The association between albumin-globulin ratio (AGR) and survival in patients treated with immune checkpoint inhibitors

Abstract

BACKGROUND:

The albumin-globulin ratio (AGR) could be a prognostic biomarker in patients with cancer, although the data is limited in patients treated with immune-checkpoint inhibitors (ICIs).

OBJECTIVES:

We aimed to evaluate the association between AGR and survival in ICI-treated patients.

METHODS:

The data of 212 advanced-stage patients were retrospectively evaluated in this cohort study. The association between AGR with overall (OS) and progression-free survival (PFS) were evaluated with multivariate analyses. Additionally, receptor operating curve (ROC) analysis was conducted to assess the AGR’s predictive power in the very early progression (progression within two months) and long-term benefit (more than twelve months survival).

RESULTS:

The median AGR was calculated as 1.21, and patients were classified into AGR-low and high subgroups according to the median. In the multivariate analyses, patients with lower AGR ( $<$ 1.21) had decreased OS (HR: 1.530, 95% CI: 1.100–2.127, $p=$ 0.011) and PFS (HR: 1.390, 95% CI: 1.020–1.895, $p=$ 0.037). The area under curve of AGR to detect early progression and long-term benefit were 0.654 (95% CI: 0.562–0.747, $p=$ 0.001) and 0.671 (95% CI: 0.598–0.744, $p<$ 0.001), respectively.

CONCLUSIONS:

In our experience, survival with ICIs was impaired in patients with lower AGR. Additionally, the AGR values could detect the very early progression and long-term benefit ICIs.

Keywords

Immune checkpoint inhibitor albumin-globulin ratio biomarker immunotherapy

1. Introduction

The immune checkpoint inhibitors (ICIs) became the fifth pillar of cancer treatment [1] and entered treatment algorithms in almost all tumors [2]. The ICIs improved outcomes in several tumors, including non-small cell lung cancer (NSCLC) [3], renal cell carcinoma (RCC) [4], melanoma [5], head and neck cancer (HNC) [6], and Hodgkin lymphoma [7]. However, the outcomes with ICIs are still far from desired, with less than 30% overall response rates in most settings [8], especially in the later treatment lines. Additionally, class-specific immune-related adverse events (IrAEs) [9] and the financial burden of the ICIs [10] make biomarker-based patient selection paramount.

Despite the stunning speed of the therapeutic field developments, the development and implementation of prognostic and predictive biomarkers was a bit slow, possibly due to limited access to complex platforms in most countries and limited tissue availability in the metastatic setting [11, 12]. So, biomarker-based treatment tailoring did not become standard other than tumor PD-L1 expression in the first-line treatment of NSCLC [3] and two tumor agnostic markers, microsatellite instability, and tumor mutational burden, which are present in a limited number of patients [13]. Additionally, these tissue-based biomarkers evaluate the tumor characteristics only and have limited power to reflect the complex host-tumor interactions [14], and have limited benefit in monitoring the dynamic anti-tumor immune response [15].

Recently, the problems and limitations with the tissue-based biomarkers with ICIs paved the way for blood-based biomarkers with a particular interest in the simple biomarkers obtained from the blood count and biochemistry parameters reflecting the immune-inflammatory status and aiding in the prognosis prediction [16]. The studies with C-reactive protein levels, neutrophil-lymphocyte ratio (NLR), and platelet-lymphocyte ratio (PLR) demonstrated poorer outcomes in patients with increased inflammatory pressure [17, 18, 19]. Similarly, the albumin-globulin ratio (AGR) could also reflect the host’s inflammatory status and have the additional advantage of reflecting nutritional status [20], although the data on the ICI-treated patients with AGR is very limited [21]. From this point, we aimed to evaluate the association of AGR values and survival outcomes in a large cohort of ICI-treated patients. Furthermore, we assessed the predictive power of AGR to predict patients garnering a long-term benefit and patients progressing in a short time under ICIs.

2. Materials and methods

2.1 Patients

The data of adult advanced (recurrent or metastatic) cancer patients treated with any ICIs between January 2014 and August 2020 were included in this retrospective study. Additionally, the data of patients treated with ICIs between August 2020 and September 2021 in two institutions (Hacettepe University and Istinye University) used for the validation of the results (validation cohort). Patients with acute or chronic infections and missing data for AGR were excluded. Additionally, patients with microsatellite instable tumors and lung cancer patients with PD-L1 expression of 50 or higher treated with ICIs in the first-line were excluded from the analyses to prevent confounding. All other patients treated within prespecified dates were included in the study due to the limited number of patients and no power analyses were conducted to determine the sample size. The patients were treated with ICIs without a biomarker-based selection (PD-L1, TMB) due to unavailability in most cases.

Demographic features, ICI types, baseline height and weight, Eastern Cooperative Oncology Group (ECOG) performance status, previous treatment lines, comorbidities, regularly used drugs were recorded together with the best response to immunotherapy and survival data. The baseline lactate dehydrogenase, neutrophil, platelet, lymphocyte, albumin, and globulin levels were collected and recorded from the routine laboratory tests ordered in the preceding one week before the first immunotherapy dose. The comorbidities were classified according to Charlson-Comorbidity Index (CCI) [22]. The AGR was calculated by dividing the albumin (gr/dL) to total protein-albumin values.

