Literature search

Fig. 1 presents the detailed study search and selection steps performed from April to June 2022 and updated in March 2025^{(Reference Page, McKenzie and Bossuyt33)}. After applying the search strategy, a total of 2641 potentially relevant publications were identified in Scopus, 930 in MEDLINE via PubMed, 735 in EMBASE, 583 in Web of Science, 550 in CINAHL, 300 in Google Scholar (grey literature) and 3 in LILACS, totalling 5742 publications. The abstracts from the databases were grouped and imported into the free online version of Endnote Reference Management. The program identified 1239 duplicates. Afterwards, a total of 4503 publications were imported into the Rayyan software, with another 418 duplicates being identified, resulting in a total of 1657 discarded duplicates and 4085 records for which the titles and abstracts were read. After reading the titles and abstracts, a total of 4046 records were excluded because they did not meet the inclusion criteria. The remaining thirty-nine publications were read in full. After reading, eighteen of them were excluded and twenty-one were selected according to the eligibility criteria.

Figure 1.

Search and selection flowchart according to PRISMA^{(Reference Page, McKenzie and Bossuyt33)}.

In March 2025, after updating the search in the databases, thirteen records were added that also met the inclusion criteria and three records from manual search. Thus, thirty-seven records were included^{(Reference Yilmaz and Açık16,Reference Wang, Armijos and Weigel24,Reference Shivappa, Hebert and Marcos51–Reference Vingrys, Hébert and Chen85)} , thirty-five scientific articles and two abstracts from conference proceedings. There are two studies each by the Iranian group Shipava et al.^{(Reference Shivappa, Hebert and Rashidkhani52,Reference Shivappa, Hebert and Neshatbini Tehrani53)} and the Brazilian group Suhett et al.^{(Reference Suhett, Hermsdorff and Ribeiro67,Reference Suhett, Vieira Ribeiro and Hermsdorff68)} included in this review. The participants in both studies from each group come from the same sample. Furthermore, Navarro et al.^{(Reference Navarro, Shivappa and Hébert60,Reference Navarro, Shivappa and Hébert61)} and Barragán-Vázquez et al.^{(Reference Barragán-Vázquez, Ariza and Pedraza62,Reference Barragán-Vázquez, Ariza and Ramírez Silva63)} each have two publications included in this review; for each group, one publication is an article and one publication is a conference abstract. For the study by Vingrys et al.^{(Reference Vingrys, Hébert and Chen85)}, we evaluated individual data from participants in five cohorts^{(Reference Boyd, Golding and Macleod86–Reference Hofman, Jaddoe and Mackenbach90)}. However, according to Cochrane^{(Reference Higgins, Thomas and Chandler32)}, the unit of interest for data interpretation is the studies that served as the origin for these articles and summaries, thus we included thirty-three studies^{(Reference Yilmaz and Açık16,Reference Wang, Armijos and Weigel24,Reference Shivappa, Hebert and Marcos51–Reference Vingrys, Hébert and Chen85)} in the qualitative synthesis and eighteen studies^{(Reference Yilmaz and Açık16,Reference Correa-Rodríguez, González-Jiménez and Rueda-Medina54,Reference Aslani, Qorbani and Hébert58,Reference Barragán-Vázquez, Ariza and Pedraza62–Reference Aljahdali, Peterson and Cantoral69,Reference Costello, Goodrich and Patterson71,Reference Jia, Wu and Su75–Reference Blaudt, de Souza Lopes and de Moura Souza79,Reference Zhang, Peng and Wang81,Reference Du, Hao and Yu83,Reference Vingrys, Hébert and Chen85)} in the quantitative synthesis (meta-analysis).

