Baseline and Longitudinal Pan-Immune-Inflammation Value Compared with Other Inflammatory Indices in Patients Receiving Immune Checkpoint Inhibitors: A Retrospective Cohort Study

Abstract

Objectives

The pan-immune-inflammation value (PIV), calculated from neutrophil, platelet, monocyte, and lymphocyte counts, is a peripheral blood biomarker of systemic immune-inflammatory status. We evaluated the prognostic value of baseline PIV and its trajectory over the first six months of treatment for overall survival (OS) and progression-free survival (PFS) in patients receiving immune checkpoint inhibitors (ICIs) and benchmarked PIV against the neutrophil-to-lymphocyte ratio (NLR), platelet-to-lymphocyte ratio (PLR), and systemic immune-inflammation index (SII).

Methods

We retrospectively analyzed 205 patients treated at a single tertiary center for non-small-cell lung cancer, renal cell carcinoma, melanoma, or other solid tumors who received an ICI (90.2% nivolumab). PIV was measured at baseline, month 3, and month 6. The outcome-independent median value (587.7) defined the PIV-low and PIV-high groups. Cox regression (per doubling of log2-transformed PIV), Kaplan–Meier analysis, 3- and 6-month landmark models, time-updated Cox models, and restricted cubic splines were used, with adjustment for age, sex, and tumor type. Discrimination was assessed using Harrell’s C-index, 1000-resample bootstrap internal validation, and time-dependent area under the curve (AUC).

Results

The median follow-up was 50.6 months; 168 deaths and 190 progression-or-death events occurred. Each doubling of baseline PIV was independently associated with worse OS (adjusted HR, 1.24; 95% confidence interval [CI], 1.14–1.36; p<0.001) and PFS (HR, 1.18; 95% CI, 1.09–1.28; p<0.001). Median OS was 11.96 vs. 5.62 months, and median PFS was 6.21 vs. 3.45 months, in the PIV-low vs. PIV-high groups (log-rank p<0.001 for both). Month-3 and month-6 absolute PIV values remained prognostic in landmark analyses, whereas the relative change in PIV from baseline to month 3 was not associated with either outcome. The association was approximately log-linear (nonlinearity p>0.05 for OS and PFS). PIV, NLR, PLR, and SII showed similar discrimination (adjusted C-index for OS: 0.650, 0.647, 0.633, and 0.652, respectively); PIV had the lowest Akaike information criterion (AIC) and Bayesian information criterion (BIC) values among the four indices.

Conclusion

Baseline and on-treatment absolute PIV values, but not early relative changes in PIV, are independently associated with survival outcomes in patients receiving ICIs, with prognostic performance comparable to that of established inflammatory indices. PIV may serve as a simple, repeatable biomarker to support risk stratification during ICI therapy.

Immune checkpoint inhibitors (ICIs) have transformed the treatment of non-small-cell lung cancer (NSCLC), renal cell carcinoma (RCC), melanoma, and several other solid tumors, but clinical benefit remains highly heterogeneous, and validated, low-cost biomarkers to guide expectations and monitoring are still lacking. Peripheral blood indices that summarize the balance between innate immune activation and lymphocyte-mediated adaptive immunity, including the neutrophil-to-lymphocyte ratio (NLR), platelet-to-lymphocyte ratio (PLR), and systemic immune-inflammation index (SII), have been consistently associated with outcomes across tumor types and treatment modalities.

The pan-immune-inflammation value (PIV), calculated as (neutrophil count [×10³/mm³] × platelet count [×10³/mm³] × monocyte count [×10³/mm³]) / lymphocyte count (×10³/mm³), incorporates the monocyte compartment in addition to the components of SII and was first shown to be prognostic in metastatic colorectal cancer treated with first-line chemotherapy.^[1] Subsequent studies extended this observation to metastatic melanoma, including specifically under ICI therapy.^[2,3] Similar findings have been reported in HER2-positive breast cancer.^[4] PIV has also been evaluated in metastatic RCC treated with nivolumab, including in a Turkish Oncology Group Kidney Cancer Consortium study.^[5] Furthermore, it has been investigated in metastatic head and neck squamous cell carcinoma treated with ICIs.^[6] Pooled analyses across ICI-treated cohorts have confirmed that elevated baseline PIV is associated with shorter OS and PFS.^[7,8] However, most available evidence is limited to a single baseline measurement, and few studies have examined whether the trajectory of PIV during treatment, rather than its value at a single time point, carries independent prognostic information. A recent study in advanced NSCLC suggested that early on-treatment changes in PIV may be associated with both efficacy and immune-related toxicity.^[9] However, longitudinal data extending to 6 months and comparative benchmarking against other established indices within the same cohort remain scarce.

