Increasing evidence on the role of inflammation in carcinogenesis and cancer development(Reference Greten and Grivennikov1–Reference Diakos, Charles and Mcmillan4) has led to the proposition that treatments targeting deregulated inflammatory responses can be used as alternative strategies for cancer prevention and therapy(Reference Crusz and Balkwill3–Reference Hou, Karin and Sun5). Within this perspective, n-3 PUFA, as essential nutrients for normal metabolism, have attracted considerable interest in cancer-preventive and anticancer effects due to their potential role in suppressing and resolving inflammation(Reference Djuricic and Calder6–Reference D’Archivio, Scazzocchio and Vari10). Indeed, the use of n-3 PUFA for patients with cancer has been recommended by the European Society for Parenteral and Enteral Nutrition to patients with cancer, although the evidence is weak(Reference Muscaritoli, Arends and Bachmann11).
n-3 PUFA exert effects against various inflammatory conditions or disorders, including cancer(Reference Djuricic and Calder6–Reference D’Archivio, Scazzocchio and Vari10,Reference Yates, Calder and Ed12) . However, inconsistencies regarding the inhibitory effects of n-3 PUFA on systemic inflammation in patients with cancer have been found in the literature, weakening the potential use of n-3 PUFA in cancer prevention and treatment. Numerous inflammatory cytokines directly contribute to carcinogenesis, and most of them are largely confined to experimental research and have limited significance in clinical practice(Reference Mantovani, Allavena and Sica2,Reference Crusz and Balkwill3,Reference Lan, Chen and Wei13,Reference Grivennikov and Karin14) . Among them, IL-6 and TNF are the most extensively studied inflammatory cytokines in clinical studies to identify associations between systemic inflammation and cancer. However, findings on the evolution of circulating IL-6 and TNF levels after n-3 PUFA use in patients with cancer are inconsistent, including those of meta-analyses(Reference Mocellin, Camargo and Nunes15–Reference Pan, Zhou and Yin20). In view of the increasing benefits to cancer treatment, only mild side effects and no convincingly serious safety issues, determining the effects of n-3 PUFA administration on IL-6 and TNF levels in cancer patients has clinical importance(Reference Nabavi, Bilotto and Russo9).
Previous meta-analyses focused on digestive system cancer and included a small number of studies (two to seven per analysis) and sample size (only eighty-six patients in some studies)(Reference Mocellin, Camargo and Nunes15–Reference Pan, Zhou and Yin20). Thus, they are prone to selection and information bias. Moreover, the potential effects of relevant variables on IL-6 and TNF levels cannot be quantitatively identified by subgroup and meta-regression analyses because of the limited number of included studies. Since the publication of previous meta-analyses, several high-quality randomised controlled trials (RCT) have been conducted to explore the effects of n-3 PUFA on IL-6 and TNF levels but did not yield consistent results. Thus, a meta-analysis of RCT was conducted for the evaluation of the effects of n-3 PUFA administration on circulating IL-6 and TNF levels in patients with various types of cancer and potential impact of relevant variables, with particular concern to the optimal patients and regimens for which n-3 PUFA administration may be highly beneficial.
Methods and materials
This meta-analysis followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement(Reference Liberati, Altman and Tetzlaff21).
Literature search and selection
The PubMed and EMBASE were searched for RCT published from January 1975 to February 2021 and updated in September 2021. The search was limited to RCT that enrolled adult humans and had no language restrictions. The following search terms were included, with combined free text and subject terms: ‘inflammation’ or ‘inflammatory;’ ‘interleukin-6’ or ‘IL-6;’ ‘tumor necrosis factor or ‘TNF;’ ‘fatty acids’ or ‘alpha-linoleic acid (ALA)’ or ‘eicosapentaenoic acid (EPA)’ or ‘docosahexaenoic acid (DHA);’ and ‘cancer’ or ‘carcinoma’ or ‘neoplasm’ or ‘tumor’ or ‘tumour’ or ‘malignancy’. The reference lists of relevant publications were manually searched for additional studies.
The search followed the Patient, Intervention, Comparison, Outcome (PICO) strategy: (1) patient (P): patients with diagnosed cancer based on acceptable criteria; (2) intervention (I): n-3 PUFA administration (regardless of type and dose); (3) comparison (C): non-n-3 PUFA administration or placebo; and (4) outcome (O): circulating IL-6 and TNF levels.