Study characteristics

We extracted data from thirty-three studies^{(Reference Yilmaz and Açık16,Reference Wang, Armijos and Weigel24,Reference Shivappa, Hebert and Marcos51–Reference Vingrys, Hébert and Chen85)} – published between 2017 and 2025 – on the association between the DII and indicators of body adiposity and BP in children and adolescents (Table 2). The studies were mostly from Asia (n = 11)^{(Reference Yilmaz and Açık16,Reference Shivappa, Hebert and Rashidkhani52,Reference Shivappa, Hebert and Neshatbini Tehrani53,Reference Karera, Witjaksono and Khusun55,Reference Aslani, Qorbani and Hébert58,Reference Aslani, Qorbani and Hébert58,Reference Kurklu, Torun and Kucukcetin64,Reference Dehghan, Nejati and Vahid72,Reference Gholamalizadeh, Ahmadzadeh and BourBour74,Reference Li, Li and Wang76,Reference Kanısoy and Kabaran80,Reference Şahin, Yardımcı and Açık84)} , followed by the USA (n = 9)^{(Reference Shivappa, Hebert and Marcos51,Reference Khan, Wirth and Ortaglia56,Reference Sen, Rifas-Shiman and Shivappa57,Reference Sethna, Alanko and Wirth66,Reference Costello, Goodrich and Patterson71,Reference Ducharme-Smith, Brady and Vizthum73,Reference Jia, Wu and Su75,Reference Zhang, Peng and Wang81,Reference Du, Hao and Yu83)} , Latin America (n = 7)^{(Reference Wang, Armijos and Weigel24,Reference Shivappa, Hebert and Rashidkhani52,Reference Barragán-Vázquez, Ariza and Ramírez Silva63,Reference Suhett, Hermsdorff and Ribeiro67–Reference Aljahdali, Peterson and Cantoral69,Reference Todendi, Salla and Shivappa78,Reference Blaudt, de Souza Lopes and de Moura Souza79,Reference Cota, Filgueiras and Pereira82)} and Europe (n = 5)^{(Reference Correa-Rodríguez, González-Jiménez and Rueda-Medina54,Reference Navarro, Shivappa and Hébert60,Reference Navarro, Shivappa and Hébert61,Reference Buckland, Northstone and Emmett70,Reference Mora-Urda, Martín-Almena and Montero López77,Reference Vingrys, Hébert and Chen85)} . One study involved several European countries as well as Oceania^{(Reference Davis, Liu and Kerr59)}.

Table 2.

Characteristics of the studies included in this systematic review

(–), no information; *multicountry (UK, Ireland, Poland and the Netherlands); ALSPAC, The Avon Longitudinal Study of Parents and Children; SWS, Southampton Women’s Survey; Lifeways, the Lifeways cross-generation cohort study; REPRO_PL, The Polish Mother and Child Cohort; GEN R,the Generation R study. BMI, body mass index; BIA, bioelectrical impedance analysis; BP, blood pressure; CDII, children’s dietary inflammatory index; DBP, diastolic blood pressure; DEXA, dual-energy X-ray absorptiometry; DII, dietary inflammatory index; FFM, fat free mass (kg); FM, fat mass (kg); FMI, fat mass index (fat mass (kg)/height (m)²); FFMI, fat free mass index (fat free mass (kg)/height (m)²); HBP, hypertensive blood pressure; MAP, mean arterial pressure; NC, neck circumference; OR, odds ratio; PR, prevalence ratios; SBP, systolic blood pressure; VAT, visceral adipose tissue; VFA, visceral fat area; WC, waist circumference; WHR, waist hip ratio; WHtR, waist-to-height ratio; 95% CI, 95% confidence interval; %TFM, total fat mass (%).

^a WHO (1998) – BMI (kg/m²): Category I under or normal weight (i.e. <25 kg/m²); Category II overweight (i.e., ≥25 kg/m² and <30 kg/m²); Category III obese (i.e., ≥30 kg/m²).

^b Hypertension and metabolic syndrome were identified according to the IDF criteria.

^c The DII (squared) was analyzed both as a continuous variable and a dichotomous variable, categorised on the basis of the median value of the DII for the controls (0·02).

^d Hypertensive BP defined as BP ≥130 mmHg systolic or ≥80 mmHg diastolic for adolescents ≥13 years and BP ≥95th percentile for age, sex and height for adolescents that were 12 years old. Elevated blood pressure defined as BP 120/<80 to 129/<80 mmHg for adolescents ≥13 years and BP ≥90th percentile to <95th percentile for age, sex and height for adolescents that were 12 years old^{(Reference Flynn, Kaelber and Baker-Smith44)}.