2.2 Statistical analyses

Descriptive statistics were expressed by the median, interquartile range (IQR; 25th-75th percentile), standard errors for continuous variables, and frequency and percentages for categorical variables. Independent group comparisons were made using Mann-Whitney U and Chi-square tests for continuous and categorical variables, respectively. The median values were used as the cut-off for the definition of AGR-high and low groups. The Spearman test was used to evaluate the correlation between the AGR values and NLR/PLR. The overall response rate (ORR) was calculated by dividing the patients with a complete or partial response to all response-evaluated patients. The overall survival (OS) time was defined as the period from treatment initiation to the last follow-up and/or death, and progression-free survival (PFS) time was defined as the period between treatment initiation to disease progression and/or death. Univariate survival analyses were conducted with Kaplan-Meier analyses and comparisons between prognostic groups were made by log-rank tests. Multivariate survival analyses were conducted by a backward Cox-regression model, including the significant parameters in the univariate analyses, and hazard ratios were calculated together with 95% confidence intervals (CI). The discriminative ability of AGR in very early progression (within two months) and long-term benefit (more than oneyear survival) was evaluated with receiver operating characteristics (ROC) curve analysis [23]. All statistical analyses were performed in SPSS, version 25.0 (IBM Inc., Armonk, NY, USA), and a type-I error level of 5% ( $p<$ 0.05) was considered as the threshold limit for statistical significance. The study was conducted following the STROBE guidelines and STROBE checklist is available as a supplement.

3. Results

3.1 Baseline characteristics

A total of 253 patients were treated within the prespecified dates. After the exclusion of the patients treated within early access programs ( $n=$ 21), patients with active infections ( $n=$ 10), patients without available albumin and globulin levels in the week before the ICI dose ( $n=$ 1), MSI-H patients ( $n=$ 3) and NSCLC patients with $\geqslant$ 50% PD-L1 expression treated in the first-line ( $n=$ 6), a total of 212 patients were included in the analyses. The study cohort’s median age was 61 (IQR 51–67) years, and 31.6% of the patients were over 65 years of age. Melanoma was the most common diagnosis (23.1%), followed by RCC (22.2%) and NSCLC (17.5%). Nivolumab was the most commonly used ICI (148/212, 69.8%) due to reimbursement in melanoma and RCC in our country (Table 1). The 21.2% of the patients were treated with a combination of ICI and chemotherapy (CT) or targeted therapy.

Table 1
Baseline patient characteristics of study population

Clinical feature	$n$ (%)
Sex
Male	145	(68.4)
Female	67	(31.6)
ECOG PS
0	133	(63.6)
1	52	(24.9)
2	22	(10.5)
3	2	(1)
Immunotherapy agent
Nivolumab	148	(69.8)
Atezolizumab	31	(14.6)
İpilimumab	18	(8.5)
Pembrolizumab	14	(6.6)
Avelumab	1	(0.5)
Primary tumor
Melanoma	49	(23.1)
RCC	47	(22.2)
NSCLC	37	(17.5)
HNC	13	(6.1)
Hodgkin’s	12	(5.7)
TCC	12	(5.7)
Other	42	(19.7)
Line of treatment
1	26	(12.3)
2	78	(36.8)
3	50	(23.6)
4 or later	58	(27.3)

${}^{*}$ RCC: renal cell carcinoma, NSCLC: non-small cell lung cancer, TCC: transitional cell cancer, HNC: Head and neck cancer. ${}^{\#}$ Other: colorectal cancer ( $n=$ 8), prostate cancer ( $n=$ 7), small cell lung cancer ( $n=$ 7), sarcoma ( $n=$ 6), mesothelioma ( $n=$ 4), hepatocellular cancer ( $n=$ 3), testicular cancer ( $n=$ 2), breast cancer ( $n=$ 3), Merkel cell carcinoma ( $n=$ 1), pancreatic cancer ( $n=$ 1).

Table 2

Comparison of baseline characteristics in the AGR low and AGR high groups

	Low AGR ( $n=$ 105) $n$ (%)	High AGR ( $n=$ 107) $n$ (%)	$p$ -value
Median age (IQR)	62 (54–68)	61 (48–66)	0.156
Age group
$<$ 65 years	67 (63.8)	78 (72.9)	0.155
$\geqslant$ 65 years	38 (36.2)	29 (27.1)
Sex
Female	24 (22.9)	43 (40.2)	0.007
Male	81 (77.1)	64 (59.8)
Primary tumor
Melanoma	21 (20)	28 (26.2)	0.347
RCC	23 (21.9)	24 (22.4)
NSCLC	23 (21.9)	14 (13.1)
Other	38 (36.2)	41 (38.3)
Line of treatment
First line	12 (11.4)	14 (13.1)	0.894
Second line	40 (38.1)	38 (35.5)
Third line or later	53 (50.5)	55 (51.4)
ECOG status
ECOG 0	62 (60.2)	71 (67)	0.308
ECOG 1 or higher	41 (39.8)	35 (33)
LDH levels
Normal	52 (49.5)	66 (61.7)	0.075
$>$ ULN	53 (50.5)	41 (38.3)
Polypharmacy
Absent	89 (84.8)	95 (88.8)	0.387
Present	16 (15.2)	12 (11.2)
ICI type
Nivolumab	78 (74.3)	70 (65.4)	0.160
Other	27 (25.7)	37 (34.6)
Median CCI (IQR)	8 (7–9)	8 (7–9)	0.209
Median BMI (IQR)	26.23 (22.90–29.41)	26.03 (23.62–29.38)	0.893
Median NLR (IQR)	4.27 (2.70–6.22)	2.95 (2–4.29)	$<$ 0.001
Median PLR (IQR)	227.33 (146.06–315.71)	170.53 (116.25–255)	0.004

${}^{*}$ AGR: albumin-globulin ratio, RCC: renal cell cancer, NSCLC: non-small cell lung cancer, ULN: upper limit of normal, ICI: immune checkpoin inhibitor, CCI: Charlson Comorbidity Index, BMI: body mass index, NLR: neutrophil-lymphocyte ratio, PLR: platelet-lymphocyte ratio.

The median BMI of the cohort was 26.12 (IQR 23.40–29.89) kg/m ${}^{2}$ , and the LDH levels were higher than normal in 43.9% of the patients. The median CCI was 8 (IQR 7–9). The median AGR was calculated as 1.21, and patients were classified into AGR-low and high subgroups according to the median AGR (1.21) cut-off. Additionally, this AGR cut-off (1.21) had 66% sensitivity and 56% specificity for OS in ROC curve analyses (Supplement). The baseline characteristics were largely similar between the AGR-low and AGR-high groups (Table 2). In correlation analyses, the AGR had a significant negative correlation with NLR ( $r=-$ 291, $p<$ 0.001) and PLR ( $r=-$ 239, $p<$ 0.001) (Fig. 1). The medians for NLR and PLR were 3.52 and 197, respectively.

