Dynamic variation in a combined inflammation–tumor marker index during neoadjuvant chemotherapy and its value for organ-preservation decisions in gastric cancer

2.1 Study design and setting

We conducted a prospective observational cohort study at Ningxia Medical University General Hospital from January 2018 through December 2023. The primary objective was to determine whether the combined inflammation–tumor marker index (CITI) measured during neoadjuvant chemotherapy (NAC) predicts pathological response and can help identify candidates for organ preservation.

Informed consent: Informed consent has been obtained from all individuals included in this study.

Ethical approval: The research related to human use has been complied with all the relevant national regulations, institutional policies and in accordance with the tenets of the Helsinki Declaration, and has been approved by the Ethics Committee of Ningxia Medical University General Hospital

2.2 Patient selection

2.3 Neoadjuvant chemotherapy protocol

All enrolled patients received neoadjuvant chemotherapy (NAC) according to institutional protocols. Two primary regimens were used:

2.4 Combined inflammation–tumor marker index (CITI)

The combined inflammation–tumor marker index (CITI) was constructed as a composite integrating inflammatory markers – C-reactive protein (CRP), albumin (ALB), and the neutrophil-to-lymphocyte ratio (NLR) – and tumor markers – carcinoembryonic antigen (CEA) and carbohydrate antigen 19-9 (CA19-9).

Peripheral blood was collected at three prespecified time points: baseline (T0, within 1 week before initiating neoadjuvant chemotherapy [NAC]); mid-NAC (T1, at cycle 2 or 3, depending on regimen); and post-NAC (T2, within 2 weeks after completion of the last NAC cycle and before surgery). For each timepoint, CRP (mg/L), albumin (g/L), CEA (ng/mL), and CA19-9 (U/mL) were natural-log transformed to reduce right skew and then standardized to z-scores, and NLR was standardized directly (unitless). Each marker was combined to yield a CITI score. We derived weights using an independent retrospective pilot cohort to predict major pathologic response (MPR; Becker 1a–1b) after NAC. After the same preprocessing, we fit a multivariable logistic regression with predictors Z _NLR, Z _CRP, Z _Albumin, Z _CEA, Z _CA19-9. To guard against overfitting, coefficients were uniformly shrunk using bootstrap-estimated calibration slope. Final weights were the shrunk coefficients rescaled to a convenient magnitude (albumin retained a negative sign). Stability and performance were assessed via bootstrap resampling (1,000 iterations) to estimate optimism-corrected discrimination (AUC) and calibration (intercept/slope).

Dynamic changes in CITI were classified as significant decline (decrease ≥ 1 point), stable (±<0.5 points), or significant rise (increase ≥ 1 point).

CITI score was calculated by the following formula:

where:

Z _NLR is the standardized (z-scored) neutrophil-to-lymphocyte ratio (unitless).

Z _CRP is the standardized C-reactive protein concentration (mg/L).

Z _Albumin​ is the standardized albumin level (g/L).

Z _CEA is the standardized carcinoembryonic antigen level (ng/mL).

Z _CA19−9 is the standardized carbohydrate antigen 19-9 level (U/mL).

w ₁ to w ₅ are the weighting coefficients.

Each biomarker is first log-transformed or z-score normalized to mitigate skew (common with tumor markers) and to ensure all variables contribute comparably.

2.5 Surgery and pathologic assessment

All patients were initially scheduled for curative-intent resection – total or subtotal gastrectomy with D2 lymphadenectomy. The choice between total and subtotal gastrectomy was determined by tumor location, extent, and the surgeon’s discretion. For patients who did not proceed to surgery, the reason – disease progression, patient preference, or medical inoperability – was recorded prospectively.

Organ Preservation Criteria: All patients were planned for curative resection (total or subtotal gastrectomy with D2 lymphadenectomy). Post-NAC decisions were made in a multidisciplinary team including a surgical oncologist, medical oncologist, gastroenterologist, diagnostic radiologist, and GI pathologist. Candidates were considered only when all of the following were present after NAC:

1. Radiology: on contrast-enhanced CT, no new lesions, no radiologic progression, no suspicious nodes; gastric wall thickening consistent with post-treatment change was permissible.

2. Endoscopy: regression to a flat scar or minimal residual lesion without contact bleeding; targeted biopsies negative or showing only minute residual atypia without definite invasive carcinoma.

3. Multidisciplinary concordance: imaging and endoscopic/biopsy findings consistent with clinical complete or near-complete response, and tumor location judged technically amenable to endoscopic therapy or function-preserving resection.