^e Blood pressure index was calculated as raw, mean blood pressure divided by the age, sex and height specific blood pressure at the 95th percentile (%tile) for all participants (95%tile measures at 17 years of age were used for those >17 years). Height and weight measurements were used to calculate age and sex standardised body mass index (BMI) z-scores (BMIz) using the CDC 2000 growth reference population in participants ≤18 years with weight status classified as ‘underweight’ <5th %tile, ‘normal’ 5^th to <85th %tile, ‘overweight’ 85th to <95th %tile or ‘obese’ ≥95th %tile. BMI was calculated as weight (kg) over squared height (m²) in participants 19–21 years with weight status classified as ‘underweight’ <18.5 kg/m², ‘normal’ 18.5–24.9 kg/m², ‘overweight’ 25–29.9 kg/m² or ‘obese’ ≥30 kg/m². Waist circumference measurements were used to calculate age and sex standardised WC z-scores (z-WC) based on NHANES III data.

^f Average CDII scores were broken into groups on the basis of the median intake of the sample. Range of scores: pro-inflammatory diet (above median: 0·05–1·22), anti-inflammatory diet (below median: −2·14 to −0·04) (lower score/quantile = more anti-inflammatory diet).

^g DII was analyzed as both dichotomous (categorised on the basis of the median value of the DII) and continuous variables. Criteria developed age- and sex-specific cut-points for WC and BP: NHANES III 1988–1994 and the 1999–2000 and 2001–2002 NHANES surveys.

^h According to the Working Group on Obesity in China, obesity was defined as a BMI of at least 28 kg/m² and overweight was defined as 24 kg/m² ≤ BMI < 28 kg/m². A person with 18.5 kg/m² ≤ BMI < 24 kg/m² was normal weight and <18.5 kg/m² constituted thinness.

ⁱ High blood pressure: ≥95th percentile of the National High Blood Pressure Education Program.

Most of the included studies were cross-sectional (n = 23)^{(Reference Yilmaz and Açık16,Reference Wang, Armijos and Weigel24,Reference Shivappa, Hebert and Marcos51–Reference Khan, Wirth and Ortaglia56,Reference Aslani, Qorbani and Hébert58,Reference Davis, Liu and Kerr59,Reference Kurklu, Torun and Kucukcetin64,Reference Sethna, Alanko and Wirth66–Reference Suhett, Vieira Ribeiro and Hermsdorff68,Reference Dehghan, Nejati and Vahid72–Reference Gholamalizadeh, Ahmadzadeh and BourBour74,Reference Li, Li and Wang76–Reference Kanısoy and Kabaran80,Reference Cota, Filgueiras and Pereira82–Reference Şahin, Yardımcı and Açık84)} ; nine studies^{(Reference Sen, Rifas-Shiman and Shivappa57,Reference Navarro, Shivappa and Hébert60–Reference Barragán-Vázquez, Ariza and Ramírez Silva63,Reference Aljahdali, Peterson and Cantoral69–Reference Costello, Goodrich and Patterson71,Reference Jia, Wu and Su75,Reference Zhang, Peng and Wang81,Reference Şahin, Yardımcı and Açık84)} were prospective (cohort) and one study was case-control^{(Reference Vahid, Bourbour and Gholamalizadeh65)}. The number of individuals included in the study samples ranged from 90 to 71 740. Regarding gender, in fourteen studies^{(Reference Yilmaz and Açık16,Reference Wang, Armijos and Weigel24,Reference Correa-Rodríguez, González-Jiménez and Rueda-Medina54,Reference Navarro, Shivappa and Hébert60,Reference Navarro, Shivappa and Hébert61,Reference Kurklu, Torun and Kucukcetin64,Reference Sethna, Alanko and Wirth66–Reference Aljahdali, Peterson and Cantoral69,Reference Costello, Goodrich and Patterson71,Reference Li, Li and Wang76,Reference Todendi, Salla and Shivappa78,Reference Blaudt, de Souza Lopes and de Moura Souza79,Reference Cota, Filgueiras and Pereira82,Reference Şahin, Yardımcı and Açık84)} there were more women than men, two studies^{(Reference Shivappa, Hebert and Rashidkhani52,Reference Shivappa, Hebert and Neshatbini Tehrani53,Reference Dehghan, Nejati and Vahid72)} included only girls, one study^{(Reference Gholamalizadeh, Ahmadzadeh and BourBour74)} included only boys, two studies^{(Reference Buckland, Northstone and Emmett70,Reference Du, Hao and Yu83)} showed an equal gender distribution and six studies^{(Reference Shivappa, Hebert and Marcos51–Reference Karera, Witjaksono and Khusun55,Reference Vahid, Bourbour and Gholamalizadeh65,Reference Vingrys, Hébert and Chen85)} did not provide this information. The minimum age was 2·8 years in the study by Sen et al.^{(Reference Page, McKenzie and Bossuyt33)} and the maximum was 22 years in the studies by Costello et al.^{(Reference Balshem, Helfand and Schünemann48)} and Li et al.^{(Reference Shivappa, Hebert and Neshatbini Tehrani53)}. The age group classifications (children and adolescents) according to the country were respected in these two studies.

