study, and e_i(j) is unexplained residual error. The assumption of the random effects of the experiment and study was that they were independent and normally distributed around 0 (N∼0). The weighing factor was calculated from the SE from each study as explained in a previous publication (Irawan et al., 2023, 2024).

Multiple interaction models were considered to account for the variability of CP, ME, and Arg in the diets, as well as the potential effects of some covariates (i.e., rearing phase, strains). However, the complex dietary scenarios or conditions across studies led to an imbalance in group representation, making the models less robust and less interpretable due to large study-specific variability. Therefore, we separated the full set and top-dressed data into two main subsets (starter and finisher phase). Since it is well known that ADG is linearly associated with the function of FI, we used FI as the main predictor in the model, with other predictors such as Arg intake (Argi, g/d), Arg levels in the diet (%), GAA levels in the diets (%), GAA intake (GAAi, g/d), ME relative percentage, and Arg/Lys ratio were initially included in the model fitting. The use of Arg/Lys was chosen because other essential AA were maintained at constant ratios relative to Lys. Consequently, variation among treatments occurred mainly in Arg supply rather than in the overall AA balance. For this reason, the Arg:Lys ratio represents the biologically most relevant predictor and was used in the analysis. The linear and quadratic models of the continuous predictors β_p (FI, GAAi, Argi) were tested independently and combined, in each “starter” and “finisher” dataset. The selected models were examined and compared for their model performance, precision, and accuracy. Model selection was based on P-value, AIC, and R² values. This was carried out using a revised k-fold cross-validation method with the caret package in R (Gareth et al., 2021), partitioning the dataset into training and testing sets. Subsequently, the mean square prediction error (MSPE), root mean square prediction error (RMSPE), ratio of RMSPE to observed SD value (RSR), and concordance correlation coefficient (CCC) were calculated as previously described (Respati et al., 2023). The CCC value was calculated in R using the epiR package and the epi.ccc function (Stevenson et al., 2021). A value of -1 indicates complete disagreement, and +1 indicates perfect agreement (Lin, 1989).

Results

Dataset characteristics and publication bias

The meta-analytic database comprised the treatment contrasts extracted from the studies listed in Table 1. Only studies used fast-growing and modern broiler strains (Ross, Cobb, Arbor Acres, and Hubbard) were retained for the analysis. Descriptive statistics for key covariates and outcomes are provided in Table 2, reported GAA inclusion covered 188 observations in the starter period and 196 observations in the finisher period. A total of 32 out of 41 studies (78%) used commercial GAA from CreAMINO^R (> 96%) while few studies used GAA from Evonik (3 studies), Tiancheng Pharmaceutical, and and Junde Tongchuang Biology (98%); all reported the GAA purity between >96%). Due to the skewed distribution of the data (ranged from 0% to 0.78%), the median GAA inclusion was 0.06% in the starter period and 0.05% in the finishing period. The average ADG was 38.23±18.24 g/d during the starter phase (n = 188) and 79.18±18.76 g/d during the finisher phase (n = 196). Concentrations of metabolizable energy (ME), CP, Lys, and Arg are summarized in Table 2. The nutritional profile of the diets across studies indicated adherence to the standard requirement of feed formulation for broiler chickens (Aviagen, 2022; NRC, 1994). For ranges of ME, CP, and AA were generally within the expected values. The minimum ME value (2775 kcal/kg) was lower, as studies was specifically designed to evaluate the low ME diet. Furthermore, because the reporting of these and other covariates was incomplete and heterogeneous across source studies, the analytic sample size varied by specific procedure (e.g., subgroup comparisons, meta-regression, and predictive modeling). Unless stated otherwise, the primary pooled estimates reported in Table 3 are based on the number of datasets (k) = 383 observations.

Table 2.

Descriptive statistics of the dataset.

Items	Starter					Finisher
	N	Min	Mean	Max	SD	N	Min	Mean	Max	SD
GAA %	188	0.00	0.06	0.45	0.11	196	0.00	0.05	0.61	0.08
AME, kcal/kg	188	2775	3002	3186	81.9	196	2800	3120	3310	81.23
CP, %	182	17.94	21.68	25.36	1.48	183	16.14	19.46	23.60	1.39
Lysine, %	187	0.80	1.29	1.50	0.15	196	0.88	1.14	1.49	0.11
Methionine, %	151	0.42	0.62	0.78	0.08	161	0.39	0.54	0.73	0.08
Arg, %	186	0.88	1.35	1.88	0.20	196	0.75	1.20	1.56	0.14
Arg/Lys	186	0.70	1.06	1.71	0.16	196	0.66	1.05	1.35	0.10
ADG, g/d	186	9.67	38.23	81.43	18.24	196	44.05	79.18	131.36	18.76
FI, g/d	186	18.22	53.61	109.00	26.66	196	71.71	139.68	334.40	52.52
FCR	186	1.01	1.37	1.93	0.20	196	1.21	1.67	2.72	0.29

N = number of data points, Min = minimum observed value, Max = maximum observed value, SD = standard of deviation, GAA = Guanidinoacetic acid, AME = apparent metabolizable energy (kcal/kg), CP = crude protein (%), ADG = average daily gain, FI = feed intake, FCR = feed conversion ratio

Table 3.

