WRITTEN REPORT Medicine Programme, Independent Work (30 p)
THE PROGNOSTIC ROLE OF C-REACTIVE PROTEIN AND ALBUMIN IN PATIENTS
UNDERGOING RESECTION OF COLORECTAL LIVER METASTASES
By: Patric Ejder
Supervisor: Prof. Bengt Isaksson
Co-supervisors: Dr. Jozef Urdzik and Dr. Petter Frühling
Date: 2018-05-15
Title: The prognostic role of C-reactive protein and albumin in patients undergoing resection of colorectal liver metastases.
By: Patric Ejder
ABSTRACT ... 2
POPULÄRVETENSKAPLIG SAMMANFATTNING PÅ SVENSKA ... 3
BACKGROUND ... 4
E PIDEMIOLOGY ... 4
P ROGNOSTIC FACTORS ... 4
S YSTEMIC INFLAMMATION AND CANCER ... 5
A IM AND HYPOTHESIS ... 6
METHODS ... 7
P ATIENTS ... 7
D ATA COLLECTION ... 8
S TATISTICS ... 8
RESULTS ... 8
C LINICOPATHOLOGICAL CHARACTERISTICS ... 8
CRP, ALBUMIN AND GPS ... 9
P ROGNOSTIC IMPACT OF CLINICOPATHOLOGICAL FACTORS ... 11
P ROGNOSTIC IMPACT OF CRP, ALBUMIN AND GPS ... 11
DISCUSSION ... 13
ACKNOWLEDGEMENT ... 17
REFERENCES ... 18
Abstract
Background and objectives
Systemic inflammation has been reported to correlate with poor survival in several tumour types but has been less studied in patients with colorectal liver metastases (CRLM). The aim of this study was to analyse the prognostic impact of the systemic inflammation markers CRP and albumin, and their combination in Glasgow Prognostic Score (GPS), in patients
undergoing potentially curative liver resection for CRLM.
Methods
A total of 849 patients undergoing liver resection between 2005 and 2015 were included in this double-centre, retrospective observational study. Preoperative systemic inflammation reflected by categorised CRP, albumin and GPS were analysed and compared with other prognostic factors through univariate and multivariate statistical methods.
Results
Patients with GPS 0 had a median survival of 62 months compared to 47 months in patients with GPS 1 and 27 months in patients with GPS 2. Multivariate analyses showed that preoperative GPS 2 was the strongest predictor of poor overall survival (HR = 2.16, 95% CI 1.1.55-3.00, P < 0.001).
Conclusions
Preoperative GPS is a strong prognostic factor in CRLM patients undergoing potentially
curative resection of liver metastases and should be included in the risk assessment and when
selecting treatment strategies.
Populärvetenskaplig sammanfattning på svenska
Cancer i tjock- och ändtarm (CRC) är en av de vanligaste och mest dödliga cancersorterna i världen. Hos många med diagnosen kommer cancern sprida sig vidare till levern, de får då så kallade kolorektala levermetastaser (CRLM). Stora framsteg i t.ex. kirurgi de senaste
decennierna har lett till att fler patienter med CRLM är möjliga att leveroperera - och potentiellt botas.
För att kunna välja rätt behandlingsstrategi för varje enskild patient finns ett behov av att bättre kunna bedöma möjligheten till långtidsöverlevnad. Många olika förslag till
riskpoängsystem har tagits fram men deras nytta och signifikans är omdiskuterade. Studier har emellertid visat att hög systemisk inflammation i kroppen korrelerar med sämre prognos för överlevnad hos patienter med CRC. Hos patienter med CRLM är sambandet dock sämre studerat.
I denna studie undersöktes om de systemiska inflammationsmarkörerna C-reactive protein (CRP) och albumin, som tillsammans bildar Glasgow Prognostic Score (GPS), har ett samband med långtidsöverlevnad hos CRLM-patienter som genomgår kirurgi av
levermetastaser. Vi tittade retrospektivt på inflammationsmarkörerna inför leveroperation hos 849 patienter från Akademiska Sjukhuset i Uppsala och Karolinska Universitetssjukhuset i Huddinge, och jämförde med andra prognostiska faktorer.