We therefore evaluated the independent prognostic value of baseline PIV for OS and PFS in a pan-cancer cohort of patients receiving ICIs, whether absolute PIV values at 3 and 6 months and relative early changes in PIV add prognostic information beyond baseline, and how PIV performs relative to NLR, PLR, and SII in terms of discrimination, model fit, and internal validity.

Methods

Study Design and Population

This was a single-center, retrospective cohort study conducted at the Department of Medical Oncology, Eskişehir Osmangazi University Faculty of Medicine, Türkiye. Consecutive adult patients (≥18 years) with a histologically or cytologically confirmed diagnosis of advanced NSCLC, RCC, melanoma, or another solid tumor who received at least one dose of an ICI at this center during the preceding 10 years were eligible if they had an available complete blood count within 7 days before ICI initiation, allowing calculation of baseline PIV, NLR, PLR, and SII, and at least one documented follow-up visit with survival status available. Patients were excluded if they had an active infection or another clinically significant acute inflammatory condition at the time of baseline blood sampling; a hematologic malignancy or disorder directly affecting peripheral blood cell counts; systemic corticosteroid use (>10 mg/day prednisone equivalent) or other immunosuppressive/anti-inflammatory therapy within the preceding 4 weeks; active autoimmune disease requiring systemic immunosuppression; missing baseline complete blood count or insufficient follow-up data; or receipt of ICI exclusively in the neoadjuvant or adjuvant setting without an evaluable advanced/metastatic-phase outcome. A total of 205 patients met all eligibility criteria and were included in the final analysis.

Biomarker Definitions

PIV was calculated as [neutrophil count (×10³/mm³) × platelet count (×10³/mm³) × monocyte count (×10³/mm³)] / lymphocyte count (×10³/mm³) using complete blood counts obtained within 7 days before treatment initiation (T0) and at approximately month 3 (T3) and month 6 (T6) of treatment. NLR [neutrophil count (×10³/mm³) / lymphocyte count (×10³/mm³)], PLR [platelet count (×10³/mm³) / lymphocyte count (×10³/mm³)], and SII [platelet count (×10³/mm³) × neutrophil count (×10³/mm³) / lymphocyte count (×10³/mm³)] were calculated from the same baseline samples. The baseline PIV cutoff (587.7) was defined as the outcome-independent cohort median, dichotomizing patients into PIV-low and PIV-high groups. Line of therapy (first-line vs. later-line ICI) was additionally recorded when documented in the medical records and was available for 58 patients (28.3%).

Outcomes

OS and PFS were the primary endpoints; follow-up was estimated using the reverse Kaplan–Meier method.

Statistical Analysis

Continuous PIV was log2-transformed, and hazard ratios (HRs) are reported per doubling of PIV. Univariable and multivariable Cox regression models, adjusted for age, sex, and tumor type, were used for the primary analysis; sensitivity analyses additionally adjusted for Eastern Cooperative Oncology Group (ECOG) performance status and albumin. Proportional hazards assumptions were tested for both the continuous and dichotomous (PIV-high vs. PIV-low) parameterizations; because violations were detected only in the dichotomous models, continuous log2-PIV was retained as the primary exposure, with dichotomous Kaplan–Meier and Cox estimates reported as supportive analyses.