If multiple studies reported outcomes on the same patient group, the one with the largest sample size was included. Abstracts, case reports, editorials, expert opinions, letters, animal studies and reviews without original data were excluded.
Data extraction and quality assessment
Two investigators independently extracted information from all eligible studies according to a standardised protocol. Disagreements were resolved through consensus with a third investigator. Data extracted from each study included the name of the first author, year of publication, nation, study design, sample size, patient inclusion criteria, cancer site, inflammatory markers, intervention, therapy duration, primary concurrent treatment, patient’s age and BMI. When only standard errors instead of standard deviations were provided in the study, standard deviation was calculated by multiplying the standard error by the square root of the sample size. In addition, some studies provided medians and ranges instead of means and standard deviation; the corresponding means and standard deviation were calculated with the method described by Hozo et al.(Reference Hozo, Djulbegovic and Hozo22)
The methodological quality of the RCT was evaluated by using the Jadad scale ranging from 0 to 7 points, including the following aspects: randomisation (0, 1 or 2), double-blinding (0, 1 or 2), concealment of allocation (0, 1 or 2), and withdrawals and dropouts (0 or 1)(Reference Jadad, Moore and Carroll23). A score of ≥ 4 indicates high quality(Reference Kjaergard, Villumsen and Gluud24). Two investigators rated each study independently and subsequently assigned a score to minimise selection bias. Disagreements were resolved through a consensus with a third investigator.
Statistical analysis
The meta-analysis was conducted using Stata statistical software (version 10.0, Stata Corporation). Given the diversity in the measurement and reporting of inflammatory markers among various laboratories, the computation of the summary estimates was used with a standardised mean difference (SMD) instead of the absolute levels of inflammatory markers. Heterogeneity across studies was tested using Q and I 2 statistics. Significant heterogeneity was indicated by P heterogeneity < 0·10 or I 2 > 60 %. A random-effects model was used when significant heterogeneity was observed; otherwise, a fixed-effect model was used to analyse the pooled results. Sensitivity analyses by changing the eligibility criteria, including the omission of one study at a time, were conducted to explore the robustness of the pooled results.
Subgroup analyses grouped by study design (double-blinded and non-double-blinded), sample size (≤ 40 and > 40), Jadad score (≤ 3 and ≥ 4), study area (America, Asia and Europe), cancer site (head and neck, gastric, colorectal, and other), fatty acid types (n-3 PUFA, DHA and EPA combined, and EPA), administration routes (oral, nasal and intravenous), therapy duration (≤ 1 month and > 1 month), primary concurrent treatment (chemotherapy, operation and others) and patient age (≤ 60 years and > 60 years) were performed using random-effects model to evaluate the effects of these variables on inflammation levels, as well as the possible sources of heterogeneity.
Meta-regression analyses (≥ 10 studies for each variable) were used to determine whether some relevant variables, including publication year, study design, Jadad score, sample size, study area, fatty acid types, administration routes, therapy duration, primary concurrent treatment, basic inflammation levels, patient age and BMI, were the possible sources of heterogeneity, as well as the existence of a linear relationship with inflammatory marker change.
Publication bias was evaluated visually by funnel plots and statistically by Egger’s and Begg’s tests. If significant publication bias was presented, the ‘trim-and-fill’ method was used to examine the expected number of studies needed to correct the asymmetry of funnel plots and compute the adjusted pooled result. Two-tailed P < 0·05 was considered statistically significant.
Results
Characteristics of included studies
A detailed flow chart of the study selection process is outlined in Fig. 1. A total of 115 potentially related articles were identified through an initial online search. After the titles and abstracts were reviewed, twenty-nine articles were selected and further examined. Nine of the twenty-nine articles were excluded for the following reasons: four had no original data(Reference Sunpaweravong, Puttawibul and Ruangsin25–Reference Gogos, Ginopoulos and Salsa28), three were not RCT(Reference Lu, Chen and Wei29–Reference Martinez, Herrera and Frias31), one did not use n-3 PUFA(Reference Mohammadzadeh, Faramarzi and Mahdavi32) and one was a study protocol(Reference Mcdonald, Bauer and Capra33). Therefore, twenty articles that satisfied the inclusion criteria were included in the meta-analysis: nineteen for IL-6(Reference Gianotti, Braga and Fortis34–Reference Haidari, Abiri and Iravani52) and fifteen for TNF(Reference Wu, Zhang and Wu36–Reference Ryan, Reynolds and Healy41,Reference Kanat, Cubukcu and Avci43,Reference Finocchiaro, Segre and Fadda44,Reference Wang, Zhang and Zhang46,Reference Golkhalkhali, Rajandram and Paliany48–Reference Mocellin, Pastore and Camargo53) . Table 1 provides the detailed characteristics of the included studies.