Figure 1.

The association between NLR (left) and PLR (right) with AGR values.

Figure 2.

The association between response (complete or partial) with OS (left) and PFS (right).

3.2 Survival analyses

During a median follow-up of 9.56 (IQR 4.97–20.91) months, 154 patients died (72.6%), and 166 patients (78.3%) had any PFS event. The median OS was 9.76 $\pm$ 1.13 months, and the median PFS was 4.34 $\pm$ 0.53 months. In patients with an evaluable response (83% of the cohort), the ORR was 30.1% (53/176), and the disease control rate was 43.2% (76/176). The OS ( $>$ NR vs. 9 $\pm$ 0.83 months, $p<$ 0.001) and PFS ( $>$ NR vs. 3.15 $\pm$ 0.22, $p<$ 0.001) were significantly better in patients with a complete or partial response (Fig. 2).

In univariate analyses, patients with higher AGR ( $>$ 1.21 vs. $<$ 1.21) had better OS (16 $\pm$ 1.91 vs. 7.66 $\pm$ 0.95 months, $p<$ 0.001) (Fig. 3). The overall survival times were increased with the increased AGR quartile (from Q1 to Q4) (Fig. 4). In contrast, patients with higher CCI ( $\leqslant$ 8 vs. $>$ 8, $p=$ 0.001), higher ECOG status (0 vs. $\geqslant$ 1, $p<$ 0.001), older age ( $<$ 65 vs. $\geqslant$ 65 years, $p=$ 0.001), higher NLR ( $<$ 3.52 vs. $>$ 3.52, $p=$ 0.002) and higher PLR ( $<$ 197 vs. was $>$ 197, $p=$ 0.049) had decreased OS. The association between polypharmacy ( $p=$ 0.095), LDH levels (normal vs. $>$ normal, $p=$ 0.105), and regular proton pump inhibitor use ( $p=$ 0.928) did not have a significant association with the OS. The PFS analyses were generally consistent with OS analyses and demonstrated statistically significant associations between AGR ( $p=$ 0.015), ECOG ( $p=$ 0.005), NLR ( $p=$ 0.033), older age ( $p=$ 0.026), and OS. The association between PLR values ( $p=$ 0.296) and CCI ( $p=$ 0.119) with PFS did not reach statistical significance.

Figure 3.

The association between AGR with OS (left) and PFS (right).

Figure 4.

The association between AGR quartiles and OS.

Table 3

The association between clinical factors with overall and progression-free survival in multivariate analyses

	Overall survival			Progression-free survival
Clinical factor	HR	95% CI	$p$ -value	HR	95% CI	$p$ -value
Patient age ( $<$ 65 vs. $\geqslant$ 65)	1.253	0.822–1.909	0.294	1.287	0.923–1.794	0.137
CCI ( $\leqslant$ 8 vs. $\geqslant$ 9)	1.400	1.000–1.959	0.050
ECOG (0 vs. $\geqslant$ 1)	1.607	1.152–2.241	0.005	1.505	1.098–2.063	0.011
AGR ( $>$ 1.21 vs. $<$ 1.21)	1.530	1.100–2.127	0.011	1.390	1.020–1.895	0.037
NLR ( $<$ 3.52 vs. $>$ 3.52)	1.365	0.977–1.907	0.068	1.212	0.873–1.681	0.251
PLR ( $<$ 197 vs. $>$ 197)	1.116	0.787–1.582	0.537

${}^{*}$ AGR: albumin-globulin ratio, CCI: Charlson Comorbidity Index, NLR: neutrophil-lymphocyte ratio, PLR: platelet-lymphocyte ratio.

Figure 5.

The ROC analyses for AGR values in progression in two months (left) and long-term benefit (right).

3.3 Multivariate analyses

A multivariate analysis model was constructed and included the following parameters: ECOG performance status (0 vs. $\geqslant$ 1), CCI ( $\leqslant$ 8 vs. $>$ 8), lower AGR ( $<$ 1.21 vs. $>$ 1.21), older age ( $<$ 65 vs. $\geqslant$ 65 years), higher NLR ( $<$ 3.52 vs. $>$ 3.52) and PLR ( $<$ 197 vs. $>$ 197) values. In the multivariate analyses, the patients with higher ECOG status (HR: 1.607, 95% CI 1.152–2.241, $p=$ 0.005), lower AGR values (HR: 1.530, 95% CI: 1.100–2.127, $p=$ 0.011) and higher CCI (HR: 1.400, 95% CI: 1.000–1.959, $p=$ 0.050) had decreased OS (Table 3). In the multivariate model for PFS, the patients with higher ECOG status (HR: 1.505, 95% CI: 1.098–2.063, $p=$ 0.011) and lower AGR values (HR: 1.390, 95% CI: 1.020–1.895, $p=$ 0.037) had significantly decreased PFS (Table 3). Additional analyses with an adjustment according to overall response rate was demonstrated an independent association of AGR with OS (HR: 1.746, 95% CI: 1.213–2.513, $p=$ 0.003 for OS, HR: 1.789, 95% CI: 1.242–2.578, $p=$ 0.002). Likewise, adjustments according to LDH levels and treatment type (ICI alone or ICI plus CT or targeted therapy) did not change the association between AGR and OS (HR: 1.683, 95% CI: 1.218–2.325, $p=$ 0.002). While we could not able to conduct separate survival analyses in the individual tumor types due to limited sample size, we made further adjustments according to the tumor type (melanoma, RCC, NSCLC or other tumors) and observed a persistent negative association between lower AGR levels and OS (HR: 1.637, 95% CI: 1.183–2.266, $p=$ 0.003) and PFS (HR: 1.402, 1.029–1.910, $p=$ 0.032).