Longitudinal PIV was evaluated using 3- and 6-month landmark Cox models, restricted to patients who were event-free at the landmark time, and time-updated Cox models. In the available-case time-updated analysis, follow-up was divided into nominal baseline-to-month-3, month-3-to-month-6, and post-month-6 intervals and was censored at the first scheduled time point without an available PIV measurement; thus, all patients contributed baseline risk time, but events occurring after such censoring were not included. In a last-observation-carried-forward (LOCF) sensitivity analysis, the most recent PIV value was carried forward to retain complete endpoint follow-up. Relative change in PIV from baseline to month 3 (ΔPIV) was modeled as the ratio of month-3 to baseline PIV. The shape of the association between baseline PIV and outcomes was examined using restricted cubic splines with a formal test for nonlinearity.

PIV, NLR, PLR, and SII were compared using identical log2-transformed Cox model specifications; because the indices were highly correlated (Spearman ρ=0.71–0.94), they were entered into separate rather than simultaneous models. Model discrimination was assessed using Harrell’s C-index, model fit using the Akaike information criterion (AIC) and Bayesian information criterion (BIC), and incremental value over a clinical model (age, sex, and tumor type) using the likelihood-ratio test. Internal validation used 1000-resample bootstrapping to obtain optimism-corrected C-indices. Prespecified subgroup analyses (tumor type; nivolumab vs. other ICIs) included formal interaction tests. Time-dependent, inverse-probability-of-censoring-weighted AUCs were calculated at 6, 12, and 24 months.

All tests were two-sided; p<0.05 was considered statistically significant. Analyses were performed using R software (version 4.3.2; R Foundation for Statistical Computing, Vienna, Austria).

Ethical Considerations

This study was approved by the Eskişehir Osmangazi University Faculty of Medicine Clinical Research Ethics Committee (Decision No. 25; January 18, 2022). The study was conducted in accordance with the Declaration of Helsinki.

Results

Study Population

A total of 205 patients were included (median age, 63 years [IQR, 56–68]; 76.6% male): 134 (65.4%) had NSCLC, 35 (17.1%) had RCC, 27 (13.2%) had melanoma, and 9 (4.4%) had another solid tumor; 90.2% received nivolumab. Median follow-up, estimated using the reverse Kaplan–Meier method, was 50.6 months; 168 deaths (82.0%) and 190 progression-or-death events (92.7%) occurred. Median OS was 8.57 months, and median PFS was 4.67 months. Median baseline PIV was 587.7 (IQR, 308.4–1392.6).

Using the outcome-independent median, 103 patients were classified as PIV-low and 102 as PIV-high. Age, sex, ECOG performance status, ICI regimen, and concurrent chemotherapy were balanced between the groups; tumor-type distribution differed significantly (p=0.001), and higher PIV was accompanied by lower albumin and higher lactate dehydrogenase (Table 1).

Table 1.

Baseline characteristics according to baseline PIV status

Characteristic	Level	Overall (n=205)	PIV-low (n=103)	PIV-high (n=102)	|SMD|	p
Age, years	Median (IQR)	63.0 (56.0–68.0)	63.0 (55.0–69.0)	62.5 (57.0–68.0)	0.04	0.764
Sex	Female	48 (23.4%)	29 (28.2%)	19 (18.6%)	0.23	0.138
	Male	157 (76.6%)	74 (71.8%)	83 (81.4%)	0.23
Tumor type	Melanoma	27 (13.2%)	16 (15.5%)	11 (10.8%)	0.14	0.001
	NSCLC	134 (65.4%)	57 (55.3%)	77 (75.5%)	0.43
	RCC	35 (17.1%)	27 (26.2%)	8 (7.8%)	0.50
	Other/SCLC	9 (4.4%)	3 (2.9%)	6 (5.9%)	0.15
ECOG PS	ECOG 0	88 (42.9%)	42 (40.8%)	46 (45.1%)	0.09	0.434
	ECOG 1	73 (35.6%)	35 (34.0%)	38 (37.3%)	0.07
	Missing	44 (21.5%)	26 (25.2%)	18 (17.6%)	0.19
ICI regimen	Other ICI	20 (9.8%)	8 (7.8%)	12 (11.8%)	0.13	0.357
	Nivolumab	185 (90.2%)	95 (92.2%)	90 (88.2%)	0.13
Concurrent chemo	No	196 (95.6%)	101 (98.1%)	95 (93.1%)	0.24	0.101
	Yes	9 (4.4%)	2 (1.9%)	7 (6.9%)	0.24
Hemoglobin	Median (IQR)	12.3 (10.6–13.6)	12.3 (10.7–13.8)	12.1 (10.5–13.3)	0.18	0.272
Albumin	Median (IQR)	4.0 (3.7–4.3)	4.1 (3.8–4.4)	3.8 (3.5–4.1)	0.53	<0.001
LDH	Median (IQR)	203 (171–256)	196 (164–231)	218 (180–302)	0.45	0.008
Baseline PIV	Median (IQR)	588 (308–1393)	308 (178–426)	1419 (906–2383)	2.42	<0.001