Most of the included studies evaluated the DII proposed by Shippava et al.^{(Reference Shivappa, Steck and Hurley25)} (n = 20)^{(Reference Wang, Armijos and Weigel24,Reference Shivappa, Hebert and Marcos51–Reference Karera, Witjaksono and Khusun55,Reference Sen, Rifas-Shiman and Shivappa57,Reference Aslani, Qorbani and Hébert58,Reference Navarro, Shivappa and Hébert60,Reference Navarro, Shivappa and Hébert61,Reference Kurklu, Torun and Kucukcetin64,Reference Vahid, Bourbour and Gholamalizadeh65,Reference Costello, Goodrich and Patterson71,Reference Dehghan, Nejati and Vahid72,Reference Gholamalizadeh, Ahmadzadeh and BourBour74–Reference Blaudt, de Souza Lopes and de Moura Souza79,Reference Zhang, Peng and Wang81,Reference Du, Hao and Yu83)} to indicate a pro-inflammatory diet as an exposure variable. Ten studies^{(Reference Yilmaz and Açık16,Reference Khan, Wirth and Ortaglia56,Reference Barragán-Vázquez, Ariza and Pedraza62,Reference Barragán-Vázquez, Ariza and Ramírez Silva63,Reference Sethna, Alanko and Wirth66–Reference Aljahdali, Peterson and Cantoral69,Reference Ducharme-Smith, Brady and Vizthum73,Reference Kanısoy and Kabaran80,Reference Cota, Filgueiras and Pereira82,Reference Vingrys, Hébert and Chen85)} used the CDII^{(Reference Khan, Wirth and Ortaglia26)}, while two studies^{(Reference Davis, Liu and Kerr59,Reference Buckland, Northstone and Emmett70)} created their own tool or adapted the DII^{(Reference Shivappa, Steck and Hurley25)} or CDII^{(Reference Khan, Wirth and Ortaglia26)}. One study used both the CDII and DII^{(Reference Du, Hao and Yu83)}. There was variation in the number of dietary parameters used in the DII or CDII. For the studies that used the DII, the number of parameters ranged from twenty-five to forty-five; two studies did not report the quantity^{(Reference Sen, Rifas-Shiman and Shivappa57,Reference Li, Li and Wang76)} , and two studies^{(Reference Karera, Witjaksono and Khusun55,Reference Şahin, Yardımcı and Açık84)} used the forty-five dietary parameters described in the original DII publication^{(Reference Shivappa, Steck and Hurley25)}. In the studies that evaluated the CDII, the number of parameters varied from sixteen^{(Reference Yilmaz and Açık16,Reference Cota, Filgueiras and Pereira82)} to twenty-five^{(Reference Khan, Wirth and Ortaglia56,Reference Barragán-Vázquez, Ariza and Pedraza62,Reference Barragán-Vázquez, Ariza and Ramírez Silva63,Reference Sethna, Alanko and Wirth66,Reference Aljahdali, Peterson and Cantoral69,Reference Kanısoy and Kabaran80,Reference Şahin, Yardımcı and Açık84)} , the number of parameters described in the original CDII publication^{(Reference Khan, Wirth and Ortaglia26)}.