Flow diagram of study selection.
Characteristics of studies included in meta-analysis
RCT, randomised controlled trials; DB, double-blind; TPN, total parenteral nutrition; PO, postoperation; NR, not reported; EN, enteral nutrition; CT, chemotherapy; RT, radiotherapy; KPS, karnofsky performance status; ENT, ear, nose and throat.
* Treatment group.
† Control group.
Results for IL-6
The pooled result indicated a significant decrease in IL-6 level after n-3 PUFA administration and showed significant heterogeneity (Fig. 2). Sensitivity analysis by omitting one RCT at a time showed that SMD ranged from 0·378 (95 % CI 0·033, 0·753) to 0·568 (95 % CI 0·175, 0·961) when the studies of Wu 2001(Reference Wu, Zhang and Wu36) and Felekis 2010(Reference Felekis, Eleftheriadou and Papadakos39) were omitted. After five studies were removed, in which means and standard deviations were extracted by reading the graphs or calculating the medians and interquartile ranges(Reference Furukawa, Tashiro and Yamamori35,Reference Silva, Trindade and Fabre42,Reference Roca-Rodriguez, Garcia-Almeida and Lupianez-Perez45,Reference Carvalho, Cruz and Viana47,Reference Feijó, Rodrigues and Viana50) , a significant decrease in IL-6 was observed (SMD, 0·573; 95 % CI 0·092, 1·053) and the heterogeneity remained significant (P heterogeneity < 0·001; I 2 = 90·5 %). Additional analysis by including two non-RCT(Reference Lu, Chen and Wei29,Reference Wigmore, Fearon and Maingay30) showed a similar result in IL-6 after n-3 PUFA administration (SMD, 0·975; 95 % CI 0·386, 1·564) with significant heterogeneity (P heterogeneity < 0·001; I 2 = 94·7 %).
Meta-analysis of n-3 PUFA administration on IL-6 in cancer patients. CI, confidence interval; SMD, standardized mean difference.
Subgroup analyses suggested significant differences existed when grouping was performed according to study area, cancer site, fatty acid types and primary concurrent treatment (Table 2). No significant linear relationship between IL-6 and some relevant variables was observed through meta-regression analysis (Table 3). The possible sources of heterogeneity were not found by subgroup and meta-regression analyses (Tables 2 and 3).
The results of subgroup analyses for IL-6
SMD, standardised mean difference; DB, double-blind; NR, not reported.
The results of meta-regression analyses
NS, no statistics.
No publication bias was observed from funnel plot and associated statistics (P Begg = 0·484; P Egger = 0·319) (Fig. 4).
Results for TNF
A remarkable decrease in TNF was observed after n-3 PUFA administration with significant heterogeneity (Fig. 3). Sensitivity analysis by omitting one RCT at a time showed that SMD ranged from 0·642 (95% CI 0·414, 0·870) to 0·771 (95% CI 0·537, 1·005) when the RCT of Finocchiaro 2012(Reference Finocchiaro, Segre and Fadda44) and Felekis 2010(Reference Felekis, Eleftheriadou and Papadakos39) were omitted. A similar result was obtained after the study of Ryan 2009(Reference Ryan, Reynolds and Healy41) was removed, in which data were extracted by reading the graph (SMD, 0·745; 95% CI 0·480, 1·011). Pooled analysis by including one non-RCT(Reference Lu, Chen and Wei29) using random-effects model revealed similar decrease in TNF levels after n-3 PUFA administration (SMD, 0·840; 95 % CI 0·508, 1·172) with significant heterogeneity (P heterogeneity < 0·001; I 2 = 81·9%).