Summary of the effects of dietary Guanidinoacetic acid and covariates on production performance of broiler chickens.

Covariates	Response variable (QM P-value & I2 and QE)
	ADG	FCR	FI
k	383	299	299
Treatments	<0.001 (99.6%; <0.001)	<0.001 (99.7%; <0.001)	<0.001 (99.9%; <0.001)
GAA dose	0.609 (99.3%; <0.001)	0.021 (99.5%; <0.001)	0.236 (99.4%; <0.001)
Rearing phase	0.943 (99.3%; <0.001)	0.343 (99.5%; <0.001)	0.003 (99.4%; <0.001)
Strain	<0.001 (99.0%; <0.001)	<0.001 (98.9%; <0.001)	<0.001 (99.3%; <0.001)
Sex	0.821 (99.3%; <0.001)	0.784 (99.5%; <0.001)	0.219 (99.4%; <0.001)
Country	0.013 (99.1%; <0.001)	0.002 (99.4%; <0.001)	0.076 (99.3%; <0.001)
Begg’s Test	0.383	0.140	0.186
Egger’s Test	0.547	0.145	0.431

Notes: Random intercept per experiment with REML estimation; k = the number of data points. QM = test of moderators; QE = test of residual heterogeneity.

The assessment of potential publication bias indicated no significant bias for ADG and FCR, as indicated by both Begg’s and Egger’s tests (Table 3). This consistency across tests provides minimal evidence of small-study effects for these key performance indicators. Leave-one-out sensitivity analyses revealed that excluding any individual study did not meaningfully alter the direction, magnitude, or statistical significance of the pooled estimates for ADG and FCR.

Overall and subgroup effects

Table 3 summarizes the effects of GAA and other covariates on broiler production performance indicators. The results show that, overall, dietary GAA supplementation was highly significant for ADG (P < 0.001; I² = 99.6%), FCR (P < 0.001; I² = 99.7%), and FI (P < 0.001; I² = 99.9%). A random-effects model confirmed the significance of GAA (Table 3). Subsequent analysis identified strain and country as significant sources of this heterogeneity (P < 0.001), whereas the GAA doses, sex, and rearing phase had no significant effect on ADG, FCR, and FI; primarily due to relatively similar dose and source of GAA. Other moderating variables that were tested and showed non-significant result was the source of GAA. The deficiency levels of Arg used in the analysis was limited to the studies used 90-95% of Arg relative the control treatment because this ranges are practically realistic in the feed formulation. Consequently, studies with Arg deficiency lower than 90% or those supplied Arg up to 180% of the Arg recommendation were disregarded. In the dataset, ME deficiency was 90-95% compared to control basal diet.

The effects of GAA, Arg, and other dietary treatments are shown in the forest plot, each compared to the control-standard diet (Fig. 1). As expected, during the starter period, broilers fed diets deficient in Arg, CP, and ME showed significant (P<0.001) reductions in ADG, with the highest reduction being shown in the Arg-deficient diet (RMD% = -4.75%). Dietary Arg or GAA supplementation markedly enhanced growth performance when added as top-dressed (ADG +4.72% and +2.17%, respectively; P<0.001) but only showed to minimally helped reduce the detrimental effect of an Arg-deficient diet (-4.75 vs. -3.64% ADG reduction), confirming its limited Arg-sparing function. Under low-energy conditions, dietary inclusion of GAA effectively restored ADG to a level equivalent to that of a ME-sufficient diet (Fig. 1), indicating that its growth-promoting effect is maintained when dietary ME is limited. In the finisher phase, similar patterns were observed, but the effect of GAA on the ME-deficient diet was not observed. The Arg supplementation during finisher did not result in improved ADG (P=4237). Meanwhile, GAA supplementation in the finisher period exhibited growth promoting effect (+1.08%; P<0.001), although the magnitude was smaller than that in the starter period.

The response pattern for FCR (Fig. 2) mirrored that of ADG. During the starter period, GAA and Arg supplementation consistently reduced FCR by 2.37% and 3.01%, respectively. In the finisher period, the FCR was reduced by 2.41% with GAA but was not affected by Arg supplementation. Moreover, FCR was higher in broilers fed Arg (+1.89%), CP (+1.88%), and ME (2.99%) deficient diets, particularly during the starter period, with no significant effect was observed in the finisher period. FI was only marginally influenced by GAA inclusion (-0.54%; P=0.001 in starter and -1.33%; P=0.002 in finisher; Fig. 3). Arg supplementation had no effect on the FI during the starter and finisher period. Dose-specific analyses (Fig. 4) showed that both low and high inclusion levels of GAA (0.06–0.12 g/kg) produced comparable effects on improving ADG and FCR both starter and finisher periods, suggesting a plateau response beyond the physiological creatine requirement.

