A literature search was conducted in the China National Knowledge Infrastructure (CNKI), Wanfang, VIP, PubMed, and Web of Science databases. The Chinese search terms included: (bone tumor OR osteosarcoma OR chondrosarcoma OR fibrosarcoma OR myeloma) AND (chemotherapy) AND (nutrition support OR dietary intervention OR enteral nutrition OR parenteral nutrition). The English search terms were: (“bone tumour” OR “osteosarcoma” OR “chondrosarcoma” OR “fibrosarcoma” OR “myeloma”) AND (“chemotherapy”) AND (“nutrition support” OR “dietary intervention” OR “enteral nutrition” OR “parenteral nutrition”).

① Primary outcomes: chemotherapy tolerability (hemoglobin) and complications. ② Secondary outcomes: prealbumin, incidence of malnutrition, length of hospital stay, body mass index (BMI), psychological state, quality of life, and self-management.

The inclusion criteria were as follows: studies had to be randomized controlled trials; participants were patients with pathologically confirmed bone tumors undergoing chemotherapy; and the intervention group received nutritional support (including enteral nutrition, parenteral nutrition, dietary guidance, or mixed nutritional support) during chemotherapy, while the control group received routine interventions.

Exclusion criteria: case reports and cohort studies; duplicate publications; studies with a sample size of less than 30; studies in which the observation group did not receive specified nutritional support; studies with incomplete or unextractable data; publications in languages other than Chinese or English; Literature that is inconsistent with the research topic and review articles.

In accordance with the established inclusion criteria, retrieved literature was first de-duplicated using Endnote literature management software. Subsequently, two researchers independently screened the titles and abstracts to exclude clearly ineligible articles. The full texts of potentially eligible articles were then reviewed to determine final inclusion. Data extraction was conducted using software such as Excel, covering basic information, intervention measures, outcome indicators, and corresponding data.

The quality of the included studies was systematically evaluated using the Cochrane risk-of-bias tool, encompassing the following domains: ① Random sequence generation: assessment of whether scientific randomization methods, such as random number tables or computer-generated sequences, were employed; ② Allocation concealment: analysis of whether group assignments were concealed using methods such as sealed envelopes or a centralized system; ③ Blinding: determination of the blinding status for participants, intervention providers, and outcome assessors; ④ Incomplete outcome data: verification of whether participant dropout or loss to follow-up compromised the integrity of the data analysis; ⑤ Selective reporting: examination of the consistency between the study protocol and the final reported outcome measures.

The risk level for each item was classified into three categories: low risk, indicated by a clear description of a standard procedure (e.g., use of a random number table); high risk, indicated by significant methodological flaws (e.g., pseudo-random allocation based on admission order); and unclear risk, where the publication provided insufficient information to permit a judgment.

Data analysis was performed using Review Manager 5.3. Dichotomous variables were evaluated using the risk ratio (RR) with its 95% confidence interval (CI), while continuous variables were evaluated using the standardized mean difference (SMD) with its 95% CI. A random-effects model was applied in cases of significant heterogeneity (P ≤ 0.10 or I² ≥ 50%); otherwise, a fixed-effect model was used. A P-value of <0.05 was considered to indicate statistical significance. Funnel plots were constructed to assess for publication bias. Supplementary note: Among the literature included in this study, the validated scales and scoring characteristics used to assess psychological status, quality of life, and self-management are as follows: Psychological status: Anxiety and depression-related scales are mainly used (such as Hamilton Anxiety Scale [HAMA], Hamilton Depression Scale [HAMD]). The scoring range is usually 0 - 56 points (for HAMA) or 0 - 53 points (for HAMD). A higher score indicates a more severe level of anxiety/depression (negative scoring) [2] [3] . Quality of life: The Quality of Life Score for Cancer Patients (QOL) is adopted, with a scoring range of 0 - 100 points. A higher score indicates a better quality of life (positive scoring) [4] [5] . Self-management ability: The Self-Management Ability Scale is used, with a scoring range of usually 0 - 40 points. A higher score indicates a stronger ability of patients to self-manage disease treatment, diet, and lifestyle (positive scoring) [3] [5] . All the above scales are mature tools with verified reliability and validity. Their scoring directions and ranges remain consistent across the included studies, ensuring the comparability of outcome indicators.

The literature search yielded 142 potentially eligible studies. After the removal of duplicates and exclusion of 131 articles upon full-text review, a final total of 11 studies were included in the analysis ( Figure 1 ). Of these 11 studies, 3 were assessed as having a high risk of bias because group allocation was not based on randomization principles. The remaining 8 were classified as having a low risk of bias for reporting explicit randomization methods ( Figure 2 ). These 8 low-risk studies complied with standardized operations in terms of random allocation methods, data integrity, and other dimensions. Their data results constitute the

main basis for the study’s conclusion that “nutritional support can significantly improve the treatment tolerance of bone tumor patients undergoing chemotherapy and reduce the risk of complications,” thus ensuring the reliability and scientificity of the analysis results.

Among the 11 included studies, one was published in English and ten in Chinese, comprising a total of 935 patients. The intervention for the observation group consisted primarily of nutritional support in six studies, while nutritional support was mentioned as part of the intervention in the remaining five. See Table 1 .

