Cancer remains a significant public health concern, with an estimated 19·3 million new cases and 10·0 million cancer-related deaths in 2020(Reference Sung, Ferlay and Siegel1). Despite advancements in diagnostic techniques and therapeutic strategies, the long-term prognosis for patients with advanced cancer remains poor(Reference Siegel, Miller and Fuchs2). Therefore, there is an urgent need to enhance the prognostic assessment of cancer patients.
Malnutrition is a prevalent issue among cancer patients(Reference Sonmez, Tezcanli and Bas3). The European Society for Clinical Nutrition and Metabolism guidelines on nutrition strongly recommend screening the nutritional status of all cancer patients(Reference Arends, Bachmann and Baracos4). Malnutrition in cancer patients has been linked to increased postoperative complications, prolonged hospitalisation, reduced tolerance to treatment, worsened survival and lower quality of life(Reference Muscaritoli, Corsaro and Molfino5). Therefore, nutritional evaluation in such patients is of paramount importance.
Several screening and assessment tools have been developed to evaluate the nutritional status of cancer patients. However, there is no universally accepted standard for defining malnutrition in this population(Reference Mendes, Barros and Rosa6,Reference Molfino, Imbimbo and Laviano7) . Among these tools, the Nutritional Risk Screening-2002 and the Patient-Generated Subjective Global Assessment (PG-SGA) were the most commonly used for nutritional evaluation in adults with cancer(Reference Crestani, Grassi and Steemburgo8). The PG-SGA numerical scores have been used internationally as the reference method for risk screening, assessment, monitoring and triaging for interventions in patients with cancer(Reference Jager-Wittenaar and Ottery9). This tool includes both patient-reported (self-reported weight change, changes in food intake, presence of nutrition impact symptoms and activities and function) and clinician-assessed (scoring weight loss, physical examination, metabolic stress and disease and its relation to nutritional requirements) components. A higher PG-SGA score indicates a higher risk of malnutrition. Patients were classified as low risk (PG-SGA ≤ 3), medium risk (PG-SGA 4–8) and high risk of malnutrition (PG-SGA > 8). The prognostic significance of this nutritional tool has been widely studied in cancer patients(Reference Tan, Read and Phan10–Reference Findlay, White and Brown16). However, the existing studies have reported inconsistent findings regarding the association between the risk of malnutrition, as estimated by the PG-SGA numerical scores, and overall survival (OS)(Reference Rodrigues, Lacerda and Chaves17,Reference Barao, Abe Vicente Cavagnari and Silva Fucuta18) . Furthermore, conflicting results have been reported regarding the prognostic significance of medium risk of malnutrition in these patients(Reference Kim, Lee and Lim11,Reference Rodrigues, Lacerda and Chaves17,Reference Barao, Abe Vicente Cavagnari and Silva Fucuta18) . Therefore, we conducted this meta-analysis to evaluate the prognostic utility of malnutrition risk, as estimated by the PG-SGA numerical scores, in cancer patients.
Methods
Search strategy
The current systematic review/meta-analysis was reported in accordance with the guidelines of Preferred Reporting Items for Systematic Reviews and Meta-Analyses(Reference Liberati, Altman and Tetzlaff19). A systematic search was performed in multiple databases, including PubMed, Web of Science, Embase, CKNI, VIP, Sinomed and Wanfang databases through 28 March 2023, without any language restrictions. Two authors independently searched the English literature using the following keywords: ‘Patient-Generated Subjective Global Assessment’ OR ‘PG-SGA’ AND ‘cancer’ OR ‘tumor’ OR ‘malignancy’ OR ‘carcinoma’ OR ‘neoplasms’ AND ‘complication’ OR ‘survival’ OR ‘mortality’ OR ‘death’. For Chinese literature, the keywords used were: ‘Zhong liu’ AND ‘ai’ AND ‘huan zhe zhu guan zheng ti ping gu’ AND ‘sheng cun’ OR ‘si wang’ AND ‘bing fa zheng’. The detailed search strategy is presented in see online supplementary material, Supplemental Text S1. In addition, the reference lists of retrieved studies and pertinent reviews were manually searched to identify additional studies.