A statistically significant association between AGR with OS (HR: 1.517, 95% CI: 1.092–2.108, $p=$ 0.013) and PFS (HR: 1.563, 95% CI: 1.126–2.171, $p=$ 0.008) was retained when the five was used as the NLR cut-off instead of the median NLR. In addition, in the alternate multivariate model with NLR $<$ 5 vs. $>$ 5 dichotomization, patients with higher NLR values ( $>$ 5) had significantly decreased OS (HR: 1.536, 95% CI: 1.091–2.164, $p=$ 0.014). Similarly, when limiting to analyses to melanoma, RCC, NSCLC, HNC, and transitional cell carcinoma (TCC) patients ( $n>$ 10 in each group, total of 167 patients) to prevent confounding by the anecdotal patients, lower AGR levels ( $<$ 1.21) were associated with a significantly inferior OS (HR: 1.454, 95% CI: 1.003–2.108, $p=$ 0.048).

3.4 Receiver operating characteristics (ROC) analyses

The receiver operating characteristics (ROC) curve analyses demonstrated a moderate ability of AGR to predict OS (area under curve (AUC): 0.637, 95% CI: 0.555–0.719, $p=$ 0.002) and PFS (AUC: 0.614, 95% CI: 0.530–0.698, $p=$ 0.018). Additional ROC analyses were performed to evaluate the performance of AGR to discriminate patients who progressed very early (within two months) and patients with a long-term benefit (survived over twelve months). The AUC of AGR to detect progression within two months was 0.654 (95% CI: 0.562–0.747, $p=$ 0.001). The median AGR values (1.21) had 71% sensitivity and 55% specificity to predict progression within two months. Similarly, AGR values had moderate performance to discriminate the patients who lived more than 12 months with a 0.671 AUC (95% CI: 0.598–0.744, $p<$ 0.001) (Fig. 5). The sensitivity and specificity of this AGR cut-off (1.21) for long-term survival ( $>$ 12 months) had 64% and 60%, respectively.

3.5 Validation cohort

We evaluated the performance of AGR as a prognostic and predictive biomarker in a validation cohort of 128 patients. The median age of the validation cohort was 60 (IQR 52–68) and 53.9% of the patients were male. The RCC (15.6%), melanoma (13.3%), and NSCLC (11.7%) were the most frequent diagnoses. Similar to the development cohort most patients (80.2%) had a good ECOG performance status (0 or 1). The median follow-up was 6.16 (IQR 3.56–9.87) months in the validation cohort. During the follow-up, 51 (39.8%) patients died and 86 patients (67.2%) had any PFS event. The median PFS was 4.01 $\pm$ 0.40 and the median OS was over 11.28 months. Higher AGR levels ( $>$ 1.21) were associated with significantly better OS (NR vs. 5.75 $\pm$ 2.71 months, $p=$ 0.006) and PFS (5.26 $\pm$ 1.03 vs. 2.60 $\pm$ 0.44 months, $p=$ 0.006) in the validation cohort (Supplement).

Similar to the development cohort, AGR values had moderate ability to predict OS (AUC: 0.676, 95% CI: 0.543–0.810, $p=$ 0.019) and PFS (AUC: 0.679, 95% CI: 0.543–0.816, $p=$ 0.027). The AUC of AGR to detect progression within two months was 0.781 (95% CI: 0.659–0.904, $p=$ 0.001). Due to low maturity, ROC analyses were not performed to evaluate the performance of AGR to predict more than 12 months survival.

4. Discussion

In this study, we observed significantly lower OS and PFS in patients with lower AGR levels and higher ECOG performance status scores. Additionally, the patients with higher CCI had decreased OS. The decreased overall survival in the AGR-low patients was independent of the response to immunotherapy, the NLR levels and the use of ICI alone or ICI plus CT or targeted therapy.

Malnutrition is a widespread problem in patients with cancer and is related to decreased performance status, poor treatment tolerance, and impaired overall survival [24, 25, 26, 27]. There are several tools to evaluate malnutrition, including anthropometric measurements and questionnaires [28, 29]. The albumin levels could also serve as a denominator of malnutrition [30]. Besides, it has close relations with the host’s inflammatory status as its levels decrease with inflammation [31, 32, 33]. These features make albumin a candidate biomarker to reflect both the nutritional and the inflammatory status. Not surprisingly, the patients treated with chemotherapy or surgery with lower albumin levels had decreased overall survival [34]. However, the data on the value of albumin in ICI-treated patients is relatively limited. In a recent study on 106 NSCLC patients treated with ICIs, patients with low albumin levels ( $<$ 3.5 gr/dL) had significantly decreased OS (6.9 vs. 19.5 months) [35]. In another study on 76 NSCLC patients treated nivolumab or durvalumab reported an inferior survival in patients with lower albumin levels ( $<$ 4.3 gr/dL) [36]. A similar study to our study was conducted by Ibrahimi et al. in a cohort of 188 patients treated with immunotherapy. However, no association with albumin levels and OS ( $p=$ 0.12) or PFS ( $p=$ 0.19) was present in the study [37].

The globulin levels reflect the combined antibody and cytokine burden in the body and increase in almost all chronic inflammatory states, including cancer [38, 39]. The cytokines produced during the inflammatory stress are reflected in the globulin constituent of the blood and have instrumental roles in tumor progression and immune exhaustion [40]. Based on this, globulin levels could be an indirect reflector of uncontrolled inflammation and tumoral progression. So, AGR has a solid biologic rationale to predict prognosis in cancer. An association between the lower AGR values and decreased OS was demonstrated in various tumors, including but not limited to urothelial cancer [41], RCC [42], hepatocellular cancer [43, 44], colorectal cancer [39], testicular cancer [45], breast cancer [46] and pancreatic cancer [47]. However, the cut-offs for AGR and sample sizes were very variable. A recent combined analysis of 14 studies encompassing 4136 patients with cancer demonstrated lower OS in patients with lower AGR values (HR: 1.87, 95% CI 1.50–2.34, $p<$ 0.001) [48], albeit patients treated with ICIs was not included in this meta-analysis. Similarly, a recent study on the 74 NSCLC patients treated with ICIs used a similar cut-off (1.17) to our study (1.21) and reported a significantly decreased OS and PFS, independent of the ECOG status, sex, and histologic type [21]. However, a more prominent negative effect on OS (HR $>$ 4) was evident in the study [21], possibly due to the smaller sample size. In addition to the prognostic value, we demonstrated the possible value of AGR values in detecting progression in the first two months of immunotherapy, a criterion to define the proposed hyper progressive disease under immunotherapy [49]. If our results could be confirmed in the prospective trials, AGR values could help to predict the hyper progressive disease under ICIs.