Fig. 3.

Forest plot of subgroup meta-analysis showing the effects of Guanidinoacetic acid (GAA) supplementation on feed intake of broiler chickens under various dietary conditions during starter (day 1-21) and finisher (day 22-42) phase. The effect size is expressed at 95% confidence intervals (lower – upper) of weighted relative means difference (RMD) to the standard - control diet. The x-axis shows the RMD; the central-blue line represents the zero effect (RMD = 0); the black-diamonds represent represents the SMD of subgroup effect. Asterisk symbol is provided to represent the level of significance (* <0.05; ** <0.01; *** <0.001). arg_def = Arg-deficient diet, arg_def+gaa = arg_def + GAA supplementation, arg+ = basal diet supplemented with Arg, gaa+ = basal diet supplemented with GAA, low_cp = low protein diet, low_cp+arg = low_cp diet supplemented with Arg, me_def = metabolizable energy-deficient diet, me_def_gaa = me_def diet supplemented with GAA.

Fig. 4.

Forest plot of subgroup meta-analysis (low vs. high dose) of Guanidinoacetic acid (GAA) supplementation effects on average daily gain (ADG), feed conversion ratio (FCR), and feed intake of broiler chickens under various dietary conditions during starter (day 1-21) and finisher (day 22-42) phase. The effect size is expressed at 95% confidence intervals (lower – upper) of weighted relative means difference (RMD) to the standard - control diet. The x-axis shows the RMD; the central-blue line represents the zero effect (RMD = 0); the blue-diamonds represent represents the SMD of subgroup effect. Asterisk symbol is provided to represent the level of significance (* <0.05; ** <0.01; *** <0.001).

Meta-regression and model performance

A total of 5 models were developed using top-dressed dataset for each rearing period (Table 4). The model structures were designed to evaluate FI vs. Arg intake as the main predictor, with each model also include GAA intake and Arg/Lys ratio. In the starter phase, the comprehensive model (Model 1; Table 4) incorporating FI, Arg, and GAA intakes, along with their quadratic and interaction terms, achieved the highest goodness-of-fit (R² = 0.935; RMSE = 5.07; CCC = 0.966; Fig. 5). Simplified models (e.g., Model 2 and Model 3) retained high explanatory power (R² = 0.909; CCC = 0.907), suggesting that Arg and FI were the dominant predictors of ADG. For the finisher phase, the best-fitting model (Model 1) achieved R² = 0.675 and CCC = 0.809, while other models yielded R² values ranging from 0.59 to 0.68 (Fig. 5). Although the predictive accuracy was lower than that of the starter phase, including GAA consistently improved model performance metrics, indicating its consistent association with growth efficiency. However, due to heterogeneous dietary conditions, the model fitted to the full dataset result in large RMSPE and RSR values. We therefore investigated whether using only the top-dressed dataset would improve model performance.

Table 4.

Model comparison and statistical performance of the selected models.