BMI was used as an outcome in all included studies except one^{(Reference Du, Hao and Yu83)}. Most of the studies used WC as a parameter for abdominal obesity^{(Reference Yilmaz and Açık16,Reference Wang, Armijos and Weigel24,Reference Correa-Rodríguez, González-Jiménez and Rueda-Medina54,Reference Sen, Rifas-Shiman and Shivappa57,Reference Aslani, Qorbani and Hébert58,Reference Navarro, Shivappa and Hébert60–Reference Kurklu, Torun and Kucukcetin64.Reference Sethna, Alanko and Wirth66,Reference Aljahdali, Peterson and Cantoral69,Reference Ducharme-Smith, Brady and Vizthum73,Reference Jia, Wu and Su75,Reference Mora-Urda, Martín-Almena and Montero López77,Reference Blaudt, de Souza Lopes and de Moura Souza79,Reference Zhang, Peng and Wang81,Reference Şahin, Yardımcı and Açık84,Reference Vingrys, Hébert and Chen85)} , while some also assessed the WHtR^{(Reference Wang, Armijos and Weigel24,Reference Correa-Rodríguez, González-Jiménez and Rueda-Medina54,Reference Aslani, Qorbani and Hébert58,Reference Mora-Urda, Martín-Almena and Montero López77,Reference Blaudt, de Souza Lopes and de Moura Souza79,Reference Zhang, Peng and Wang81,Reference Şahin, Yardımcı and Açık84)} , the WHR^{(Reference Correa-Rodríguez, González-Jiménez and Rueda-Medina54,Reference Aslani, Qorbani and Hébert58)} and NC^{(Reference Aslani, Qorbani and Hébert58)}. BP was also evaluated as an outcome variable, based on SBP and DBP^{(Reference Yilmaz and Açık16,Reference Wang, Armijos and Weigel24,Reference Correa-Rodríguez, González-Jiménez and Rueda-Medina54,Reference Sethna, Alanko and Wirth66,Reference Aljahdali, Peterson and Cantoral69,Reference Buckland, Northstone and Emmett70,Reference Ducharme-Smith, Brady and Vizthum73,Reference Mora-Urda, Martín-Almena and Montero López77)} , mean arterial pressure^{(Reference Davis, Liu and Kerr59,Reference Buckland, Northstone and Emmett70,Reference Mora-Urda, Martín-Almena and Montero López77)} , high BP^{(Reference Jia, Wu and Su75,Reference Todendi, Salla and Shivappa78)} , the diagnosis of hypertension^{(Reference Kurklu, Torun and Kucukcetin64,Reference Sethna, Alanko and Wirth66,Reference Mora-Urda, Martín-Almena and Montero López77,Reference Du, Hao and Yu83)} or even as a continuous variable^{(Reference Sen, Rifas-Shiman and Shivappa57,Reference Kurklu, Torun and Kucukcetin64,Reference Sethna, Alanko and Wirth66)} . Finally, indicators of body adiposity such as TFM%, FM, FFM, skin folds, the FMI, the FFMI and the visceral fat area were evaluated in some studies^{(Reference Correa-Rodríguez, González-Jiménez and Rueda-Medina54,Reference Sen, Rifas-Shiman and Shivappa57,Reference Barragán-Vázquez, Ariza and Pedraza62,Reference Barragán-Vázquez, Ariza and Ramírez Silva63,Reference Sethna, Alanko and Wirth66–Reference Suhett, Vieira Ribeiro and Hermsdorff68,Reference Buckland, Northstone and Emmett70,Reference Costello, Goodrich and Patterson71,Reference Gholamalizadeh, Ahmadzadeh and BourBour74,Reference Li, Li and Wang76,Reference Mora-Urda, Martín-Almena and Montero López77,Reference Vingrys, Hébert and Chen85)} . However, owing to heterogeneity among the studies, it was not possible to carry out a quantitative synthesis of these data.

Risk of bias and assessment of methodological quality

On the basis of the NOS, the included publications presented moderate quality, with scores ranging from four to six stars for cross-sectional studies and from six to eight stars for cohort and case-control studies (Table 3)^{(Reference Wells, Shea and O’Connell46,Reference Patra, Bhatia and Suraweera47)} .

Table 3.

Results of the methodological quality assessment of the studies included in the systematic review according to the Newcastle–Ottawa scale^{(Reference Wells, Shea and O’Connell46,Reference Patra, Bhatia and Suraweera47)}

–, undetermined; NA, not applicable. Cross-sectional studies: maximum score of seven asterisks (or stars). Moderate quality: 4–6 stars, low quality: ≤3 stars.

^a Cohort

^b Case-control studies: maximum score of nine asterisks (or stars). Moderate quality: 6–8 stars, low quality: ≤5 stars

Association between the DII and cardiometabolic risk factors

DII and BMI

The results of the meta-analyses on pro-inflammatory diet and BMI from analyses involving linear and logistic regression models are presented in Figs. 2 and 3, respectively. In studies that used linear regression, there was a small but significant association between the DII and BMI in the evaluated sample. In other words, in children and adolescents, each unit of the DII was associated with a 0·06 kg/m² higher mean BMI (β = 0.06; 95% CI 0·03–0·06) (Fig. 2). However, in studies that used logistic regression, adolescents with a higher DII (quartile four or above the median, depending on the study) had a 62% greater chance of developing excess weight compared with those with a lower DII (quartile one or below the median; OR = 1·62, 95% CI 1·38–1·86; Fig. 3).