Study selection
Two authors independently selected studies based on the following criteria for inclusion: (1) population: adult patients diagnosed with cancer; (2) comparator: risk of malnutrition, as estimated using the PG-SGA numerical scores; (3) comparison: medium and high risk of malnutrition (PG-SGA score >4) v. low risk of malnutrition (PG-SGA score ≤3); (4) outcomes of interest: OS or postoperative complications defined by the Clavien–Dindo classification system; (5) type of study: either retrospective or prospective cohort and (6) reported a multivariable adjusted hazard ratio (HR) or OR with 95 % CI for the abovementioned outcomes. In cases where multiple publications were derived from the same population, only the study with the most comprehensive information was included. Articles from the same cohort but with specific type of cancer were included in subgroup analysis. The criteria for exclusion were (1) risk of malnutrition was estimated using other nutritional assessment tools; (2) lack of outcomes of interest; (3) reported of the unadjusted risk estimate; (4) not selecting the low risk of malnutrition (PG-SGA score ≤3) as the reference group; (5) overlapping participants with other articles and (6) inclusion of meeting abstracts, reviews or cross-sectional studies.
Data extraction and Quality assessment
Data extracted from the individual studies included: first author’s name, publication year, origin of patients, study design, cancer type, sample size, proportion of male participants, age at enrollment, assessing risk of malnutrition, risk of malnutrition prevalence, outcome measures, length of follow-up, fully adjusted relative risk and adjustment for variables. To assess the methodological quality of the included studies, a nine-point Newcastle-Ottawa Scale was used(Reference Wells, Shea and O’Connell20). The overall quality was categorised as low (<4 points), moderate (4-6 points) or high (≥7 points), respectively. Two independent authors performed data extraction and quality assessment. Any disagreements were resolved through consensus or discussion with the corresponding author.
Statistical analyses
All meta-analyses were undertaken using Stata 12·0 (Stata Corporation). For OS (time-to-event data), the prognostic value was expressed by pooling the adjusted HR with 95 % CI for the medium and high risk of malnutrition v. low risk of malnutrition group. The pooled adjusted OR with 95 % CI was used to summarise the association between risk of malnutrition with postoperative complications. Study heterogeneity was assessed using the I 2 statistic and Cochran’s Q test. An I2 statistics of <50 % and/or a P value >0·10 for the Cochran Q test indicated no significant heterogeneity, and a fixed-effect model was used for meta-analysis. If significant heterogeneity was present, a random-effects model was used. Sensitivity analysis was carried out by repeating the analyses after removing one study at a time. Subgroup analyses were undertaken based on study design (retrospective or prospective), cancer type (all types of cancer or gastrointestinal cancer or specific cancer), number of patients (≥500 or <500), age at enrollment (≥60 years or <60 years), geographical region (East Asia or other areas), degree of risk of malnutrition (medium or high) and length of follow-up (≥1 year or < 1 year). Publication bias was evaluated using the Begg’s test(Reference Begg and Mazumdar21) and Egger’s test(Reference Egger, Davey Smith and Schneider22). To investigate the potential influence of publication bias, a trim-and-fill analysis was performed.
Results
Search results and studies’ characteristics
Figure 1 summarises the process of study selection. Out of 1425 potentially relevant articles identified in the initial literature search, 627 remained after excluding duplicates. After evaluating the titles and abstracts, 562 articles were subsequently excluded. Sixty-four articles were retrieved for full-text assessment. After applying the predefined inclusion and exclusion criteria, nineteen studies reporting on twenty articles(Reference Tan, Read and Phan10–Reference Barao, Abe Vicente Cavagnari and Silva Fucuta18,Reference Mauricio, Xiao and Prado23–Reference da Silva Couto, Gonzalez and Martucci32) were finally included in this meta-analysis. Among these, Zhang(Reference Zhang, Tang and Fu27) and Ruan(Reference Ruan, Wang and Zhang29) reported on all types of cancer and a colorectal cancer subgroup from the same cohort.