Model	Predicted equation	N	R2	RMSE	RMSPE	RSR	MB	SB	CCC
Fullset - Starter
Model 1	-10.57 + 0.064×fi + 0.005×fi2 + 6.89×arg – 2.77×Arg2 + 0.02×GAAi – 9.63×GAA% + 0.07×arg%+ 0.249×AME% - 23.29×Arg_lys	188	0.935	5.12	12.4	26.2	1.357	-1.39	0.966
Model 2	6.09 + 0.512×fi – 17.43×Argi + 0.001×GAAi - 7.26×Arg_lys	188	0.909	6.2	15	31.8	0.737	0.497	0.952
Model 3	-1.48 + 0.614×fi + 9.65×Argi	188	0.907	6.34	15.3	32.5	0.592	1.63	0.95
Model 4	-7.71 + 90.58×Argi - 22.78× Argi2- 0.013×GAAi + 2.96×GAA% + 0.013×Argi + 0.256×AME% - 29.4×Arg_lys	188	0.851	7.53	18.2	38.5	0.196	-13.4	0.921
Model 5	14.63 + 44.45×Argi - 0.006×GAAi + 0.075×Argi + 0.19×AME% - 32.1×Arg_lys	188	0.89	7.25	17.5	37.1	0.183	-26.9	0.913
Model 6	CON=5.23 + 74.24×Argi; GAA= 5.23+74.548×Argi	188	0.815	11.39	27.5	58.4	-1.99	-54.1	0.728
Fullset -Finisher
Model 1	-64.78 + 0.79×fi - 0.001×fi2 + 2.16×Arg% + 0.92× Arg2 + 0.009×GAAi - 8.14×GAA% + 0.22×Argi + 0.525×AME% - 16.21×Arg_lys	196	0.675	12.2	15.3	66.4	3.8	-7.78	0.806
Model 2	26.15 + 0.36×fi + 7.74×Argi + 0.005×GAAi - 6.021×Arg_lys	196	0.603	14.7	18.5	80	4.26	-4.01	0.749
Model 3	25.55 + 0.338×fi + 5.79×Argi	196	0.611	13.3	16.7	72.3	3.79	-12.9	0.768
Model 4	6.31 + 30.62×Argi - 1.81× Argi2+ 0.009×GAAi - 13.09×GAA% - 0.009×Argi + 0.59×AME% - 26.98×Arg_lys	196	0.618	11.5	14.5	62.6	2.08	-43.1	0.743
Model 5	15.32 + 24.47×Argi + 0.002×GAAi - 0.082×Argi + 0.606×AME% - 25.49×Arg_lys	196	0.59	11.9	14.9	64.6	2.19	-42	0.734
Model 6	CON=20.34 + 74.11×Argi; GAA=20.34+74.718×Argi	196	0.677	12.4	15.6	67.4	1.06	-62.1	0.624
Topdress - Starter
Model 1	27.12 + 0.32×fi + 0.002×fi2 + 0.012×GAAi - 14.62×Arg_lys	116	0.936	4.17	11.50	25.20	0.51	-6.40	0.967
Model 2	30.23 + 0.59×fi + 0.011×GAAi - 23.68×Arg_lys	116	0.921	4.62	12.70	27.90	0.17	-7.89	0.959
Model 3	42.91 + 29.37×Argi + 9.88× Argi2+ 0.018×GAAi - 32.79×Arg_lys	116	0.941	4.01	11.00	24.20	0.20	-7.95	0.969
Model 4	43.01 + 46.24×Argi + 0.016×GAAi - 38.53×Arg_lys	116	0.931	4.32	11.90	26.10	-0.08	-8.80	0.963
Model 5	CON=1.943 + 46.59×Argi; GAA= 1.943+47.708×Argi	116	0.899	5.23	14.40	31.60	0.16	-9.52	0.947
Topdress -Finisher
Model 1	54.59 - 0.21×fi + 0.003×fi2 + 0.002×GAAi + 3.82×Arg_lys	121	0.689	10.8	13.8	56.7	-1.6	-23.4	0.825
Model 2	21.52 + 0.45×fi + 0.0013×GAAi - 1.66×Arg_lys	121	0.666	11.1	14.2	58.2	-1.41	-28.6	0.807
Model 3	65.003 + 19.51×arg_int + 6.14×arg_int2 + 0.004×GAAi - 32.37×Arg_lys	121	0.669	11.1	14.1	57.8	-1.6	-29.1	0.808
Model 4	53.1 + 39.75×arg_int + 0.004×GAAi - 35.6×Arg_lys	121	0.671	11.0	14	57.5	-1.48	-29.5	0.809
Model 5	CON=15.57 + 39.013×Argi; GAA=15.57+39.967×Argi	121	0.657	11.3	14.4	58.9	-1.31	-29.2	0.802

N = number of observations (data points), R² = regression coefficient between observed vs. predicted values, RMSE = root mean square error; RMSPE = root mean square prediction error; MB = mean bias; SB = slope bias; RSR = RMSPE-observations SD ratio; CCC = concordance correlation coefficient. In the equation, fi = feed intake, Arg = % Arg in the diet, Argi = Arg intake (g/d), GAAi = GAA intake (mg/d), GAA% = % GAA content in the diet, Arg% indicates relative changes of Arg levels due to experimental design in each study (Arg in the control – standard diet is referred as 100%), AME% indicates relative changes of AME levels due to experimental design in each study (AME in the control – standard diet is referred as 100%), Arg_lys = ratio of Arg to lysine in the diet. The equations for each phase and subset were selected by considering quadratic and linear terms, feed intake vs. Arg intake comparison as predictors, and the interaction with treatment groups [with or without GAA supplementation] as categorical predictor.

Fig. 5.

Plot of the observed vs. predicted values for the selected models (Table 4). The red-dashed and black solid lines represent the fitted regression line for the relationship between the predicted and observed values and the identity line (y = x), respectively.

As a result, analysis based on top-dressed GAA datasets yielded superior predictive performance compared to those using feed-mixed supplementation. In the starter phase, the top-dress Model 1 achieved R² = 0.936 and CCC = 0.967 (Fig. 6), indicating an excellent agreement between predicted and observed ADG values. Even the simplest model based solely on Arg intake (Model 5) explained 89.9% of the variation (R² = 0.899; CCC = 0.947), indicating a strong linear dependency between Arg intake and growth. In the finisher phase, model performance was moderate (R² = 0.69; CCC = 0.825), consistent with reduced marginal returns from additional creatine synthesis at later growth stages.

Fig. 6.

Plot of the observed vs. predicted values for the selected models (see Table 4). The red-dashed and black solid lines represent the fitted regression line for the relationship between the predicted and observed values and the identity line (y = x), respectively.