Figure 2.

Meta-analysis to quantify the association between DII and BMI levels in children and adolescents (linear regression). DII, dietary inflammatory index; F, females; M, males; 95% CI, 95% confidence interval.

Figure 3.

Meta-analysis to quantify the association between pro-inflammatory diet (higher DII scores) and overweight (BMI) in adolescents (logistic regression). DII, dietary inflammatory index; F, females; M, males; 95% CI, 95% confidence interval.

DII and abdominal obesity

Figs. 4 and 5 show the association between the DII and abdominal obesity, based on WC and the WHtR, respectively. Adolescents with a higher DII (quartile four) were 45% more likely to have abdominal obesity, according to WC, compared with those with a lower DII (quartile one) (OR = 1·45, 95% CI 1·10–1·81; Fig. 4). Similarly, adolescents with a higher DII (quartile four) were 76% more likely to have abdominal obesity, according to the WHtR, compared with individuals with a lower DII (quartile one) (OR = 1·76, 95% CI 1·38–2·14; Fig. 5).

Figure 4.

Meta-analysis to quantify the association between pro-inflammatory diet (higher DII scores) and abdominal obesity (WC) in adolescents (logistic regression). DII, dietary inflammatory index; F, females; M, males; 95% CI, 95% confidence interval.

Figure 5.

Meta-analysis to quantify the association between pro-inflammatory diet (higher DII scores) and abdominal obesity (WHtR) in adolescents (logistic regression). DII, dietary inflammatory index; F, females; M, males; 95% confidence interval.

Association between the CDII and cardiometabolic risk factors

There were no statistically significant associations between the CDII and cardiometabolic risk factors. Specifically, there was no association between the CDII and BMI (β = 0·01; 95% CI −0·01 to 0·03; Fig. S3). Similarly, linear regression analyses revealed no significant associations between a pro-inflammatory dietary profile and abdominal obesity (Figs. S4 and S5). In addition, there were no associations between the CDII and SBP (Fig. S6) or DBP (Fig. S7).

Grade

The certainty of the evidence for the outcomes (body adiposity indicators and BP) was downgraded from low to very low due to the inconsistency and methodological quality of the included publications. The level of evidence was not downgraded owing to imprecision, indirect evidence or publication bias. Therefore, the development of more robust studies is necessary to increase confidence in the results reported in this systematic review and meta-analysis.

Discussion

This systematic review and meta-analysis synthesised evidence from observational studies assessing the association between the inflammatory potential of the diet (based on the DII or CDII) and indicators of body adiposity and BP in children and adolescents. Overall, the findings suggest that a pro-inflammatory dietary profile, denoted by a higher DII, is positively associated with increased body adiposity (based on BMI, WC and the WHtR), particularly in adolescents. When considering children and adolescents, the DII is also associated with BMI, but with a small effect. There are no significant associations between the CDII and cardiometabolic risk factors, or between the DII or CDII and BP.

The results of this systematic review and meta-analysis suggest that a diet with a greater inflammatory potential, as indicated by the DII and CDII, may be associated with higher levels of body adiposity in children and adolescents. The interpretation of these findings should consider that dietary habits have cumulative and gradual effects on the body’s inflammatory status, even during childhood. A diet rich in ultra-processed foods, added sugars, saturated fats and food additives can trigger low-grade inflammatory responses, modulate immunometabolic function and promote adipose tissue accumulation, particularly in central regions of the body^{(Reference Minihane, Vinoy and Russell91,Reference Méndez López, González Llerena and Vázquez Rodríguez92)} .

However, the strength and direction of the associations with body adiposity varied between the indices. There were more consistent and clinically relevant findings observed for the DII – especially among adolescents – for both general indicators (such as BMI) and central adiposity markers (such as WC and the WHtR), while the CDII did not show significant associations with cardiometabolic risk factors. The findings related to the CDII were more heterogeneous and showed unexpected directions, which may reflect limitations in adapting the CDII to paediatric populations, both in how the index is constructed – consideration of nutrients not always suited to children’s needs – and in how dietary data are collected and reported in studies involving children^{(Reference Suhett, Hermsdorff and Cota17,Reference Sethna, Alanko and Wirth66)} .