Continuous variables compared with the Wilcoxon rank-sum test, categorical variables with Fisher’s exact test. Absolute SMD ≥0.10 indicates potential imbalance. ECOG missing in 44 patients. PIV: Pan-immune-inflammation value; SMD: Standardized mean difference; ICI: Immune checkpoint inhibitor; LDH: Lactate dehydrogenase. ECOG PS: Eastern Cooperative Oncology Group performance status; IQR: Interquartile range; NSCLC: Non-small-cell lung cancer; RCC: Renal cell carcinoma; SCLC: Small-cell lung cancer.

Baseline PIV and Survival

Each doubling of baseline PIV was associated with a 25% increase in the hazard of death (univariable HR, 1.25; 95% CI, 1.15–1.36; p<0.001) and a 20% increase in the hazard of progression or death (HR, 1.20; 95% CI, 1.11–1.30; p<0.001). These associations remained significant after adjustment for age, sex, and tumor type (OS: HR, 1.24; 95% CI, 1.14–1.36; PFS: HR, 1.18; 95% CI, 1.09–1.28; both p<0.001) and were consistent in ECOG- and albumin-adjusted sensitivity analyses (Table 2). Median OS was 11.96 vs. 5.62 months, and median PFS was 6.21 vs. 3.45 months, in the PIV-low vs. PIV-high groups (log-rank p<0.001 for both; Fig. 1). In dichotomous models, high PIV was associated with worse OS (HR, 1.81; 95% CI, 1.31–2.50) and PFS (HR, 1.50; 95% CI, 1.11–2.03), although these estimates were considered supportive given the detected proportional hazards violation (Fig. 2).

Table 2.

Primary and sensitivity Cox regression analyses for baseline PIV

Outcome	Analysis	n/events	HR	95% CI	p	C-index
OS	Univariable	205/168	1.25	1.15–1.36	<0.001	0.636
OS	Primary adjusted (age, sex, tumor type)	205/168	1.24	1.14–1.36	<0.001	0.650
OS	ECOG-adjusted	161/135	1.25	1.11–1.41	<0.001	0.638
OS	Albumin-adjusted	189/154	1.25	1.14–1.37	<0.001	0.684
OS	High vs low PIV, adjusted	205/168	1.81	1.31–2.50	<0.001	0.637
PFS	Univariable	205/190	1.20	1.11–1.30	<0.001	0.620
PFS	Primary adjusted (age, sex, tumor type)	205/190	1.18	1.09–1.28	<0.001	0.645
PFS	ECOG-adjusted	161/151	1.20	1.06–1.35	0.004	0.632
PFS	Albumin-adjusted	189/175	1.18	1.09–1.29	<0.001	0.656
PFS	High vs low PIV, adjusted	205/190	1.50	1.11–2.03	0.008	0.623

HRs per doubling of PIV unless otherwise indicated. ECOG model used complete cases (n=161); albumin model used n=189. PIV components were not entered simultaneously due to collinearity. CI: Confidence interval; C-index: Harrell concordance index; ECOG: Eastern Cooperative Oncology Group; HR: Hazard ratio; OS: Overall survival; PFS: Progression-free survival; PIV: Pan-immune-inflammation value.

Figure 1.

Kaplan–Meier curves for OS and PFS according to the outcome-independent median baseline PIV threshold (587.7). Shaded areas represent 95% CI; risk tables are displayed below each panel.