Predictive application and model validation

The final predictive equations derived from the top-performing models were applied to forecast ADG and cumulative body weight in Ross 308 broilers (Fig. 7) because 73% (30/41) studies used Ross broiler strains. During the starter phase, the regulation of body weight on age showed a higher slope for the GAA-supplementation compared to the control (46.59 and 47.71, respectively), along with higher coefficient of determination (R² = 0.899). For the finisher phase, slopes were 39.01 (control) and 39.97 (GAA), with corresponding R² values of 0.66 and 0.64, respectively. The model predicted indicated that GAA inclusion at 0.06% was associated with an averaged 60 gram increase in final BW (Fig. 7). These models accurately predicted growth trajectories closely aligned with commercial Ross 308 performance benchmarks, demonstrating the applicability of the derived equations for practical nutritional modeling.

Fig. 7.

The relationship between Arg intake with or without Guanidinoacetic acid (GAA) supplementation on average daily gain (g/d) of broiler chickens during starter (day 1-21) and finisher (day 22-42) period, predicted using equation from Model 5 of top-dressed dataset (Table 4). The regression equation for starter period: CON = 1.943 + 46.593 × Arg intake (R² = 0.87; RMSE = 1.12), GAA = 1.943 + 47.708 × Arg intake (R² = 0.92; RMSE = 0.13); and for finisher period: CON = 15.573 + 39.013 × Arg intake (R² = 0.65; RMSE = 1.97), GAA = 15.573 + 39.967 × Arg intake (R² = 0.67; RMSE = 0.34). The lower quadrant represents the prediction of the application of the models on body weight (BW) of commercial broiler chickens (Ross 308; as hatched) during the starter and finisher periods.

Discussion

Arginine-sparing effects of GAA

This meta-analysis provides evidence that dietary GAA supplementation enhances broiler growth performance, regardless of dietary nutrient sufficiency. Pooled results from 41 studies (383 treatment units) demonstrate that adding GAA increases weight gain and improves feed conversion, even amid substantial within-study variation, indicating the consistency of effects across heterogeneous trial conditions. Nevertheless, our results indicated the the minimal response of broilers fed diets with GAA supplementation under diets deficient in Arg but the effect was more pronounced under ME diets. Our results indicated that under Arg-deficient diets (90-95% Arg levels), GAA-fed broilers alleviated the ADG from -4.75% to -3.64% reduction in the starter and from -3.67% to -2.1% in the finisher periods. Dose–response analysis demonstrated minimal additional benefits beyond a supplementation level of 0.06%, suggesting a plateau effect and indicating that higher inclusion rates may not further enhance growth performance. Meta-regression models identified dietary Arg intake and FI as the primary predictors of weight gain, with the inclusion of GAA intake adding significant explanatory power. A quadratic model based on feed intake, Arg, and GAA intake demonstrates higher ADG and final BW (2.11%) in commercial Ross 308 growth. This small to modest ADG recovery supports the concept that GAA can spare only the fraction of arginine used for creatine synthesis (DeGroot et al., 2018), whereas arginine’s other physiological roles in protein synthesis, nitric oxide production, and polyamine metabolism remain limiting.

It has been well conceptualized that GAA exerts arginine-sparing effect via creatine biosynthesis. GAA serves as the direct biochemical precursor of creatine, a key metabolite involved in maintaining intracellular energy equilibrium (Asiriwardhana and Bertolo, 2022). In avian species, endogenous GAA is formed through the transfer of an amidino group from Arg to Gly, a reaction catalyzed by L-Arg:Gly amidinotransferase (Khajali et al., 2020). This pathway utilizes a considerable portion of dietary Arg, thereby influencing the overall AA utilization in broilers. Supplying GAA in the diet reduces the metabolic requirement for Arg for endogenous creatine synthesis (DeGroot et al., 2018), thereby conserving Arg for physiological functions essential to broiler growth. From a nutritional standpoint, exogenous GAA effectively spares dietary Arg, enabling greater utilization of Arg-derived substrates for protein deposition, nitric oxide synthesis, and immunological competence (Verhelle and Saremi, 2025).

The Arg-sparing property of GAA has been consistently confirmed in feeding studies involving broiler chickens. Dilger et al. (2013) reported that dietary inclusion of 0.12% GAA in an Arg-deficient diet fully restored growth rate and feed efficiency to levels equivalent to those achieved under Arg-adequate feeding. This finding indicates that GAA can effectively substitute for the Arg required for endogenous creatine biosynthesis. The Arg-sparing effect was also reported in a study examining increasing GAA supplementation (0.06–0.18%), whereas GAA alleviated the adverse effects of Arg deficiency, although it did not completely achieve the performance of an Arg-adequate diet, regardless of the inclusion levels (Fosoul et al., 2018). Subsequent trials have corroborated these results, showing that broilers maintained on Arg-restricted diets supplemented with GAA display nearly normal growth performance and FCR (Majdeddin et al., 2020; Asiriwardhana and Bertolo, 2022). Collectively, these studies support our findings that dietary GAA helped to alleviate the growth-detrimental effects of Arg deficiency, but full recovery of performance might not be expected.