GPS var den starkaste prognostiska faktorn av de inkluderade i studien; patienter med GPS 0 hade en medianöverlevnad på 62 månader jämfört med 27 månader hos patienter med GPS 2.
Systemisk inflammation - speglat av värdet på GPS - kan därför lämpligen beaktas vid val av
behandlingsstrategi hos CRLM-patienter. Den bakomliggande orsaken till sambandet mellan
systemisk inflammation och cancer kräver dock mer forskning. Kanske skulle cancerpatienter
med hög systemisk inflammation gynnas av att behandlas med antiinflammatoriska läkemedel
istället för att bara sikta in behandlingen mot tumörerna?
Background
Epidemiology
Colorectal cancer (CRC) is one of the most common cancers in the world. Yearly, there are 1.4 million new cases of CRC and it causes about 0.7 million deaths worldwide (1,2). The incidence varies ten-fold between different parts of the world, more commonly appearing in countries characterized by high income and development. In the Western world, it is the second largest reason of cancer death (1–3).
Up to half of CRC patients will develop metastases, and the liver is the most common place for metastases (3–5). Some decades ago, these CRC patients with liver metastases (CRLM) were seen as incurable as they had a 5-year survival of under 1% with supportive treatments (6). Big advancements have been made; today's multidisciplinary approach with cornerstone of aggressive surgical resections have led to a 5-year survival of about 50% after resection (7,8). The group of patients with primarily unresectable metastases can be reversed into resectable ones by increasingly used oncological treatment with chemotherapy and biological agents (9). In combination with better diagnostic imaging and improved surgical techniques such as ablation and pre-operative portal vein embolisation, this has led to increasing number of patients that are suitable for resection - and possibly can be cured (3,8–10).
Prognostic factors
In order to prognosticate survival and choose the right treatment strategy for each CRLM patient, a number of different clinical risk scores have been developed. Two examples are Nagashima and Fong scores, including factors such as nodal status of the primary tumour and number of metastases (11,12). In a review of risk score models, Spelt and co-workers
identified 15 models containing 25 different prognostic factors (13). Six of these prognostic factors were included in five or more models (Table I). Many of these clinical risk scores were developed before important advancements were made e.g. chemotherapy, that can influence the reliability of predictive risk scores (14,15). In addition, many of the risk scores have unweighted score systems, which means that they give the same point for every
prognostic parameter included and do not take into account that one parameter could be of
higher influence than another (13).
TABLE I. Prognostic factors often used in traditional risk scores for long-term survival in CRLM patients after liver surgery, identified by Spelt and co-workers(13)
Prognostic factor Number of studies
(of a total of 15) Number of liver metastases
Lymph node metastases in primary tumour Maximum size of metastases
Interval between CRC operation and detection of liver metastases Preoperative CEA level
12 10 6 5 5
Extrahepatic tumour 5
Today, patient response to neoadjuvant chemotherapy is used to prognosticate and select patients before liver surgery. Patients whose disease progress despite chemotherapy have a poorer outcome (7,16). Naturally, factors of general importance for survival such as comorbidity and age are included in the risk assessment. Prognostic factors presented in Table I are included in many of the older clinical risk scores but their significance today is discussed (13–15). In a review, Settmacher and co-workers suggest a simplified scoring system that divides CRLM patients into a low or high risk group (17). All patients with extrahepatic tumour at the time of liver surgery are regarded as high risk, as well as patients who have both three or more liver metastases and more than three lymph node metastases from the primary tumour. All other patients are regarded as low risk, and the final score is available only after the resection of liver metastases.
Consequently, there is still a need of better scoring systems that easily could be measured before liver resection in order to prognosticate survival, and thus be helpful in the decision of suitable patients for resection and choose the right treatment strategy for each patient. In a recent study from 2018, Sasaki and co-workers examine if a prognostic tool named Tumor Burden Score (TBS) has prognostic value in CRLM patients undergoing liver resection (18).
