Type of marker | Marker | Cancer type | Timepoint | Object | Cases | Main findings | Efficacy of marker | Refs | |||
---|---|---|---|---|---|---|---|---|---|---|---|
Radiomic-based biomarker | 18F-FDG PET/CT signature | NSCLC | Pre-treatment | Patients | 99,47,48 | 18F-FDG PET/CT signatures pre-treatment identified patients benefiting from ICIs | AUC = 0.86, 0.83, 0.81 | [21] | |||
 | PD-L1 DLS | NSCLC | Pre-treatment | Patients | 697 | Combined with clinical data, DLS was capable of accurately predicting DCB, PFS, and OS in different cohorts | C-index = 0.70–0.87 | [22] | |||
 | rADC | Glioblastoma | Post-treatment | Patients | 44 | Patients with rADC ≥ 1.63 showed longer OS | HR = 0.41, P = 0.02 | [23] | |||
 | Radiomic score of tumor-infiltrating CD8 + T cells | Advanced solid tumors | Pre-treatment | Patients | 137 | Higher radiomic scores at baseline correlated to a higher proportion of patients with objective response or SD at 6 months and longer OS | P = 0.025, 0.013 for objective response and SD respectively HR = 0.58, P = 0.0081 for OS | [24] | |||
 | Maximum 89Zr-labeled CD4 ratio (tumor to heart) | 7 different tumor models | Pre-treatment | Mice | 35 | The 89Zr-labeled CD4 ratio > 9 was associated with longer OS | P = 0.0018 | [25] | |||
 | 68 Ga-grazytracer | Colon Cancer | 12 days post-tumor inoculation | Mice | 12 | The high 68 Ga-grazytracer uptake group showed smaller tumor volumes compared with the low uptake group | P < 0.05 | [26] | |||
 | SUVmax of 89Zr-labeled atezolizumab | Bladder cancer, NSCLC, and TNBC | Pre-treatment | Patients | 22 | Patients with CR had a higher SUVmax compared to those with progressive disease The geometric mean SUVmax correlated to PFS and OS | P = 0.00021 HR = 11.7, P = 0.000028 for PFS; HR = 6.3, P = 0.0027 for OS | [27] | |||
 | 89Zr-labeled pembrolizumab | Advanced melanoma or NSCLC | Pre-treatment | Patients | 18 | The tumor SUVmax was associated with ICI response, PFS, and OS | P trend = 0.014 P = 0.0025 for PFS P = 0.026 for OS | [28] | |||
Blood-based biomarker | CTCs | NSCLC | Pre- and 4 weeks post-treatment | Patients | 104 | The presence of CTCs independently predicted the lack of durable response to ICIs at baseline and 4 weeks after treatment | OR 0.28, P = 0.02 at baseline; OR 0.07, P < 0.01 at four weeks after treatment | [29] | |||
 | CTC heterogeneity | Metastatic genitourinary cancer | Pre- and on-treatment | Patients | 81 | The B and D subtypes were associated with shorter OS at baseline and on C2D1.baseline Increasing CTC heterogeneity correlated to worse OS during the treatment | P < 0.0001–0.013 P = 0.045 | [30] | |||
 | PD-L1 expression on CTCs | Metastatic melanoma | Pre-treatment | Patients | 25 | Patients with PD-L1 + CTCs had longer PFS PD-L1 + CTCs were independent predictors of PFS | PFS, 26.6 vs. 5.5 months, P = 0.018 HR = 0.229, P = 0.026 | [31] | |||
 | PD-L1 expression on CTCs | NSCLC | 8 weeks post-treatment | Patients | 45 | Patients with PD-L1 positivity rates ≥ 7.7% at week 8 had longer PFS | P < 0.01 | [32] | |||
 | Ki67 level of circulating PD-1 + CD8 + T cells | Melanoma | Pre- and 6 weeks post-treatment | Patients | 29 | Higher Ki67 levels of circulating PD-1 + CD8 + T cells at baseline showed worse OS Patients with the ratio (PD-1 + Ki67 + CD8 + T cell to tumor burden) > 1.94 at 6 weeks post-treatment showed better outcomes in overall response rate, PFS, and OS | P = 0.02 P < 0.05 | [33] | |||
 | TCR diversity and clonality of PD1 + CD8 + T cells | NSCLC | Pre- and post-treatment | Patients | 25, 15 | Patients with higher TCR diversity pre-ICI had better responses and longer PFS in the combined dataset Patients with increased TCR clonality post-ICI had longer PFS and OS | The optimal Youden’s index = 0.81, Sensitivity = 0.87, Specificity = 0.94 PFS, HR = 0.28; 95% CI 0.11–0.74, P = 0.002 OS, HR = 0.23, 95% CI 0.07–0.79; P = 0.034 | [34] | |||
 | TMR | NSCLC | Pre- and post-treatment | Patients | 34 | TMR could distinguish responders and non-responders Patients with TMR > 0.39 had longer PFS | AUC = 87% Median PFS, 103 vs. 35 days, P = 0.0079 | [35] | |||
 | LIPS | Multiple recurrent or metastatic cancer types | Pre-treatment and after the first application | Patients | 56, 33 | The signature predicted OS benefit accurately The low-risk group had longer OS in the training and validation cohort | C index 0.74 vs. 0.71 Training cohort, HR = 0.26, 95% CI 0.12–0.56, P = 0.00025; Validation cohort, HR = 0.30, 95% CI 0.10–0.91, P = 0.024 | [36] | |||
 | ctDNA | NSCLC, Melanoma, Colorectal Cancer | 8 weeks post-treatment | Patients | 15 | Detection of ctDNA at week 8 correlated with shorter PFS and OS | Median PFS, 11 vs. 2 months, HR 10.2, P = 0.001 OS, HR = 15, P = 0.004 | [37] | |||