Consistent with previous experimental findings, the subgroup analysis from the present meta-analysis demonstrated that inclusion of GAA in Arg-restricted diets effectively ameliorates the performance disparity between Arg-deficient and Arg-adequate groups. This outcome indicates that exogenous GAA can successfully substitute for endogenously synthesized GAA, thereby compensating for the creatine requirement necessary to sustain normal energy metabolism and growth. The observed convergence in BW gain and feed efficiency further supports the hypothesis that dietary GAA compensates for the metabolic demand on Arg associated with creatine biosynthesis. Physiologically, this substitution minimizes the Arg flux through the amidinotransferase pathway, conserving more of the AA for anabolic processes such as muscle protein accretion, nitric oxide formation, and immune system maintenance (Asiriwardhana and Bertolo, 2022). Collectively, these findings confirm that GAA can be used as an Arg-sparing and energy-enhancing supplement and provide a quantitative foundation for precision-feeding recommendations.

Contribution of GAA to cellular energy homeostasis

In contrast to the effect of GAA on Arg-deficient diet, the more pronounced response under ME-deficient diets suggests that GAA’s primary benefit is linked to improved phosphocreatine-mediated ATP buffering and energy utilization efficiency, which becomes particularly advantageous when dietary energy supply is constrained. In addition to its Arg-sparing capacity, GAA exerts a physiological role by expanding the intramuscular creatine and phosphocreatine reservoirs, which are integral to maintaining cellular energy stability (El-Aziz et al., 2025). Within metabolically active tissues such as skeletal muscle, the creatine–phosphocreatine system functions as a rapid energy-buffering mechanism. During periods of elevated ATP demand, phosphocreatine donates a high-energy phosphate group to ADP via the creatine kinase reaction, thereby sustaining ATP availability and preventing fluctuations in the cellular energy charge (Asiriwardhana and Bertolo, 2022; Wyss and Kaddurah-Daouk, 2000). Enhancing dietary creatine availability through GAA supplementation strengthens the phosphagen energy-buffering system in skeletal muscle. Experimental evidence in poultry indicates that GAA inclusion elevated intramuscular concentrations of creatine and phosphocreatine, concurrently stimulating creatine kinase activity and improving the overall adenine nucleotide balance. Specifically, elevated ATP content and reduced ADP and AMP levels were observed, reflecting a more efficient energy transfer within muscle fibers (Brault et al., 2003; Lemme et al., 2007).

Consistent with these findings, Portocarero and Braun (2021) reported that broilers fed a dietary GAA diet exhibited greater total high-energy phosphate metabolites than those fed a control diet, suggesting that GAA intake enhances the stability of muscular energy homeostasis. Findings from our meta-regression further suggested that the enhanced growth performance associated with GAA supplementation is primarily driven by improved energy utilization efficiency. Notably, GAA supplementation resulted in a significant improvement in FCR without a corresponding increase in voluntary FI, indicating that supplemented birds derived greater ME from each unit of feed consumed. In addition, dietary GAA fortifies the creatine-phosphagen system, diminishing the energetic cost of maintenance and increasing the proportion of dietary energy directed toward lean tissue accretion. In broilers fed energy-restricted diets, inclusion of 0.06% GAA effectively counteracted growth suppression (Pirgozliev et al., 2022), highlighting its capacity to buffer cellular energy under suboptimal conditions. A recent study estimated the ME-contributing effects of 0.06% GAA to be 88 kcal/kg diet (Salgado et al., 2023). Although another study reported inconsistent results in which a minimal growth recovery effect of GAA was reported from broilers fed 50-100 kcal/kg ME-deficient diets (Verhelle and Saremi, 2025), this study also found no difference between adequate vs. ME-restricted diets, indicating that the broilers might physiologically have sufficient energy supply. Nevertheless, the energy homeostasis activity and energy-sparing effects of GAA warrant further investigation across various conditions, feed ingredients, and dietary ME levels.

Dose-response effects of GAA and predictive growth

Our analysis demonstrated distinct dose–response effects of GAA on ADG between the starter and finisher periods, whereas a low dose of GAA (∼0.06%) resulted in a similar improvement, whereas high GAA inclusion was only found in the starter period. The lack of difference between low and high GAA supplementation during the starter period suggests a plateau at higher inclusion levels. Dilger et al. (2013) reported that 0.12% GAA represents the threshold beyond which additional supplementation does not further improve growth in Arg-deficient chicks. Similarly, Ahmadipour et al. (2018b) and Mohebbifar et al. (2022) observed that inclusion levels above 0.15-0.20% failed to provide incremental benefits. This plateau likely reflects physiological constraints on creatine uptake, as muscle creatine transporters reach saturation and excess GAA or creatine is either excreted or fails to contribute further to metabolic processes. Excessive GAA supplementation can also induce feedback inhibition of endogenous creatine synthesis (Asiriwardhana and Bertolo, 2022; Dinesh et al., 2021; Tossenberger et al., 2016), underscoring that most benefits can be achieved within a low-to-moderate inclusion range.