Note: ① Tolerability; ② Complications; ③ Prealbumin; ④ Incidence of malnutrition; ⑤ Length of hospital stay; ⑥ Body mass/BMI; ⑦ Psychological state; ⑧ Quality of life; ⑨ Self-management.

Three studies [8] [10] [12] reported on hemoglobin levels, encompassing a total of 160 patients (n = 80 in the observation group and n = 80 in the control group). Subgroups were divided by nutritional types, which are “routine care + nutritional support”, “dietary nursing”, and “enteral + parenteral nutrition vs enteral nutrition”. The meta-analysis revealed significant heterogeneity among these studies (I² = 81%, P = 0.005); therefore, a random-effects model was employed. The forest plot indicates that nutritional support improved hemoglobin levels in bone tumor patients during chemotherapy (MD = 5.70, 95% CI: 4.30 - 7.10, P < 0.00001), as shown in Figure 3 .

Five studies [4] [7] [8] [11] [12] reported on complications, involving a total of 418 patients (211 in the observation group and 207 in the control group). The meta-analysis revealed low heterogeneity among the five studies (I² = 0%, P = 0.85); therefore, a fixed-effects model was employed. The forest plot indicated that nutritional support significantly reduced the incidence of complications in bone tumor patients during chemotherapy (OR = 0.33, 95% CI: 0.20 - 0.55, P < 0.0001), as shown in Figure 4 .

Two studies [7] [12] reported on prealbumin levels, encompassing a total of 138 patients (71 in the observation group and 68 in the control group). The nutritional types are “combined nutrition vs. simple parenteral nutrition” and “routine care + nutritional support”. The meta-analysis revealed significant heterogeneity between the two studies (I² = 96%, P < 0.00001); therefore, a random-effects model was employed. The forest plot indicated that nutritional support improved prealbumin levels in bone tumor patients undergoing chemotherapy (MD = 26.76, 95% CI: 23.43 - 30.10, P < 0.00001) ( Figure 5 ).

Three studies [6] [9] [11] reported the incidence of malnutrition, encompassing a total of 407 patients (186 in the observation group and 221 in the control group). A meta-analysis revealed low heterogeneity among these studies (I² = 0%, P = 0.65); therefore, a fixed-effects model was employed. The forest plot ( Figure 6 ) indicated that nutritional support significantly reduced the incidence of malnutrition in bone tumor patients during chemotherapy (OR = 0.45, 95% CI: 0.29 - 0.72, P = 0.0007).

Three studies [7] [8] [11] reported data on the length of hospital stay for a total of 298 patients (151 in the observation group and 147 in the control group). Subgroups were divided by tumor types, which are “multiple myeloma”, “sacral tumor” and “osteosarcoma”. A meta-analysis revealed significant heterogeneity among the three studies (I² = 96%, P < 0.00001); therefore, a random-effects model was employed. The forest plot indicates that nutritional support was associated with a shorter length of hospital stay for patients with bone tumors (MD = −0.72, 95% CI: −1.10 - −0.34, P = 0.0002) ( Figure 7 ).

Two studies [6] [10] reported on BMI, encompassing a total of 128 patients (66 in the observation group and 62 in the control group). The nutritional types are “dietary nursing” and “kitchen food supplements vs. flavoring supplements”. The meta-analysis revealed significant heterogeneity between the two studies (I² = 95%, P < 0.0001); therefore, a random-effects model was employed. The forest plot indicated that nutritional support improved the BMI of patients with bone tumors (MD = 0.80, 95% CI: 0.05 - 1.55, P = 0.04), as shown in Figure 8 .

Two studies [2] [3] reported on psychological state scores, involving a total of 170 patients (85 in the observation group and 85 in the control group). All are aimed at patients with multiple myeloma, with consistent tumor types. The meta-analysis revealed significant heterogeneity between the two studies (I² = 95%, P < 0.0001); therefore, a random-effects model was employed. The forest plot indicated that nutritional support alleviated negative emotions in bone tumor patients undergoing chemotherapy (MD = 11.60, 95% CI: 10.16 - 13.04, P < 0.00001), as shown in Figure 9 .

Two studies [4] [5] reported quality of life scores, comprising a total of 130 patients (65 in the observation group and 65 in the control group). Subgroups divided by research quality. The meta-analysis revealed significant heterogeneity between the two studies (I² = 95%, P < 0.0001); therefore, a random-effects model was employed. The forest plot indicated that nutritional support improved the quality of life for bone tumor patients during chemotherapy (MD = 18.05, 95% CI: 14.10 - 21.99, P < 0.00001), as shown in Figure 10 .

Two studies [3] [5] reported quality of life scores, including a total of 160 patients (n = 80 in the observation group, n = 80 in the control group). Meta-analysis revealed significant heterogeneity between the two studies (I² = 83%, P = 0.02); therefore, a random-effects model was employed. The forest plot indicated that nutritional support improves the self-management ability of patients with bone tumors during chemotherapy (MD = 1.59, 95% CI: 1.28 - 1.90, P < 0.00001) (see Figure 11 ).

Funnel plot ( Figure 12 ) shows that, the comparison of complications between the two groups revealed an asymmetrical distribution of effect sizes. The effect size points were relatively dispersed, suggesting potential publication bias in the research results.