4. Supportive biomarker signal: ΔCITI ≥ 1.0 from baseline to post-NAC served as supportive evidence but was not used as a sole determinant; discordant CITI dynamics led to intensified assessment rather than automatic exclusion.

If findings were discordant, patients were not offered non-operative watch-and-wait. The default recommendation was standard gastrectomy. Rarely, if all anatomic criteria were favorable and uncertainty was limited to biomarkers alone, a limited/function-preserving resection could be considered after additional targeted biopsies and MDT consensus.

Surveillance after organ-preserving management: Every 3 months in year 1, then every 4–6 months in year 2, with high-definition endoscopy and targeted biopsies of the scar or any suspicious area. CT chest/abdomen/pelvis every 3–4 months in year 1, then every 6 months in year 2. CEA, CA19-9, CRP, albumin, and NLR were obtained at each visit. CITI was re-calculated to track dynamics during surveillance. Any biopsy-proven regrowth, new/enlarging suspicious nodes, or consistent rising CITI coupled with new radiologic/endoscopic suspicion prompted expedited salvage gastrectomy.

2.6 Sample size calculation

Sample size and power. We used a two-sample comparison of proportions to estimate the sample size. We hypothesized that patients with a favorable decline in CITI would have a pathologic complete response (pCR) rate of approximately 35 % versus 15 % in those without a favorable decline. With a two-sided α = 0.05 and 80 % power, the required sample size was ∼74 patients per subgroup. Allowing for ∼10–15 % dropout or non-evaluable cases, we targeted a total enrollment of 160–180 patients. Ultimately, 179 patients were recruited within the prespecified timeframe.

2.7 Statistical analysis

Patient demographics, treatments, and outcomes were stored in a secure electronic database. Continuous variables were summarized as mean ± standard deviation (SD) or median [interquartile range, IQR], as appropriate; categorical variables as counts (percentages). Changes in CITI across time (baseline, mid-NAC, post-NAC) were assessed with paired t tests or Wilcoxon signed-rank tests, as appropriate. Between-group differences in CITI for responders versus non-responders were tested with unpaired t tests (normally distributed) or Mann–Whitney U tests (non-normal). For multiple pathological-response categories (pCR, near-pCR, partial, minimal/no response), one-way ANOVA or Kruskal–Wallis tests were used with multiplicity-adjusted post-hoc comparisons. Disease-free survival (DFS) and overall survival (OS) were estimated by the Kaplan–Meier method from the date of surgery, and survival curves for favorable versus unfavorable CITI groups were compared using log-rank tests. Multivariable Cox proportional-hazards models adjusted for baseline stage, histology, and other relevant covariates; results are reported as hazard ratios (HRs) with 95 % confidence intervals (CIs). All tests were two-sided, with α = 0.05 considered statistically significant.

Multiplicity and subgroup analyses: Subgroup comparisons were pre-specified as exploratory and reported with nominal, unadjusted p-values. Given the number of contrasts, there is an increased risk of type I error (false positives). We therefore emphasize effect sizes and 95 % CIs and interpret subgroup findings as hypothesis-generating pending confirmation in larger, independent datasets.

Analyses were performed using R (version 4.4) statistical software. Logistic regression (major pathologic response), Kaplan–Meier estimation, log-rank testing, Cox proportional-hazards modeling, and PH-assumption checks were implemented using the packages (stats, survival, boot, survminer, ggplot2, dplyr, and broom).

3.1 Patient enrollment and baseline characteristics

Between January 2018 and December 2023, 179 patients with cT2–T4 and/or cN+ gastric adenocarcinoma were enrolled. Table 1 summarizes baseline demographics and tumor characteristics. The mean age was 57.8 ± 9.3 years, and 67.0 % were male. Tumors were most commonly located proximally (33.5 %) or in the distal stomach (35.8 %). Most patients had clinical T3 disease (55.9 %), and 41.3 % presented with N2 lymph-node involvement. The median baseline CITI was 4.0 (IQR 3.2–5.2), indicating a moderate composite inflammatory–tumor marker burden at diagnosis.

3.2 Treatment delivery and compliance

All patients received neoadjuvant chemotherapy with either FLOT (5-fluorouracil, leucovorin, oxaliplatin, docetaxel) or SOX (S-1, oxaliplatin) (Table 2). Overall, 88.3 % completed the planned number of cycles. The most common grade 3–4 adverse events were neutropenia (23.5 %) and nausea/vomiting (11.2 %). Dose reductions were required in 17.9 % of patients, and 15.6 % experienced treatment delays longer than seven days. Eight patients (4.5 %) discontinued NAC owing to disease progression, and 10 (5.6 %) discontinued because of unacceptable toxicity.