We observed an association between poorer ECOG status and decreased OS and PFS. The ECOG status is a general nominator of overall status and disease and symptom burden [50]. As in our study, a strong correlation between the ECOG status and survival was consistently demonstrated in patients treated with chemotherapy [51, 52] or immunotherapy [53]. The CCI could also serve similarly and successfully quantify the comorbidity burden [22] and aid in treatment individualization. Although relatively understudied in the ICI-treated patients, in a recent study on 66 NSCLC patients, patients with higher CCI had decreased OS. Like our study, the association between the PFS and CCI did not reach statistical significance [54], possibly due to the limited sample size.

We additionally evaluated the prognostic role of NLR and PLR in survival outcomes with ICIs. We used the median NLR and PLR values as the cut-offs for NLR and PLR. While the univariate analyses were statistically significant, the association between the NLR or PLR with OS was not statistically significant in the multivariate analyses. However, when we restructure the model with a different NLR cut-off (five), the association of NLR with OS and PFS was significant. Our results support the recent observations pointing out a more robust negative association with survival with higher NLR cut-offs [55, 56]. However, the possibility of lower sensitivity should be kept in mind with this strategy.

Our study has several limitations. First, the study’s retrospective nature and a heterogeneous patient group with a modest patient number prevented us from conducting additional subgroups analyses. Additionally, the limited number of cases caused relatively wide confidence intervals in the multivariate analyses and prevented us from conducting linear logistic regression analyses instead of binary analyses with a patient dichotomization according to an AGR cut-off. Most of our patients were treated in the second or later lines of the treatment that limited our results’ generability to patients treated in the clinical trials and countries with access to immunotherapy in the earlier lines. The additional adjustments according to pathologic characteristics like PD-L1 expression levels and tumor mutational burden were not possible due to a lack of data in most cases. However, PD-L1 levels are not routinely used in the patient selection for the tumor types and treatment lines of our cohort. Despite these limitations, we demonstrated the promise of AGR as a possible biomarker for prognosis prediction.

In conclusion, in our experience, survival with ICIs was impaired in patients with lower AGR values. Additionally, the AGR values could detect the patients with very early progression and patients garnering long-term benefit ICIs. If prospective studies validate our observations, AGR could be a readily available biomarker for treatment tailoring in ICI-treated patients.

Funding

The authors received no financial support for this article.

Compliance with ethical standards

All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. The study was approved by the ethics committee of Hacettepe University.

Consent to participate

N/A

Consent for publication

N/A

Availability of data and material

The data was not reposited in the data repositories.

Code availability

N/A.

Author contributions

Conception: Deniz Can Guven, Oktay Halit Aktepe, Mustafa Erman, Suayib Yalcin, Sercan Aksoy, Saadettin Kilickap.

Interpretation or analysis of data: Deniz Can Guven, Oktay Halit Aktepe, Melek Seren Aksun, Taha Koray Sahin, Gozde Kavgaci, Enes Ucgul, Ibrahim Yahya Cakir, Hasan Cagri Yildirim, Gurkan Guner, Serkan Akin, Neyran Kertmen, Saadettin Kilickap.

Preparation of the manuscript: All Authors.

Revision for important intellectual content: All Authors.

Supervision: Mustafa Erman, Suayib Yalcin, Sercan Aksoy, Saadettin Kilickap.

Supplementary data

The supplementary files are available to download from http://dx.doi.org/10.3233/CBM-210349.

sj-pdf-1-cbm-10.3233_CBM-210349.pdf - Supplemental material

Supplemental material, sj-pdf-1-cbm-10.3233_CBM-210349.pdf

sj-xlsx-1-cbm-10.3233_CBM-210349.xlsx - Supplemental material

Supplemental material, sj-xlsx-1-cbm-10.3233_CBM-210349.xlsx

sj-xlsx-2-cbm-10.3233_CBM-210349.xlsx - Supplemental material

Supplemental material, sj-xlsx-2-cbm-10.3233_CBM-210349.xlsx

Footnotes

Conflict of interest

The authors declare that they have no conflict of interest.

References

Abdel-Rahman

, ECOG performance score 0 versus 1: Impact on efficacy and safety of first-line 5-FU-based chemotherapy among patients with metastatic colorectal cancer included in five randomized trials, International Journal of Colorectal Disease 34 (2019), 2143–2150.

H.J.

Chon

H.J.

and Kim

, Peripheral blood-based biomarkers for immune checkpoint inhibitors, International Journal of Molecular Sciences 22 (2021).

Ansell

S.M.

Lesokhin

A.M.

Borrello

Halwani

Scott

E.C.

Gutierrez

Schuster

S.J.

Millenson

M.M.

Cattry

Freeman

G.J.

Rodig

S.J.

Chapuy

Ligon

A.H.

Zhu

Grosso

J.F.

Kim

S.Y.

Timmerman

J.M.

Shipp

M.A.

and Armand

, PD-1 blockade with nivolumab in relapsed or refractory hodgkin’s lymphoma, New England Journal of Medicine 372 (2014), 311–319.

Azab

Kedia

Shah

Vonfrolio

Naboush

Mohammed

and Bloom

S.W.

, The value of the pretreatment albumin/globulin ratio in predicting the long-term survival in colorectal cancer, International Journal of Colorectal Disease 28 (2013), 1629–1636.