 | bTMB | NSCLC | Pre-treatment | Patients | 152 | The bTMB-high group reached higher ORR values and longer OS | ORR, 35.7% vs. 5.5%, P < 0.0001 OS, 23.9 vs. 13.4 months, HR = 0.66, P = 0.18 | [38] | |||
 | bTMB | NSCLC | Pre-treatment | Patients | 50 | bTMB levels ≥ 6 was associated with better PFS and ORR | PFS, HR = 0.39, P = 0.01 ORR, 39.3% vs. 9.1%, P = 0.02 | [39] | |||
 | GIN | 18 cancer types | 6 weeks post-treatment | Patients | 44 | GIN of cfDNA depicted the ICI efficacy at week 6 | HR (NRs vs. Rs) = 5.74, P = 0.001 | [40] | |||
 | Specific open regions of chromatin | Gastric cancer | Pre-treatment | Patients | 32, 52 | Patients with high chromatin openness tended to respond to ICIs and had better prognoses | Discovery cohort, Sensitivity 100.0%, Specificity 90.9%, P < 0.001 Validation cohort, Sensitivity 88.9%, Specificity 58.8%, P < 0.001 AUC = 0.717 | [41] | |||
 | Lung dynamics index | NSCLC | Pre- and within 4 weeks post-treatment | Patients | 22 | The index differentiated patients with DCB from NDB and correlated with PFS | AUC = 0.93 PFS, HR = 11.38, Wald P = 0.006 | [42] | |||
 | LIF | Multiple unresectable or metastatic cancer types | Pre-treatment | Patients | 95, 292 | The LIF-low group had longer PFS, OS, and DCB | Median PFS, 7.4 vs. 1.7 months, 95% CI 2.9–11.9 vs. 1.3–2.1 months, P < 0.0001 Median OS, 21.7 vs. 4.3 months, 95% CI 12‒31.4 vs. 3.4–5.1 months, P < 0.0001 DCB, 41.7% vs. 6.4%, P < 0.0001 AUC = 0.622 | [43] | |||
 | HIC | NSCLC | Pre-treatment | Patients | 284, 877 | The HIC-H group had longer OS in all ICI regimens and ICI monotherapy | Median OS, not-reached vs. 5.0 months, HR = 0.38, P < 0.0001 for all ICI regimens OS, 16.8 vs. 2.8 months, HR = 0.36, P < 0.0001 for ICI monotherapy | [44] | |||
 | CRAFITY score | HCC | Pre-treatment | Patients | 190, 102 | Patients with a low CRAFITY score had the longest OS and best radiological responses | P < 0.001, C index = 0.62 | [45] | |||
 | Circulating exosomal PD-L1 | Melanoma | Pre- and 3–6 weeks post-treatment | Patients | 39 | High levels of circulating exosomal PD-L1 pre-treatment were associated with poor clinical outcomes Responders showed increased exosomal PD-L1 levels at week 3–6 Patients with the fold change value > 2.43 at week 3–6 had better prognoses | P = 0.0018 P = 0.00001 P < 0.05 | [46] | |||
 | Circulating exosomal CD73 | Melanoma | 4 weeks post-treatment | Patients | 41 | Circulating exosomal CD73 greatly increased in non-responders at week 4 compared with baseline | P = 0.0041 | [47] | |||
 | EV-score | Gastric cancer | Pre- and at the first month post-treatment | Patients | 112 | Baseline EV-score could characterize 6-month PD or death EV-score changes at the first month after treatment could predict prognosis | AUC = 0.729, 0.630 PFS, HR = 0.3677, P = 0.0471 OS, HR = 0.4568, P = 0.1828 | [48] | |||
Microbial biomarker | Microbiota composition | Cutaneous melanoma | Pre-treatment | Patients | 94, 5 microbiome datasets | Baseline microbiota composition correlated to the outcome one year after ICI initiation in a cohort of 94 patients Optimized algorithms predicted outcomes across five cohorts consistently | P = 0.006 AUC = 0.54–1.00 | [49] | |||
 | SCFA | Metastatic or advanced solid tumors | Pre-treatment | Patients | 52 | Responders had higher levels of fecal and serum SCFAs | P < 0.05 | [50] | |||
 | SCFA | Multiple myeloma | Pre-treatment | Patients | 85 | Lower baseline levels of butyrate and propionate were associated with longer PFS | P = 0.0015; P = 0.0029 | [51] | |||
Exhaled breath | Molecular profiles | NSCLC | Pre-treatment | Patients | 92, 51 | Baseline data significantly differentiated different responses at 3 months | AUC = 0.89, 0.85 | [52] | |||
 | SpiroNose exhaled breath data | NSCLC | 6 weeks post-treatment | Patients | 62, 32 | The eNose was capable of distinguishing objective responders in the early stage | Training, AUC = 0.95, Sensitivity = 100%, Specificity = 73%Validation, AUC = 0.97 | [53] | |||
Other characteristics | Gender | Advanced or metastatic tumors | - | Patients | 11,351 | Men and women had different ICI outcomes | P = 0.0019 | [54] | |||
 | BMI | Melanoma | - | Patients | 207, 331 | Obese patients had improved PFS and OS in the immunotherapy cohort | HR = 0.75, 0.64 | [55] | |||
 | BMI | Melanoma | - | Patients | 423 | To observe the association between BMI and survival outcomes | NS | [56] | |||
 | Body composition | Melanoma | - | Patients | 287 | Patients featured with sarcopenic obesity showed inferior PFS and those featured with high total adipose tissue index had shorter PFS | HR = 1.4, P = 0.04; HR = 1.7, P = 0.04 | [57] |