Figure 2.

Adjusted association of baseline PIV with OS and PFS across the prespecified multivariable and sensitivity models. Hazard ratios are reported per doubling of PIV.

Longitudinal PIV

Absolute PIV retained prognostic value throughout treatment. Each doubling of month-3 PIV was associated with shorter OS (adjusted HR, 1.46; 95% CI, 1.26–1.70; p<0.001) and PFS (HR, 1.21; 95% CI, 1.03–1.41; p=0.019); month-6 PIV showed similar associations (OS: HR, 1.33; 95% CI, 1.09–1.64; p=0.006; PFS: HR, 1.23; 95% CI, 1.02–1.50; p=0.032). In contrast, relative change in PIV from baseline to month 3 (ΔPIV) was not associated with OS (HR, 1.04; 95% CI, 0.92–1.18; p=0.535) or PFS (HR, 0.99; 95% CI, 0.88–1.11; p=0.835). The available-case time-updated analysis retained 123 deaths and 174 PFS events and showed associations with OS (HR, 1.41; 95% CI, 1.24–1.62; p<0.001) and PFS (HR, 1.30; 95% CI, 1.17–1.45; p<0.001). Corresponding LOCF estimates, based on all 168 deaths and 190 PFS events, were HR, 1.38 (95% CI, 1.23–1.55; p<0.001) for OS and HR, 1.28 (95% CI, 1.16–1.42; p<0.001) for PFS (Table 3; Fig. 3).

Table 3.

Baseline, landmark, and time-updated PIV analyses

Outcome	Analysis	n/events	HR	95% CI	p	C-index
OS	Baseline PIV, adjusted	205/168	1.24	1.14–1.36	<0.001	0.650
OS	Month-3 PIV, 3-month landmark	148/113	1.46	1.26–1.70	<0.001	0.681
OS	Month-6 PIV, 6-month landmark	74/55	1.33	1.09–1.64	0.006	0.686
OS	T0-to-T3 ΔPIV, 3-month landmark	148/113	1.04	0.92–1.18	0.535	0.607
OS	Time-updated PIV, available-case intervals	205/123	1.41	1.24–1.62	<0.001	—
OS	Time-updated PIV, LOCF sensitivity	205/168	1.38	1.23–1.55	<0.001	—
PFS	Baseline PIV, adjusted	205/190	1.18	1.09–1.28	<0.001	0.645
PFS	Month-3 PIV, 3-month landmark	130/116	1.21	1.03–1.41	0.019	0.661
PFS	Month-6 PIV, 6-month landmark	70/60	1.23	1.02–1.50	0.032	0.647
PFS	T0-to-T3 ΔPIV, 3-month landmark	130/116	0.99	0.88–1.11	0.835	0.635
PFS	Time-updated PIV, available-case intervals	205/174	1.30	1.17–1.45	<0.001	—
PFS	Time-updated PIV, LOCF sensitivity	205/190	1.28	1.16–1.42	<0.001	—

HRs are reported per doubling of absolute PIV; ΔPIV HRs are reported per doubling of the month-3/baseline ratio. Landmark models were adjusted for age, sex, and tumor type. In the available-case time-updated analysis, all 205 patients contributed baseline risk time, but follow-up was censored at the first scheduled time point without an available PIV measurement; therefore, 123 deaths and 174 PFS events were retained. The LOCF sensitivity analysis carried the most recent PIV value forward and retained all 168 deaths and 190 PFS events. Time-updated analyses were adjusted for age, sex, and tumor type and should be interpreted as supportive.

CI: Confidence interval; C-index: Harrell concordance index; HR: Hazard ratio; LOCF: Last observation carried forward; OS: Overall survival; PFS: Progression-free survival; PIV: Pan-immune-inflammation value; T0: Baseline; T3: Month 3; ΔPIV: Relative change in PIV from baseline to month 3.

Figure 3.

Longitudinal association of PIV with OS and PFS. Absolute PIV at baseline, month 3 and month 6 was prognostic, whereas relative baseline-to-month-3 change was not.