Bilen

M.A.

Martini

D.J.

Liu

Lewis

Collins

H.H.

Shabto

J.M.

Akce

Kissick

H.T.

Carthon

B.C.

Shaib

W.L.

Alese

O.B.

Pillai

R.N.

Steuer

C.E.

C.S.

Lawson

D.H.

Kudchadkar

R.R.

El-Rayes

B.F.

Master

V.A.

Ramalingam

S.S.

Owonikoko

T.K.

and Harvey

R.D.

, The prognostic and predictive impact of inflammatory biomarkers in patients who have advanced-stage cancer treated with immunotherapy, Cancer 125 (2019), 127–134.

Guven

D.C.

Sahin

T.K.

Dizdar

and Kilickap

, Predictive biomarkers for immunotherapy efficacy in non-small-cell lung cancer: current status and future perspectives, Biomarkers in Medicine 14 (2020), 1383–1392.

Castillo-Martínez

Castro-Eguiluz

Copca-Mendoza

E.T.

Pérez-Camargo

D.A.

Reyes-Torres

C.A.

Ávila

E.A.

López-Córdova

Fuentes-Hernández

M.R.

Cetina-Pérez

and Milke-García

M.D.P.

, Nutritional assessment tools for the identification of malnutrition and nutritional risk associated with cancer treatment, Revista de Investigación Clínica 70 (2018), 121–125.

Charlson

M.E.

Pompei

Ales

K.L.

and MacKenzie

C.R.

, A new method of classifying prognostic comorbidity in longitudinal studies: Development and validation, Journal of Chronic Diseases 40 (1987), 373–383.

Chen

Shao

Yao

Zhuang

Wang

Xing

and Xu

, Preoperative albumin to globulin ratio predicts survival in clear cell renal cell carcinoma patients, Oncotarget 8(29) (2017).

10.

Chi

Xie

Jia

Liu

Sun

Chen

and Yi

, Prognostic value of albumin/globulin ratio in survival and lymph node metastasis in patients with cancer: A systematic review and meta-analysis, Journal of Cancer 9 (2018), 2341–2348.

11.

Deng

Pang

Miao

R.-C.

Chen

Zhou

Y.-Y.

J.-B.

Liu

S.-S.

Zhang

J.-Y.

and Liu

, Prognostic significance of pretreatment albumin/globulin ratio in patients with hepatocellular carcinoma, OncoTargets and Therapy 9 (2016), 5317–5328.

12.

Dewys

W.D.

Begg

Lavin

P.T.

Band

P.R.

Bennett

J.M.

Bertino

J.R.

Cohen

M.H.

Douglass

H.O.

, Jr. Engstrom

P.F.

Ezdinli

E.Z.

Horton

Johnson

G.J.

Moertel

C.G.

Oken

M.M.

Perlia

Rosenbaum

Silverstein

M.N.

Skeel

R.T.

Sponzo

R.W.

and Tormey

D.C.

, Prognostic effect of weight loss prior to chemotherapy in cancer patients. Eastern Cooperative Oncology Group, Am J Med 69 (1980), 491–497.

13.

X.-J.

Tang

L.-L.

Mao

Y.-P.

Sun

Zeng

M.-S.

Kang

T.-B.

Jia

W.-H.

Lin

A.-H.

and Ma

, The pretreatment albumin to globulin ratio has predictive value for long-term mortality in nasopharyngeal carcinoma, PloS One 9 (2014), e94473.

14.

Ferris

R.L.

Blumenschein

Fayette

Guigay

Colevas

A.D.

Licitra

Harrington

Kasper

Vokes

E.E.

Even

Worden

Saba

N.F.

Iglesias Docampo

L.C.

Haddad

Rordorf

Kiyota

Tahara

Monga

Lynch

Geese

W.J.

Kopit

Shaw

J.W.

and Gillison

M.L.

, Nivolumab for recurrent squamous-cell carcinoma of the head and neck, New England Journal of Medicine 375 (2016), 1856–1867.

15.

Gjoerup

Brown

C.A.

Ross

J.S.

Huang

R.S.P.

Schrock

Creeden

Fabrizio

and Tolba

, Identification and utilization of biomarkers to predict response to immune checkpoint inhibitors, Aaps J 22 (2020), 132.

16.

Gruys

Toussaint

M.J.M.

Niewold

T.A.

and Koopmans

S.J.

, Acute phase reaction and acute phase proteins, Journal of Zhejiang University. Science. B 6 (2005), 1045–1056.

17.

Guner

and Seker

K.G.

, The role of preoperative albumin to globulin ratio in predicting prognosis in testicular cancer patients, Actas Urol Esp (Engl Ed) 44 (2020), 469–476.

18.

Gupta

and Lis

C.G.

, Pretreatment serum albumin as a predictor of cancer survival: A systematic review of the epidemiological literature, Nutr J 9 (2010), 69.

19.

Haslam

and Prasad

, Estimation of the percentage of US patients with cancer who are eligible for and respond to checkpoint inhibitor immunotherapy drugs, JAMA Network Open 2 (2019), e192535–e192535.

20.

Ibrahimi

Mukherjee

Roman

King

Machiorlatti

and Aljumaily

, Effect of body mass index and albumin level on outcomes of patients receiving anti PD-1/PD-L1 therapy, Journal of Clinical Oncology 36 (2018), 213–213.

21.

Jang

R.W.

Caraiscos

V.B.

Swami

Banerjee

Mak

Kaya

Rodin

Bryson

Ridley

J.Z.

L.W.

and Zimmermann

, Simple prognostic model for patients with advanced cancer based on performance status, Journal of Oncology Practice 10 (2014), e335–e341.

22.

Jiang

Peng

Chen

Liu

Cao

Wang

and Wu

, Peripheral blood biomarkers associated with outcome in non-small cell lung cancer patients treated with nivolumab and durvalumab monotherapy, Front Oncol 10 (2020), 913.

23.