Flow chart showing the process of study selection
The descriptive characteristics of the eligible studies are shown in Table 1. These studies were published from 2015 to 2023 and originated from Brazil, Chile, Australia, South Africa, France, Korea, Iran, Taiwan and China. Eight articles(Reference Tan, Read and Phan10,Reference Gallois, Artru and Lievre12,Reference Barao, Abe Vicente Cavagnari and Silva Fucuta18,Reference Mauricio, Xiao and Prado23,Reference Von Geldern, Salas and Alvayay26,Reference Nikniaz, Somi and Naghashi28,Reference de Sousa, Silva and de Carvalho30,Reference Argefa and Roets31) adopted the prospective designs, while the remaining articles used retrospective designs. Four articles(Reference Tan, Read and Phan10,Reference De Groot, Lee and Ackerie14,Reference Von Geldern, Salas and Alvayay26,Reference Zhang, Tang and Fu27) included all types of cancer, while the others focused on specific types such as oesophageal cancer, gastric cancer, colorectal cancer, hepatocellular carcinoma, gynaecologic cancer, nasopharyngeal carcinoma, oral cancer, head and neck cancer and multiple myeloma. The included studies enrolled a total of 9286 patients with cancer, with sample sizes ranging from 70 to 3547 cases. The prevalence of medium and high risk of malnutrition, as estimated by the PG-SGA numerical scores, varied between 16·0 %(Reference De Groot, Lee and Ackerie14) and 71·6 %(Reference Nikniaz, Somi and Naghashi28). The quality of the studies included is summarised in see online supplementary material, Supplemental Table S1. According to the Newcastle-Ottawa Scale criteria, two articles(Reference De Groot, Lee and Ackerie14,Reference Argefa and Roets31) were classified as moderate quality, while the rest were deemed to be of high quality.
Main characteristic of the included studies
P, prospective; OS, overall survival; R, retrospective; ISS, International Staging System; LDH, lactate hydrogenase; FISH, fluorescence in situ hybridization; ASCT, autologous stem cell transplantation; CRC, colorectal cancer; TNM, tumour node metastasis; HCC, hepatocellular carcinoma; NLR, neutrophil-to-lymphocyte ratio; NPC, nasopharyngeal carcinoma; HNC, head and neck cancer; CCI, Charlson Comorbidity Index; KPS, Karnofsky performance status; NLR, neutrophil-to-lymphocyte ratio.
* Results pooling from the sub-group using a fixed-effect model.
Overall survival
Fifteen studies(Reference Tan, Read and Phan10–Reference Barao, Abe Vicente Cavagnari and Silva Fucuta18,Reference Von Geldern, Salas and Alvayay26–Reference Nikniaz, Somi and Naghashi28,Reference de Sousa, Silva and de Carvalho30–Reference da Silva Couto, Gonzalez and Martucci32) examined the association between risk of malnutrition as measured by the PG-SGA and OS. As shown in Fig. 2, medium and high risk of malnutrition was associated with a significantly worse OS (HR 1·98; 95 % CI 1·77, 2·21) compared with those with low risk of malnutrition, without significant heterogeneity (I 2 = (Reference Zhang, Tang and Fu27,Reference Nikniaz, Somi and Naghashi28,Reference de Sousa, Silva and de Carvalho30–Reference da Silva Couto, Gonzalez and Martucci32) 32·9 %; P = 0·105). Sensitivity analysis demonstrated the credibility of the original risk summary. Sub-group analysis based on the degree of risk of malnutrition showed that the pooled HR of OS was 1·55 (95 % CI 1·17, 2·06) for medium risk of malnutrition and 2·65 (95 % CI 1·90, 3·70) for high risk of malnutrition, respectively (Fig. 3). Moreover, medium and high risk of malnutrition significantly predicted OS in each predefined sub-group (Table 2). However, Begg’s test (P = 0·023) and Egger’s test (P = 0·027) suggested the presence of publication bias. Despite this, the pooled HR for OS remained statistically significant (HR 1·88; 95 % CI 1·24, 2·84) after imputing three potentially missing studies using the trim-and-fill analysis (see online supplementary material, Fig. S1).