The optimal dose, however, might be influenced by nutritional design of the diets, i.e., birds consuming highly Arg-deficient feeds may benefit from slightly higher GAA levels, whereas lower doses are sufficient in energy and Arg-standard diets, as evidenced in previous experiments in which no further improvement on Arg-deficient diet was found with GAA >0.12% (Córdova-Noboa et al., 2018b; DeGroot et al., 2019; Sharma et al., 2022). Supporting this, several studies have documented a pronounced tapering of growth benefits at GAA inclusion levels above 0.15% in feed (Ahmadipour et al., 2018a; Mohebbifar et al., 2022; Asiriwardhana & Bertolo, 2022). Consistently, meta-regression models incorporating quadratic terms for GAA indicated diminishing returns at higher supplementation levels. These findings suggest that the practical optimal inclusion range lies between 0.06% and 0.12%, beyond which additional GAA is unlikely to enhance growth performance and may unnecessarily increase feed costs. Consequently, the GAA dose–response relationship is context-dependent, but our pooled analysis and meta-regression provide a practical guideline for formulating typical commercial broiler diets.

Furthermore, our mixed-effects modeling highlights the pivotal role of diet composition in determining growth responses. Arg intake and FI emerged as the strongest predictors of daily gain, and including GAA intake as an additional covariate improved model fit and ADG. Our modeling indicated strong agreement between predicted and observed body weights in commercial Ross 308 broilers during the starter phase, with a CCC value of 0.97. Predictions for the finisher phase remained robust (CCC ∼0.84) but exhibited greater variability, reflecting both the increased biological heterogeneity of older birds and their reduced sensitivity to GAA supplementation. This finding suggests that GAA provides additional explanatory power by providing extra energy-buffering capacity. This finding supports the mechanistic framework in which GAA improves broiler performance both via Arg-sparing and an independent effect on cellular energy metabolism.

Practical implications and recommendations

Given the observed plateau in ADG response to GAA, this meta-analysis recommends a low to moderate inclusion of GAA in broiler diets. An inclusion of 0.06% appears sufficient to capture the majority of performance benefits, provided that birds maintain adequate FI. This recommended inclusion level is confirmed and consistent with observed efficacy reported across multiple studies (Dilger et al., 2013; Sharma et al., 2022). From a practical standpoint, nutritionists should balance expected performance gains against feed costs, as supplementation beyond 0.06% is unlikely to be economically justified given the minimal incremental growth benefit. Although not specifically examined in this meta-analysis due to small sample size, Arg level in the diet may be reduced to 90% from the recommended level with the use of 0.06% dietary GAA without compromising broiler performance. By conserving Arg for creatine synthesis, GAA provides formulation flexibility. This application might be useful when lower corn and soybean use is needed in feed formulation due to limited availability. Critical AA, particularly methionine, must not be compromised; in fact, methionine levels should be maintained at or slightly above requirement to support GAA’s methylation demand and ensure efficient creatine synthesis.

Our analysis indicated that GAA supplementation is particularly beneficial under challenged conditions. In contemporary poultry production, diets are often formulated with reduced CP levels to lower feed costs and nitrogen emissions, and are supplemented with crystalline AA to meet essential requirements. Under such scenarios, GAA can mitigate the growth penalties associated with lower protein by enhancing creatine-mediated energy availability, thereby supporting optimal performance despite constrained nutrient supply. Under environmental stressors such as heat, cold, or high stocking density, GAA supplementation may confer benefits beyond growth promotion. In heat-stressed broilers, GAA has been reported to attenuate elevations in core body temperature and circulating stress hormones (Li et al., 2023), while also enhancing intestinal integrity and barrier function (Peng et al., 2023). In cold-stress models, such as those that induce ascites, GAA supplementation has been shown to enhance cardiovascular performance and improve survival (Majdeddin et al., 2019; Boroumandnia et al., 2021). Similarly, under high stocking density, which typically depresses growth. Alaa et al. (2024) reported that GAA-supplemented flocks exhibited higher weight gain and enhanced antioxidant status compared with controls. Therefore, under challenging rearing conditions, supplementation with GAA at approximately 0.06% may enhance flock resilience. In practical terms, producers confronting heat stress or intensive housing can consider GAA as a complementary strategy to conventional management practices, such as improved ventilation or cooling systems, to mitigate the negative impacts of environmental stressors.