Jin

Yang

Liu

and Li

, Association of the neutrophil to lymphocyte ratio and clinical outcomes in patients with lung cancer receiving immunotherapy: A meta-analysis, BMJ Open 10 (2020), e035031.

24.

Keller

, Nutritional laboratory markers in malnutrition, Journal of Clinical Medicine 8 (2019), 775.

25.

Kuzuya

Izawa

Enoki

Okada

and Iguchi

, Is serum albumin a good marker for malnutrition in the physically impaired elderly, Clinical Nutrition 26 (2007), 84–90.

26.

Landskron

De la Fuente

Thuwajit

and Hermoso

M.A.

, Chronic inflammation and cytokines in the tumor microenvironment, Journal of Immunology Research 2014 (2014), 149185–149185.

27.

Larkin

Chiarion-Sileni

Gonzalez

Grob

J.-J.

Rutkowski

Lao

C.D.

Cowey

C.L.

Schadendorf

Wagstaff

Dummer

Ferrucci

P.F.

Smylie

Hogg

Hill

Márquez-Rodas

Haanen

Guidoboni

Maio

Schöffski

Carlino

M.S.

Lebbé

McArthur

Ascierto

P.A.

Daniels

G.A.

Long

G.V.

Bastholt

Rizzo

J.I.

Balogh

Moshyk

Hodi

F.S.

and Wolchok

J.D.

, Five-year survival with combined nivolumab and ipilimumab in advanced melanoma, New England Journal of Medicine 381 (2019), 1535–1546.

28.

Lee

C.-S.

Devoe

C.E.

Zhu

Fishbein

J.S.

and Seetharamu

, Pretreatment nutritional status and response to checkpoint inhibitors in lung cancer, Lung Cancer Management 9 (2020), LMT31.

29.

Leuenberger

Kurmann

and Stanga

, Nutritional screening tools in daily clinical practice: The focus on cancer, Support Care Cancer 18(Suppl 2) (2010), S17–27.

30.

Zhang

Yan

Song

Wang

Chen

Jiao

and Wang

, Pretreatment neutrophil-to-lymphocyte ratio (NLR) may predict the outcomes of advanced non-small-cell lung cancer (NSCLC) patients treated with immune checkpoint inhibitors (ICIs), Frontiers in Oncology 10 (2020).

31.

Liam

C.-K.

Mallawathantri

and Fong

K.M.

, Is tissue still the issue in detecting molecular alterations in lung cancer, Respirology 25 (2020), 933–943.

32.

Lo Russo

Moro

Sommariva

Cancila

Boeri

Centonze

Ferro

Ganzinelli

Gasparini

Huber

Milione

Porcu

Proto

Pruneri

Signorelli

Sangaletti

Sfondrini

Storti

Tassi

Bardelli

Marsoni

Torri

Tripodo

Colombo

M.P.

Anichini

Rivoltini

Balsari

Sozzi

and Garassino

M.C.

, Antibody-Fc/FcR interaction on macrophages as a mechanism for hyperprogressive disease in non-small cell lung cancer subsequent to PD-1/PD-L1 blockade, Clin Cancer Res 25 (2019), 989–999.

33.

Lusted

L.B.

, Decision-making studies in patient management, New England Journal of Medicine 284 (1971), 416–424.

34.

Martins

Sofiya

Sykiotis

G.P.

Lamine

Maillard

Fraga

Shabafrouz

Ribi

Cairoli

Guex-Crosier

Kuntzer

Michielin

Peters

Coukos

Spertini

Thompson

J.A.

and Obeid

, Adverse effects of immune-checkpoint inhibitors: Epidemiology, management and surveillance, Nature Reviews Clinical Oncology 16 (2019), 563–580.

35.

McMillan

D.C.

Watson

W.S.

O’Gorman

Preston

Scott

H.R.

and McArdle

C.S.

, Albumin concentrations are primarily determined by the body cell mass and the systemic inflammatory response in cancer patients with weight loss, Nutr Cancer 39 (2001), 210–213.

36.

Motzer

R.J.

Escudier

McDermott

D.F.

George

Hammers

H.J.

Srinivas

Tykodi

S.S.

Sosman

J.A.

Procopio

Plimack

E.R.

Castellano

Choueiri

T.K.

Gurney

Donskov

Bono

Wagstaff

Gauler

T.C.

Ueda

Tomita

Schutz

F.A.

Kollmannsberger

Larkin

Ravaud

Simon

J.S.

L.-A.

Waxman

I.M.

and Sharma

, Nivolumab versus everolimus in advanced renal-cell carcinoma, New England Journal of Medicine 373 (2015), 1803–1813.

37.

Muscaritoli

Lucia

Farcomeni

Lorusso

Saracino

Barone

Plastino

Gori

Magarotto

Carteni

Chiurazzi

Pavese

Marchetti

Zagonel

Bergo

Tonini

Imperatori

Iacono

Maiorana

Pinto

Rubino

Cavanna

Di Cicilia

Gamucci

Quadrini

Palazzo

Minardi

Merlano

Colucci

and Marchetti

, Prevalence of malnutrition in patients at first medical oncology visit: The PreMiO study, Oncotarget 8 (2017), 79884–79896.

38.

Nakanishi

Masuda

Yamaguchi

Sakamoto

Horimasu

Mimae

Nakashima

Miyamoto

Tsutani

Iwamoto

Fujitaka

Miyata

Hamada

Okada

and Hattori

, Albumin-globulin ratio is a predictive biomarker of antitumor effect of anti-PD-1 antibody in patients with non-small cell lung cancer, Int J Clin Oncol 25 (2020), 74–81.

39.

Nixon

A.B.

Schalper

K.A.

Jacobs

Potluri

Wang

I.M.

and Fleener

, Peripheral immune-based biomarkers in cancer immunotherapy: Can we realize their predictive potential, Journal for immunotherapy of cancer 7 (2019), 325.

40.

Oiseth

S.J.

and Aziz

M.S.