Meta-analysis of n-3 PUFA administration on TNF in cancer patients. CI, confidence interval; SMD, standardized mean difference.
The filled funnel plot for IL-6 (a) and TNF (b). Open circles are for original data, and solid squares are for imputed “filled” values.
Subgroup analyses revealed the differences in sample size, Jadad score and cancer site might affect n-3 PUFA efficacy (Table 4). The sources of heterogeneity were not found by subgroup (Table 4) and meta-regression (Table 3) analyses.
The results of subgroup analyses for TNF
SMD, standardized mean difference; DB, double-blind; NR, not reported.
The funnel plot (Fig. 4) and Begg’s (P = 0·038) and Egger’s test (P = 0·094) indicated the occurrence of publication bias for TNF. The ‘trim-and-fill’ method showed the need for five additional studies to correct the funnel plot asymmetry (Fig. 4). The SMD corrected using the fixed- and random-effects models were 0·469 (95 % CI 0·337, 0·601) and 0·467 (95 % CI 0·192, 0·742), respectively, which indicated the slight contribution of publication bias to the pooled results.
Discussion
This meta-analysis assessed the effects of n-3 PUFA administration on circulating IL-6 and TNF levels in patients with cancer. The results indicated that n-3 PUFA can reduce IL-6 and TNF levels. Sensitivity analyses by changing the eligibility criteria further strengthened the robustness of the results.
n-3 PUFA possess anti-inflammatory and inflammation-resolving activities, possibly related to the inhibition of IL-6 and TNF production. The inhibitory effects of n-3 PUFA on IL-6 and TNF have been reported in multiple inflammatory diseases. However, the results are inconsistent in patients with cancer. The current finding supported the role of n-3 PUFAs in reducing circulating IL-6 and TNF levels in patients with cancer. Partially consistent with current results, most previous meta-analyses reported significant decrease in IL-6(Reference Mocellin, Camargo and Nunes15–Reference Zhao and Wang18) and TNF(Reference Mocellin, Fernandes and Chagas16,Reference Zhao and Wang18,Reference Yan, Li and Yang19) , whereas few reported non-significant changes in IL-6(Reference Yan, Li and Yang19) and TNF(Reference Yu, Liu and Zhang17) after n-3 PUFA administration in digestive system cancer. Similar findings were observed in breast, lung and colorectal cancers examined in previous non-RCT(Reference Martinez, Herrera and Frias31,Reference Cerchietti, Navigante and Castro54–Reference Alfano, Imayama and Neuhouser56) . Favourable evidence were provided by in vitro studies. Findings showing that n-3 PUFA use can decrease IL-6 and TNF secretion were reported in various human(Reference Khalfoun, Thibault and Watier57–Reference Trebble, Arden and Stroud59) and other mammary cultured cells(Reference Feng, Wang and Yang60). Although not affected IL-6 and TNF secretion, n-3 PUFA can promote pro-resolving responses in human monocytes(Reference Jaudszus, Gruen and Watzl61). Based on current published literature, the effects of reducing IL-6 and TNF levels in patients with cancer should be regarded as convincing.
When the suppressive action of n-3 PUFA on inflammatory cytokines (IL-6 and TNF) in patients with cancer was established, the pros and cons of n-3 PUFA use should be weighed.
IL-6 is a multifaceted pleiotropic cytokine mainly produced by cancer and stromal cells in patients with cancer and has a wide range of target cells because of its trans-signalling mechanism. IL-6 has carcinogenic actions in experimental cancer models and patients with cancer(Reference Grivennikov and Karin14,Reference Rodriguez-Vita and Lawrence62–Reference Lacina, Brabek and Kral64) . Raised IL-6 levels indicate poor prognosis in patients with several types of cancer(Reference Crusz and Balkwill3). Anti-IL-6 therapy can target cancer by suppressing cancer growth, metastasis, metabolism and cachexia(Reference Crusz and Balkwill3,Reference Hou, Karin and Sun5,Reference Lacina, Brabek and Kral64) .