Sources of heterogeneity and limitations

While the advantages of GAA supplementation are well documented, the magnitude of its effects is highly variable across studies, as reflected by the pronounced between-study heterogeneity observed in our analysis (I² > 90%). This variability likely arises from multiple interacting factors influencing GAA responses in broilers, particularly between-study rather than within-study variance. Such large heterogeneity is typically expected in feeding studies involving broiler chickens (Ningsih et al., 2023; Weaver et al., 2022; Wei et al., 2024; Yano et al., 2025). Although variations in broiler strain, sex, and age can markedly influence the magnitude and consistency of the effect, their contribution to the high heterogeneity value was minimal, as evidenced by the non-significant effects of these factors, except the strain effect, which was more likely due to between-study effects rather than the strain itself. The sex of broilers could also influence the outcome. Male broilers typically exhibit faster growth and higher FI than females, which may render them more receptive to the enhanced energy buffering provided by GAA (Al-Abdullatif et al., 2024). In this meta-analysis, no study used exclusively female broilers. This suggests minimal bias from the sex of broilers, as the male-dominant sex is representative of many of the broiler industry standards, and also evidence from the non-statistical effect of sex. Another potential heterogeneity and bias may arise from dietary composition and ingredients. Although dietary formulations for major broiler strains such as Ross, Cobb, and Arbor-Acres have relatively similar standards, sources of feed ingredients, and different formulation approaches, these differences can yield a minor to moderate bias, particularly when the experimental design involves different degrees of nutrient deficiency.

Statistical assessments revealed minimal small-study bias for growth performance, although some asymmetry was observed for FI outcomes. Leave-one-out sensitivity analyses confirmed that no individual study disproportionately influenced the results. Nevertheless, caution is warranted, as the published literature may underrepresent null or negative findings. Unpublished trials with neutral GAA effects likely exist, particularly under diets that already meet optimal nutrient and energy requirements, where the scope for additional gains is limited. Given the growing interest in GAA within the poultry industry, all study outcomes, including neutral or negative results, must be reported to improve the accuracy of effect estimates.

Despite the overall robustness of the findings, several limitations should be acknowledged. The substantial residual heterogeneity observed across studies suggests that additional, unmeasured factors, such as genetic line, microclimate, disease pressure, or pellet quality, may influence the magnitude of GAA effects. Furthermore, potential interactions with other feed additives, such as betaine, vitamins, or probiotics, remain poorly characterized. In particular, diets that marginally meet methionine requirements or contain high levels of alternative methyl donors (e.g., betaine or choline) could influence the efficiency of GAA utilization. The potential interaction, additive, or antagonistic effects with other feed additives should be evaluated. Although our predictive equations show considerable promise, they require validation under independent conditions. Field trials conducted in commercial production settings would be especially valuable to confirm both biological responses and economic returns. Additionally, more detailed mechanistic studies, such as tracing Arg and methyl group partitioning and assessing creatine transporter expression, are needed to elucidate further how GAA interacts with avian metabolism and influences performance and tissue physiology.

Conclusions

On the basis of the cumulative results, dietary supplementation with GAA improved ADG and FCR in broiler chickens during both the starter and finisher phases. The response plateau around 0.12% indicates that low or moderate inclusion is sufficient for most practical applications, but lower inclusion (0.06%) is economically preferable. Dietary GAA exerts an moderate Arg-sparing effect at certain levels. Additional studies are required to verify the sparing and energy-saving effects under commercial and challenging conditions. Under normal nutritional and physiological conditions, the simulation study using the Ross 308 performance objective showed that GAA inclusion at 0.06% improved final BW by 2.11% (60 g) compared with the control diet. Predictive models derived from meta-regression accurately estimate starter growth and provide useful guidance for finisher-phase performance, supporting precision ration formulation. Nevertheless, substantial heterogeneity and residual variance underscore the need for careful application and further validation. Incorporating GAA into contemporary precision-nutrition strategies offers both economic and environmental advantages in broiler production.

Data availability

All data included in this study are available upon request by contacting the corresponding author.

CRediT authorship contribution statement

Min Gao: Writing – original draft, Validation, Investigation, Data curation, Conceptualization. Mohamed El-Sherbiny: Writing – review & editing, Validation, Data curation. Bartosz Kierończyk: Writing – review & editing, Validation, Data curation. Hao Guo: Writing – review & editing, Validation, Data curation. Montaser Elsayed Ali: Writing – review & editing, Validation, Data curation. Abdel Moneim Eid Abdel-Moneim: Writing – review & editing, Validation, Data curation, Conceptualization. Mhd. Adanan Purba: Writing – review & editing, Validation, Data curation. Luthfi Adya Pradista: Writing – review & editing, Validation, Data curation. Wahyu Subagio Saputro: Writing – review & editing, Validation, Data curation. Adi Ratriyanto: Writing – review & editing, Validation, Data curation. Wara Pratitis Sabar Suprayogi: Writing – review & editing, Validation, Data curation. Yulianri Rizki Yanza: Writing – review & editing, Validation, Data curation, Conceptualization. Agung Irawan: Writing – review & editing, Writing – original draft, Visualization, Supervision, Software, Resources, Project administration, Methodology, Investigation, Funding acquisition, Formal analysis, Data curation, Conceptualization.

Disclosures

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.