Pooled adjusted hazard ratio with 95 % CI of overall survival for medium and high risk of malnutrition v. those with low risk of malnutrition
Sub-group analysis on overall survival based on the medium (A) and high (B) risk of malnutrition, respectively
Results of sub-group analysis on overall survival
Postoperative complications
Three studies(Reference Mauricio, Xiao and Prado23–Reference Tsai, Lai and Huang25) examined the association between risk of malnutrition, as estimated by the PG-SGA, and postoperative complications. As shown in Fig. 4, medium and high risk of malnutrition was associated with an increased risk of postoperative complications (OR 4·65; 95 % CI 1·61, 13·44) compared with those with low risk of malnutrition, with significant heterogeneity (I 2 = 81·2 %; P = 0·005). Sensitivity analysis confirmed the robustness of the originally statistical significance of the pooled risk summary.
Pooled adjusted OR with 95 % CI of postoperative complications for medium and high risk of malnutrition v. those with low risk of malnutrition
Discussion
This systematic review and meta-analysis first evaluated the association between the risk of malnutrition, as estimated by the PG-SGA numerical scores, and adverse outcomes in cancer patients. Overall, the studies included in this analysis were of high methodological quality. Our meta-analysis revealed that the medium and high risk of malnutrition, as measured by the PG-SGA numerical scores, was significantly associated with poorer OS in cancer patients. Specifically, cancer patients with a medium to high risk of malnutrition had approximately twice the risk of reduced OS compared with those with a low risk of malnutrition. The association was even stronger in high-risk malnourished patients (HR 2·65) compared with medium-risk malnourished patients (HR 1·55). Further stratified analysis indicated that medium and high risk of malnutrition consistently correlated with poorer OS, irrespective of study design, cancer type, sample size, degree of malnutrition risk and length of follow-up.
In addition to OS, the risk of malnutrition, as measured by the PG-SGA numerical scores, was found to be linked to a higher risk of postoperative complications. According to our meta-analysis, cancer patients with medium and high risk of malnutrition had a 4·65-fold increased risk of postoperative complications. A randomized, single-blind clinical trial also demonstrated that medium risk of malnutrition was associated with a higher risk of complications in patients with head and neck cancer(Reference de Carvalho, Silva and Andre33). These complications can result in higher mortality and morbidity rates among cancer patients undergoing surgery. Furthermore, serious postoperative complications can also prolong hospital stays. These is evident in patients with risk of malnutrition and head and neck cancer(Reference Findlay, White and Brown16), colorectal cancer(Reference Karin, Bogut and Hojsak34) and gynecological cancer(Reference Laky, Janda and Kondalsamy-Chennakesavan35).
There is no consensus on which specific nutritional assessment tool best predicts survival outcomes in cancer patients. Several systematic reviews and meta-analyses have evaluated the value of malnutrition in predicting OS in cancer patients, including the Controlling Nutritional Status (CONUT) score(Reference Kheirouri and Alizadeh36), Prognostic Nutritional Index (PNI)(Reference Bullock, Greenley and McKenzie37), Geriatric Nutritional Risk Index (GNRI)(Reference Lv, An and Sun38) and Global Leadership Initiative on Malnutrition (GLIM)(Reference Xu, Jie and Sun39). Interestingly, the relative risk magnitude for OS was similar in GNRI (HR 1·95), PNI (HR 1·89) and GLIM (HR 1·90). This indicates that the risk of malnutrition, as estimated by the PG-SGA numerical scores, has similar prognostic power in patients with cancer. However, the prognostic value was stronger for PG-SGA-defined high risk of malnutrition (HR 2·65) in the current study compared with the previous GLIM-defined severe malnutrition (HR 1·68). One possible explanation for this finding may be the higher sensitivity and specificity of the PG-SGA numerical scores compared with the GLIM-defined malnutrition(Reference Ruan, Wang and Zhang29). It is important to note that these findings were based on indirect comparisons. Further research is needed to fully understand the prognostic significance of malnutrition in various types of cancer, and it may be beneficial to analyze data separately for each specific cancer type.