TNF is another key cytokine that associates inflammation with cancer. TNF has been initially found to have anticancer functions because of its capability to induce haemorrhagic necrosis in tumours. Existing data indicate that TNF is a poor apoptosis inducer with weak cytotoxic or cytostatic effects on malignant cells(Reference Lebrec, Ponce and Preston65). Only high-dose TNF administration can be used as a cytotoxic agent to kill tumour cells(Reference Lebrec, Ponce and Preston65,Reference Szlosarek, Charles and Balkwill66) . Moreover, TNF is a pro-cancer cytokine that favours tumour growth and metastasis(Reference Grivennikov and Karin14,Reference Lebrec, Ponce and Preston65) . Elevated circulating TNF concentration and expression are present in various pre-cancerous and malignant diseases. Chronic, low-level TNF exposure is linked to a pro-malignant phenotype (growth, invasion and metastasis)(Reference Szlosarek, Charles and Balkwill66).
Regardless of the underlying mechanism linking IL-6 and TNF with cancer, substantial evidence now exists to suggest that reductions in IL-6 and TNF levels are associated with the benefits of cancer treatment.
Conforming to the results of subgroup analyses, the anti-inflammatory action of n-3 PUFA may be subject to multiple factors. The disparity influence of race and ethnicity in various aspects of cancer, including treatment response, was observed(Reference Grenade, Phelps and Villalona-Calero67,Reference Shavers and Brown68) . Compared with their American and European counterparts, Asians can attain more benefits from n-3 PUFA administration. Additionally, differences in serum IL-6 and TNF levels were reported among people with different ethnic origins or regions(Reference Stowe, Peek and Cutchin69) and people with different serum n-3 PUFA levels(Reference Sekikawa, Steingrimsdottir and Ueshima70). Previous studies supported different effects of n-3 PUFA administration on C-reactive protein in diverse populations(Reference Pan, Zhou and Yin20). The disparate impacts of n-3 PUFA on circulating IL-6 and TNF in various regions were understandable, although no study directly compared the effects of n-3 PUFA administration on IL-6 and TNF levels in diverse populations.
A site-specific association between cancer and inflammation was reported(Reference Toiyama, Fujikawa and Koike71,Reference Guo, Pan and Du72) . The present analysis supported the diverse effects of cancer sites on IL-6 and TNF levels. Unexpectedly, an increasing trend in IL-6 and a borderline result for TNF were observed in head and neck cancer. Chronic inflammation involving IL-6 and TNF was related to the development and progression of head and neck cancer(Reference Bonomi, Patsias and Posner73). None of the included RCT reported remarkable increases in IL-6 and TNF in head and neck cancer(Reference Casas-Rodera, Gómez-Candela and Benítez38,Reference Felekis, Eleftheriadou and Papadakos39,Reference Roca-Rodriguez, Garcia-Almeida and Lupianez-Perez45,Reference Carvalho, Cruz and Viana47,Reference Solis-Martinez, Plasa-Carvalho and Phillips-Sixtos49) . According to current knowledge, the present results for IL-6 and TNF in head and neck cancer should be interpreted with caution. The possible reasons are as follows: firstly, the power to disclose these potential benefits of the therapy because of the small sample size is insufficient. Secondly, the possibility of low IL-6 and TNF levels in head and neck cancer minimises the extent of their reduction. Finally, substantial effects of n-3 PUFA are lacking.
Substantial differences between the anti-inflammatory effects of EPA and DHA were found(Reference Gorjao, Azevedo-Martins and Rodrigues74). The use of DHA alone was not reported in the included articles. A borderline result was observed for EPA use alone in IL-6. Alternatively, a substantial impact of TNF was observed. Subgroup analysis suggested that the combination of EPA and DHA have higher benefits than EPA alone. However, a recent network meta-analysis(Reference Vors, Allaire and Mejia75) and a head-to-head comparison study(Reference Allaire, Couture and Leclerc76) did not find remarkable differences in IL-6 and TNF levels between DHA and EPA. Results of in vitro studies were mixed(Reference Khalfoun, Thibault and Watier57,Reference Jaudszus, Gruen and Watzl61,Reference Weldon, Mullen and Loscher77,Reference So, Wu and Lichtenstein78) . Some were suggestive of more potent in EPA(Reference Khalfoun, Thibault and Watier57,Reference Sierra, Lara-Villoslada and Comalada79) , whereas others showed different results(Reference Jaudszus, Gruen and Watzl61,Reference Weldon, Mullen and Loscher77) . Additionally, no remarkable differences in some inflammation-related genes expressing in human immune cells were found between the effects of EPA and DHA(Reference Vors, Allaire and Marin80). Thus, whether EPA or DHA is superior to the other in terms of anti-inflammatory activity remains unclear.