Dose–Response Relationship

Restricted cubic spline analyses showed no significant evidence of nonlinearity for OS (p=0.079) or PFS (p=0.129), supporting an approximately log-linear association between baseline PIV and adverse outcomes across the observed range (Fig. 4).

Figure 4.

Adjusted restricted cubic spline associations of baseline PIV with OS and PFS. The median PIV was used as the reference; shaded areas represent 95% CI, and rugs show the observed PIV distribution.

Comparison With Other Inflammatory Indices

PIV was strongly correlated with NLR (ρ=0.814), PLR (ρ=0.713), and SII (ρ=0.943). All four indices were independently associated with OS and PFS, with modest differences in discrimination. For OS, adjusted C-indices were 0.650 (PIV), 0.647 (NLR), 0.633 (PLR), and 0.652 (SII); for PFS, the corresponding values were 0.645, 0.647, 0.630, and 0.648, respectively. PIV had the lowest AIC and BIC values for both outcomes, with SII providing a similar model fit; each index significantly improved model fit over the clinical model comprising age, sex, and tumor type. Bootstrap internal validation (1000 resamples) showed limited optimism, with corrected OS C-indices of 0.633 (PIV), 0.629 (NLR), 0.614 (PLR), and 0.634 (SII) (Suppl Table S1).

Subgroup and Treatment-Interaction Analyses

The prognostic effect of PIV was consistent across the NSCLC, RCC, and melanoma subgroups, with no significant tumor-type interaction for OS (p=0.156) or PFS (p=0.084). A stronger association between PIV and survival was observed among the 20 patients receiving a non-nivolumab ICI compared with the 185 patients receiving nivolumab (OS interaction p=0.003; PFS interaction p=0.004); given the small non-nivolumab subgroup, this finding is exploratory (Suppl Table S2; Fig. 5).

Figure 5.

Adjusted restricted cubic spline associations of baseline PIV with OS and PFS. The median PIV was used as the reference; shaded areas represent 95% CI, and rugs show the observed PIV distribution.

Time-Dependent Discrimination

Biomarker-enhanced models consistently outperformed the clinical model alone. At 6 months, OS AUCs were 0.715 (PIV), 0.708 (NLR), 0.690 (PLR), and 0.725 (SII); corresponding PFS AUCs were 0.731, 0.734, 0.730, and 0.738. At 24 months, PIV yielded the highest OS (0.785) and PFS (0.807) AUCs, although these late estimates were less stable because only 31 OS controls and 17 PFS controls remained (Suppl Table S3; Fig. 6).

Figure 6.

Time-dependent dynamic AUCs at 6, 12, and 24 months for the clinical model and models additionally containing PIV, NLR, PLR, or SII. Estimates were obtained using inverse-probability-of-censoring weights; 31 OS controls and 17 PFS controls remained at 24 months.

Discussion

In this cohort of 205 patients receiving ICIs across NSCLC, RCC, melanoma, and other solid tumors, baseline PIV was independently and consistently associated with both OS and PFS, with an approximately log-linear dose–response relationship. These findings are consistent with prior reports linking elevated PIV to worse outcomes with ICI therapy in melanoma.^[2,3] Similar associations have been reported in RCC.^[5] PIV has also been linked to outcomes in head and neck squamous cell carcinoma.^[6] Comparable findings have been reported in colorectal cancer.^[1] Pooled multi-tumor analyses have likewise demonstrated adverse prognostic associations with elevated PIV.^[7,8] The magnitude of the association observed in our study (adjusted HR per doubling of PIV: 1.24 for OS and 1.18 for PFS) is broadly comparable to pooled estimates from meta-analyses of PIV in ICI-treated populations.^[7,8]

The principal contribution of this study is the longitudinal analysis extending to 6 months. Absolute PIV measured during treatment at months 3 and 6 remained independently prognostic, and time-updated modeling confirmed that current PIV, rather than only its baseline value, was associated with risk throughout follow-up. This finding is consistent with a small number of prior reports suggesting that on-treatment PIV dynamics may provide additional information in patients with NSCLC treated with ICIs.^[9] Notably, however, the relative early change in PIV from baseline to month 3 was not associated with either outcome in our cohort. This dissociation between absolute PIV levels and relative changes suggests that PIV may function more as a marker of a patient’s underlying immune-inflammatory state at any given time than as a treatment-response biomarker whose early trajectory predicts benefit. This distinction has direct implications for how PIV might be used clinically, for example, for risk stratification at each restaging visit rather than as an early on-treatment response signal.