, Cancer immunotherapy: A brief review of the history, possibilities, and challenges ahead, Journal of Cancer Metastasis and Treatment 3 (2017), 250–261.

41.

Oken

M.M.

Creech

R.H.

Tormey

D.C.

Horton

Davis

T.E.

McFadden

E.T.

and Carbone

P.P.

, Toxicity and response criteria of the eastern cooperative oncology group, American Journal of Clinical Oncology 5 (1982), 649–656.

42.

Pressoir

Desné

Berchery

Rossignol

Poiree

Meslier

Traversier

Vittot

Simon

Gekiere

J.P.

Meuric

Serot

Falewee

M.N.

Rodrigues

Senesse

Vasson

M.P.

Chelle

Maget

Antoun

and Bachmann

, Prevalence, risk factors and clinical implications of malnutrition in french comprehensive cancer centres, British journal of cancer 102 (2010), 966–971.

43.

Reck

Rodríguez-Abreu

Robinson

A.G.

Hui

Csöszi

Fülöp

Gottfried

Peled

Tafreshi

Cuffe

O’Brien

Rao

Hotta

Leiby

M.A.

Lubiniecki

G.M.

Shentu

Rangwala

and Brahmer

J.R.

, Pembrolizumab versus chemotherapy for PD-L1 – positive non – small-cell lung cancer, New England Journal of Medicine 375 (2016), 1823–1833.

44.

Sehgal

Gill

R.R.

Widick

Bindal

McDonald

D.C.

Shea

Rangachari

and Costa

D.B.

, Association of performance status with survival in patients with advanced non-small cell lung cancer treated with pembrolizumab monotherapy, JAMA Network Open 4 (2021), e2037120–e2037120.

45.

Shinde

Bhandare

M.S.

Chaudhari

Sarodaya

Agarwal

and Shrikhande

, Preoperative albumin-globulin ratio and its association with perioperative and long-term outcomes in patients undergoing pancreatoduodenectomy, Dig Surg 38 (2021), 275–282.

46.

Takamori

Takada

Shimokawa

Matsubara

Fujishita

Ito

Toyozawa

Yamaguchi

Okamoto

Yoneshima

Tanaka

Okamoto

Tagawa

and Mori

, Clinical utility of pretreatment Glasgow prognostic score in non-small-cell lung cancer patients treated with immune checkpoint inhibitors, Lung Cancer 152 (2021), 27–33.

47.

Vaddepally

R.K.

Kharel

Pandey

Garje

and Chandra

A.B.

, Review of indications of FDA-approved immune checkpoint inhibitors per NCCN guidelines with the level of evidence, Cancers (Basel) 12 (2020).

48.

Van Cutsem

and Arends

, The causes and consequences of cancer-associated malnutrition, European Journal of Oncology Nursing 9(Suppl 2) (2005), S51–63.

49.

Vano

Y.-A.

Oudard

M.-A.

Têtu

Thibault

Aboudagga

Scotté

and Elaidi

, Optimal cut-off for neutrophil-to-lymphocyte ratio: Fact or fantasy? A prospective cohort study in metastatic cancer patients, PloS One 13 (2018), e0195042–e0195042.

50.

Ventola

C.L.

, Cancer Immunotherapy, Part 3: Challenges and future trends, P & T: A Peer-Reviewed Journal for Formulary Management 42 (2017), 514–521.

51.

Verma

Sprave

Haque

Simone

C.B.

Chang

J.Y.

Welsh

J.W.

and Thomas

C.R.

, A systematic review of the cost and cost-effectiveness studies of immune checkpoint inhibitors, Journal for Immunotherapy of Cancer 6 (2018), 128.

52.

Wei

Wang

Zhou

and Tian

, Low pretreatment albumin-to-globulin ratio predicts poor prognosis in gastric cancer: Insight from a meta-analysis, Frontiers in Oncology 10 (2021), 623046–623046.

53.

Xuan

Yang

Tang

Zhao

Shao

and Wang

, Combination of the preoperative albumin to globulin ratio and neutrophil to lymphocyte ratio as a novel prognostic factor in patients with triple negative breast cancer, Cancer Management and Research 11 (2019), 5125–5131.

54.

Zeng

Zhu

Wang

Zeng

Long

and Zhu

, Effect of comorbidity on outcomes of patients with advanced non-small cell lung cancer undergoing Anti-PD1 immunotherapy, Med Sci Monit 26 (2020), e922576.

55.

Zhang

Zhou

L.-Q.

Z.-S.

Shen

Liu

L.-B.

Wang

Chen

X.-Y.

Fan

Zhang

Han

W.-K.

and Jin

, Prognostic significance of preoperative albumin-globulin ratio in patients with upper tract urothelial carcinoma, PloS One 10 (2015), e0144961.

56.

Zhang

Liu

Yang

Chen

and Luo

, The pretreatment albumin to globulin ratio, a validated biomarker, predicts prognosis in hepatocellular carcinoma, Journal of B.U.ON. 21 (2016), 925–934.

The association between albumin-globulin ratio (AGR) and survival in patients treated with immune checkpoint inhibitors

Abstract

BACKGROUND:

OBJECTIVES:

METHODS:

RESULTS:

CONCLUSIONS:

Keywords

1. Introduction

2. Materials and methods

2.1 Patients

2.2 Statistical analyses

3. Results

3.1 Baseline characteristics

Table 1 Baseline patient characteristics of study population

3.4 Receiver operating characteristics (ROC) analyses

3.5 Validation cohort

4. Discussion

Funding

Compliance with ethical standards

Consent to participate

Consent for publication

Availability of data and material

Code availability

Author contributions

Supplementary data

sj-pdf-1-cbm-10.3233_CBM-210349.pdf - Supplemental material

sj-xlsx-1-cbm-10.3233_CBM-210349.xlsx - Supplemental material

sj-xlsx-2-cbm-10.3233_CBM-210349.xlsx - Supplemental material

Footnotes

Conflict of interest

References

Table 1
Baseline patient characteristics of study population