Therapy or administrative duration may be one vital factor influencing the anti-inflammatory properties of n-3 PUFA. Previous data seemed to support significant decrease in IL-6 and TNF levels from longer duration of therapy(Reference Sunpaweravong, Puttawibul and Ruangsin25,Reference Ryan, Reynolds and Healy41,Reference Hershman, Unger and Crew81) . Inconsistent findings were provided by current subgroup analyses. A significant decrease in TNF was observed after long-term therapy, whereas a significant decrease in IL-6 was observed after short-term therapy. The optimal duration of n-3 PUFA was poorly determined and may have been influenced by multiple factors. Among the most critical factors, particular attention should be paid to the anti-inflammatory pathways of n-3 PUFA. Incorporation into cell membrane phospholipids can rapidly modify cell function, and 26 weeks are needed to alter the gene expression profiles to anti-inflammatory status in human blood mononuclear cells(Reference Bouwens, van de Rest and Dellschaft82). Additionally, n-3 PUFA can act directly on inflammatory cells by decreasing inflammatory cytokine production through the activation of free fatty acid receptors 1 and 4(Reference Kimura, Ichimura and Ohue-Kitano83) and enzymatically produce specialised pro-resolving mediators to orchestrate the resolution of inflammation(Reference Basil and Levy84). The precise anti-inflammatory pathways of n-3 PUFA are complex(Reference Djuricic and Calder6,Reference Calder8) and vary under various conditions, including doses and proportions(Reference Katan, Deslypere and van Birgelen85). Thus, varying the duration of n-3 PUFA administration for the desired effects is conceivable.
Doses and proportions are the other two fundamental factors influencing the anti-inflammatory properties of n-3 PUFA(Reference Zárate, Jaber Vazdekis and Tejera7,Reference Dasilva, Pazos and Garcia-Egido86) . The optimal doses have not yet been firmly established(Reference Calder8). Different inflammatory conditions possibly require different doses(Reference Calder, Albers and Antoine87). The threshold(Reference Kew, Banerjee and Minihane88) and dose–response relationship within a certain range(Reference Ras, Demonty and Zebregs89,Reference Flock, Skulas-Ray and Harris90) of n-3 PUFA administration have been reported. However, the dose–response relationship and diverse effects of different proportions of n-3 PUFA have not been analysed because of limited data.
Additionally, differences in administration routes, primary concurrent treatment, basic inflammatory factor levels and patient age cannot independently predict the effectiveness of n-3 PUFA administration on IL-6 and TNF levels according to the results of subgroup and meta-regression analyses.
Some limitations should be noted when interpreting the findings of this meta-analysis. Firstly, the number of patients and studies is small, and thus are prone to selection and publication biases. Secondly, substantial variations on potential confounders were present, such as patient enrolment, cancer site, n-3 PUFA types and dosage, and therapy duration. Thirdly, the subgroup results were defined post hoc, and the means of the studies instead of individual patient’s data were used as data points. Finally, the means and standard deviations in some studies were extracted through figures or calculated from data with non-normal distribution. These limitations may have reduced the statistical power, leading to false or spurious results.
Despite that the optimal regimens using n-3 PUFA were not identified, the present result supports the use of n-3 PUFA for patients with cancer, possibly excluding head and neck cancer, because of their anti-inflammatory properties. More benefits were observed in Asian, EPA and DHA combined, independent of administration routes, therapy duration and primary concurrent treatment. Further studies are needed to determine optimal patients and regimens that will highly benefit from the use of n-3 PUFA.
Acknowledgements
None.
This research received no specific grant from any funding agency, commercial or not-for-profit sectors.
X. X. and Y. G. conceived and designed the study; B. M., H. H. and Y. Z. selected the studies; X. L., N. L. and Y. G. extracted the data; X. X. and Y. G. performed the statistical analyses; X. X., B. M., X. L. and Y. G. wrote the manuscript. All authors read and revised the manuscript. The final manuscript was approved by all authors.