When benchmarked against NLR, PLR, and SII, PIV showed comparable but not clearly superior discrimination. SII performed similarly across most metrics, consistent with the substantial overlap between the two indices (ρ=0.943), given their shared neutrophil and platelet components. PIV did have the lowest AIC and BIC values among the four indices and the highest time-dependent AUC at 24 months, suggesting a possible advantage in overall model fit and longer-term discrimination. However, these differences were modest and should be interpreted cautiously given the correlated nature of the indices and the exploratory nature of the late time-point estimates.

The exploratory treatment-interaction finding, namely a numerically stronger PIV–survival association among the 20 patients receiving a non-nivolumab ICI, should not be overinterpreted given the small subgroup size and warrants confirmation in larger cohorts with greater diversity in ICI agents before any agent-specific inference can be drawn.

Limitations

This was a retrospective, single-center study without external validation, which limits generalizability. Proportional hazards violations in the dichotomized PIV models mean that dichotomous cutoff-based estimates should be interpreted as supportive rather than primary. Treatment line was recorded in only 58 patients (28.3%), and PD-L1 expression was not available; consequently, these variables could not be incorporated into full-cohort multivariable models, and residual confounding by treatment selection and tumor biology cannot be excluded. Because PIV, NLR, PLR, and SII share overlapping hematologic components, they were not entered simultaneously into the same model. Landmark analyses involved smaller event-free cohorts at later time points, and the available-case and LOCF time-updated analyses depend on assumptions regarding missing on-treatment PIV measurements. Residual confounding by other unmeasured inflammatory or nutritional factors also remains possible.

Strengths

Strengths include the longitudinal design with three prespecified PIV measurement time points, a prespecified outcome-independent cutoff, rigorous assessment of proportional hazards assumptions, head-to-head comparison with three established inflammatory indices in the same patients, and internal validation using bootstrap resampling and time-dependent AUC methodology.

Conclusion

Baseline and on-treatment absolute PIV values are independently associated with OS and PFS in patients receiving ICIs, with a log-linear dose–response relationship and prognostic performance comparable to that of NLR, PLR, and SII. Early relative changes in PIV do not appear to add prognostic information. If externally validated, PIV, as a simple, low-cost, repeatable blood-based index, could support ongoing risk stratification throughout ICI treatment, complementing rather than replacing existing inflammatory biomarkers.

Disclosures

Ethics Approval

Ethics committee approval for this study was obtained from the Eskişehir Osmangazi University (Approval no: 25; date: January 18, 2022)

Conflict of Interest

The authors declare that there is no conflict of interest.

Financial Disclosure

This study received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Use of AI for Writing Assistance

During the preparation of this manuscript, the authors used Claude (Anthropic, PBC) to improve the language, clarity, and readability of the text and to assist with English-language translation. No artificial intelligence tool was used for study design, data collection, statistical analysis, or interpretation of the results. After using this tool, the authors reviewed and edited the content as needed and take full responsibility for the content of the publication.

Authorship Contributions

Concept – N.D., F.S., L.D.; Design – N.D., B.Y., F.S.; Supervision – B.Y., F.S., L.D.; Materials – F.K., Y.S., B.Y., M.D.; Data collection &/or processing – N.D., F.K., Y.S., D.B.G., M.D., T.B.G., B.K., C.I.K., F.Kemik; Analysis and/or interpretation – N.D., B.Y., F.S., L.D.; Literature search – N.D., D.B.G., T.B.G., B.K., C.I.K., F.Kemik; Writing – N.D.; Critical review – N.D., F.K., Y.S.K., D.B.G., B.Y., M.D., T.B.G., B.K., C.I.K., F.Kemik, F.S., L.D.

Peer-review

Externally peer-reviewed.