3.3 CITI patterns during neoadjuvant chemotherapy

As summarized in Table 3, the median CITI declined from 4.0 (IQR 3.2–5.2) at baseline to 3.1 (IQR 2.4–4.1) post-NAC. Overall, 53.1 % of patients experienced a ≥1-point decrease in CITI from baseline to post-NAC, whereas 17.9 % showed a net increase. Patients with a ≥1-point decline were more likely to complete all planned chemotherapy cycles without major dose modifications. Conversely, those with stable or rising CITI more often had poorer treatment tolerance or disease progression during NAC.

3.4 Pathologic response and surgical outcomes

Following NAC, 164 of 179 patients (91.6 %) underwent surgical resection (Table 4). Total gastrectomy was performed in 58.3 % of cases, and 36.8 % underwent subtotal gastrectomy. An R0 resection (negative margins) was achieved in 90.8 % of surgical patients. A pathologic complete response (pCR) occurred in 16.6 %, and an additional 9.2 % had near-pCR. The median number of lymph nodes examined was 24 per specimen.

3.5 Organ-preservation considerations

A subset of 15/179 (8.4 %) met MDT criteria for potential organ preservation, 11/15 (73.3 %) ultimately underwent an organ-preserving approach (endoscopic resection, function-preserving gastrectomy, or non-operative surveillance) (Table 5). During a median 12-month follow-up (range 6–24), 2/11 (18.2 %) experienced local regrowth and were successfully salvaged surgically (Table 6). Given the small size of this subgroup and relatively short follow-up, these outcomes are presented as descriptive and should be interpreted cautiously; standard gastrectomy remains the default approach, with organ-preservation reserved for rigorously vetted cases under close surveillance.

Although baseline CITI was not significantly different among eventual responders versus non-responders (p = 0.20), the change in CITI from baseline to post-NAC was strongly correlated with pathologic outcomes (Table 6). Patients who achieved a pCR or near pCR demonstrated a mean CITI drop of −1.7 ± 1.2, compared to −0.7 ± 1.0 among those with partial or minimal/no response (p < 0.01). On multivariate logistic regression, a ≥1-point decrease in CITI emerged as an independent predictor for major pathologic response (OR = 2.76, 95 % CI 1.40–5.43; p = 0.02).

3.6 Survival outcomes

Table 7 illustrates OS and DFS for the entire cohort and stratified by favorable (≥1-point drop) versus unfavorable CITI changes. At a median follow-up of 24 months, the 3-year OS was 66.8 % for those with favorable CITI versus 52.7 % for unfavorable CITI (p = 0.04). Similarly, the 3-year DFS was 56.3 % vs 41.0 % (p = 0.06), respectively. A Cox proportional hazards model adjusted for baseline stage and histology confirmed that a significant decrease in CITI was associated with improved OS (HR = 0.60, 95 % CI 0.40–0.90; p = 0.01). In multivariable Cox models further adjusting for chemotherapy regimen (FLOT versus SOX) and baseline CITI, the association between a ≥1-point ΔCITI decline and overall survival remained independent (adjusted HR = 0.62, 95 % CI 0.43–0.91; p = 0.015), while disease-free survival showed a consistent, directionally favorable pattern (adjusted HR = 0.76, 95 % CI 0.55–1.06; p = 0.10). Baseline CITI was itself associated with outcomes (per-unit HR for OS = 1.12, 95 % CI 1.03–1.22; p = 0.008; per-unit HR for DFS = 1.09, 95 % CI 1.01–1.18; p = 0.028), and regimen did not materially alter inference (Supplementary Table S1).

3.7 Additional subgroup analyses

As summarized in Table 8, associations between CITI, NAC response, and outcomes varied by tumor location, histologic subtype, and regimen. Intestinal-type tumors had higher rates of major pathologic response (pCR + near-pCR) than diffuse-type tumors (31.6 % versus 15.6 %). Patients treated with FLOT exhibited a numerically higher pCR rate than those receiving SOX (18.3 % versus 8.5 %). Importantly, a baseline CITI < 4.0 was associated with a greater likelihood of achieving pCR or near-pCR (32.5 % versus 18.2 % for CITI ≥ 4.0) and with longer disease-free survival (DFS; log-rank p < 0.05). These findings highlight heterogeneity in treatment response across clinical subgroups and support individualized therapeutic strategies.