Moreover, the smaller number of studies using the CDII, combined with methodological variability, may have contributed to unstable and inconsistent estimates. This underscores the need for greater standardisation in the application of the CDII and for validation across specific age groups and pubertal stages. This endeavour would ensure that the CDII accurately reflects the inflammatory potential of the diet in different contexts of child development.

It is also noteworthy that the association between the DII and adiposity indicators was more evident among adolescents, which may reflect not only hormonal and behavioural changes characteristic of this developmental stage but also greater stability in dietary patterns compared with younger children^{(Reference Burrows, Whatnall and Patterson93)}. Although the effect sizes observed in samples comprising children and adolescents were generally small, the consistency of the associations across different outcomes and age groups of children suggests a potentially meaningful trend from the clinical and epidemiological perspectives^{(Reference Shivappa, Godos and Hebert94,Reference Larruy-García, Mahmood and Miguel-Berges95)} .

Our findings are consistent with published evidence in adult populations: there is a clear association between a higher DII and increased adiposity and metabolic disorders^{(Reference Aslani, Sadeghi and Bagheri96–Reference Bakhshimoghaddam, Bagheri and Daneshzad98)}. However, in paediatric populations, the evidence has been less consistent and is limited by methodological heterogeneity. In their review, Suhett et al.^{(Reference Suhett, Hermsdorff and Cota17)} reported associations with adiposity, but their analyses were more limited in scope, and they did not perform quantitative data synthesis. Our review has built on this foundation by incorporating more recent studies, including those using the CDII, and providing quantitative estimates of effect. Furthermore, it stands out for examining the differences between the DII and CDII, demonstrating that associations observed for the DII are more consistent than those for the CDII.

In contrast, outcomes related to BP were less consistent and not significant for either the DII or CDII. This may suggest that the inflammatory effects of diet on haemodynamic parameters are not yet detectable in younger populations or are influenced by other external factors such as physical activity levels, sleep quality, stress or genetic predisposition^{(Reference Flynn, Kaelber and Baker-Smith44,Reference Kelishadi, Bahreynian and Qorbani99)} . It is also possible that variability in BP measurement methods, cut-offs and statistical adjustments may have hampered the detection of more robust associations.

Some limitations of this systematic review and meta-analysis should be acknowledged. Most of the included studies were cross-sectional, which limits causal inference. Although we observed associations, we could not confirm causality. Pro-inflammatory diets may contribute to weight gain, but it is equally plausible that individuals with excess weight adopt less healthy eating patterns, emphasising the need for longitudinal studies to clarify the direction and temporality of these associations^{(Reference Suhett, Hermsdorff and Cota17,Reference Ji, Hong and Chen100)} . In addition, we noted substantial heterogeneity in the dietary assessment methods, outcome definitions and statistical models used. This heterogeneity, combined with the moderate-to-low methodological quality of many studies, led us to downgrade the certainty of evidence according to the GRADE approach. Subgroup analyses and meta-regressions were not feasible owing to the small number of comparable studies per outcome. Publication bias could not be formally assessed.

Despite these limitations, this review has several strengths. We followed a pre-registered protocol in PROSPERO, adhered to the PRISMA reporting guidelines^{(Reference Page, McKenzie and Bossuyt33)} and applied rigorous methodological tools such as the NOS and the GRADE framework to evaluate the quality and certainty of the evidence. The search strategy was comprehensive, with no language or publication date restrictions, and included both the DII and CDII as dietary exposures.

These findings reinforce the importance of promoting anti-inflammatory dietary patterns from early life as a preventive strategy against overweight and cardiometabolic risk. Public health policies that aim to reduce the consumption of ultra-processed and pro-inflammatory foods, while encouraging balanced and nutrient-rich diets, may contribute to long-term cardiovascular health.

Future studies should prioritise longitudinal designs, standardised definitions of exposures and outcomes, and adequately powered samples. In addition, the CDII should be further validated and standardised in paediatric populations to improve comparability across studies. The development of standardised cut-off points and harmonised analytical approaches will also contribute to more robust and generalisable evidence in this field.