The Oncology Nutrition Dietetic Practice Group of the American Dietetic Association uses the PG-SGA as the standard for nutritional evaluation in cancer patients(Reference Bauer, Capra and Ferguson40). Compared with other nutritional assessment tools, the PG-SGA criteria enable a more objective evaluation of nutritional status and the identification of nutritional impact symptoms. Unlike other tools, the PG-SGA relies less on subjective responses from individuals. The PG-SGA numerical scores can indicate changes over time. A study found that for every point increase in PG-SGA score, there was a 4 % higher risk of death in cancer patients receiving a cachexia support service(Reference Bland, Zopf and Harrison41). In patients with nasopharyngeal carcinoma, a multivariate-adjusted Cox regression analysis showed that each point increase in PG-SGA score was associated with a 7 % decrease in OS(Reference Wang, Yang and Ge42). These findings further support the prognostic significance of the PG-SGA numerical scores in cancer patients.
The present study has important implications for clinical practice. The PG-SGA can serve as a promising nutritional screening tool and prognostic indicator of patients’ survival in patients with various types of cancer. For cancer patients at high risk of malnutrition, the PG-SGA numerical scores may provide more accurate prognostic information compared with other nutritional assessment tools. The clinical relevance of the PG-SGA numerical scores lies in its ability to identify patients who are at risk of malnutrition. By identifying and addressing nutritional challenges early, healthcare professionals can implement timely interventions to improve nutritional status and potentially enhance treatment outcomes. Furthermore, regular reassessment using the PG-SGA enables healthcare professionals to track changes in nutritional status and adjust interventions accordingly. However, further research is needed to explore the prognostic value of PG-SGA-defined the risk of malnutrition, particularly through separate analysis of primary cancer types.
Several limitations need to be mentioned in our study. First, the inclusion of retrospective studies in the meta-analysis may have been influenced by their inherent selection bias. Second, there was significant heterogeneity in certain sub-group analyses. This variation could potentially be attributed to differences in clinicopathologic characteristics, types of cancer, study design and follow-up intervals. Third, the results of Begg’s and Egger’s tests revealed the presence of publication bias. However, the trim-and-fill analysis showed that the prognostic value of PG-SGA-defined the risk of malnutrition may have been only slightly overestimated. Finally, this systematic review and meta-analysis has not been prospectively registered in PROSPERO or any other international databases prior to its publication.
Conclusions
This systematic review/meta-analysis provides evidence that medium and high risk of malnutrition, as estimated by the PG-SGA numerical scores, is significantly linked to poorer OS and an increased risk of postoperative complications in oncology patients. Evaluating the numerical scores of the PG-SGA numerical scores can offer crucial prognostic information for these patients.
Conflict of interest
The authors declare that they have no competing interests.
Financial support
This work was supported by (1) Medical Clinical Science and Technology Development Fund of Jiangsu University (JLY2021167); (2) Key Project Fund of Jiangsu Provincial Health Commission (ZD2022052, ZD2023016); (3) Zhenjiang Social Development Fund (SH2022032, SH2022090) and (4) Suqian Science and Technology Support Project Fund (S2022).
Authorship
Study conception/design and interpretation of data: Y.F. and H.Z.; literature search, data extraction, quality assessment and statistical analysis: J.F.Z. and W.X.; writing the manuscript: W.X.; revising the manuscript: Y.F. All authors approved the final manuscript.
Ethics of human subject participation
This study adopted the study-level data rather than individual participants; ethical approval was not required.
Supplementary material
For supplementary material accompanying this paper visit https://doi.org/10.1017/S1368980024000788
Open access