TBS was calculated based on number of metastases and size of largest metastasis, and the study showed that TBS was a strong prognostic tool for overall survival. Nevertheless, TBS needs further validation, which this study will look into.
Systemic inflammation and cancer
Cancer-associated inflammation is reported to have a big role in the progress and outcome of
tumours (19). In CRC and plenty of other cancer types; markers of systemic inflammation
have been reported to be a strong negative prognostic factor (20–23). Elevated C-reactive
protein (CRP) and low serum albumin level are often used as markers for systemic
inflammation. Together they form the Glasgow Prognostic Score (GPS) where patients can get 0, 1 or 2 points depending on their CRP and albumin body concentrations (Table II). The higher score a patient gets, the more systemic inflammation - and a more negative prognosis.
A modified version of GPS (mGPS) has been developed on the basis of evidence that
hypoalbuminemia itself has no significant association with cancer-specific survival (21,24).
TABLE II. Glasgow Prognostic Score (GPS, Modified Glasgow Prognostic Score (mGPS)
Factors GPS score mGPS score
CRP ≤ 10 mg/l and albumin ≥ 35 g/l Albumin < 35 g/l
CRP > 10 mg/l
0 1 1
0 0 1
CRP > 10 mg/l and albumin < 35 g/l 2 2
While the connection between systemic inflammation and poor survival prognosis in CRC and other cancer types is well described, the biological mechanisms behind the activation of systemic inflammation are not clarified (20–23). It may just be a reflection of a nonspecific immune response due to burden, damage and necrosis from the tumours. On the other hand, systemic inflammation could be the thing that drives tumour progression. Evidence suggest that systemic inflammation is associated with an increased activation of the innate immune system and down regulation of the adaptive responses, possibly leading to a more favourable growing environment for tumours (25,26). The mechanisms behind activation of systemic inflammation and its role in tumour growth remain unclear but could also possibly be a combination of these theories as well as other mechanisms.
In CRLM patients, the connection between systemic inflammation and survival prognosis is not as well researched as in CRC patients. Yet, in a recent study CRP was found to be the strongest independent predictor of survival, superior to factors such as number of liver metastases > 5 and size of largest liver metastases > 5 cm (27). Although that and some other recently published studies suggest a connection between systemic inflammation and survival prognosis, more studies are needed to validate it (27–29).
Aim and hypothesis
The aim of this study was to analyse the prognostic impact of the GPS factors CRP and
albumin in patients with colorectal liver metastases undergoing possible curative liver
surgery, and further to compare them with other prognostic factors. The hypothesis was that
high CRP value and hypoalbumineria before liver surgery indicate a poor long-term survival.
Methods
Patients
Patients with confirmed CRLM who underwent first resection of liver metastases with curative intent were included in this retrospective observational study. All patients were operated between 2005 and 2015 at Uppsala University Hospital in Uppsala, Sweden, or Karolinska University Hospital in Stockholm, Sweden.
Data for these patients were collected from patient records and local databases. Clinical and tumour-related data examined included: sex, centre of surgery (Stockholm or Uppsala), age at surgery, BMI, date of eventual death, location of primary tumour (colon or rectum),
embryonic origin of primary tumour (midgut or hindgut), size of largest metastasis, number of metastases, synchronous (defined as metastases detected within 6 months after the diagnosis of primary tumour) or metachronous (detected later than 6 months after primary tumour diagnosis) metastases.
Liver surgery related data examined included: date of surgery, ASA physical status classification before surgery, postoperative complications according to the Clavien-Dindo Score (30), length of hospital stay after surgery, continued care (defined as if patients were sent to another hospital after surgery), PVE (portal vein embolisation), peroperative ablation, two stage-surgery (liver surgery planned into two separate operations), resection type (minor
< 3 liver segment, major 3-4 segments, extended > 4 segments), blood lost during surgery and operation time.
As indicators of systemic inflammation, CRP and serum albumin levels taken within 3 months before liver surgery were collected. The GPS and mGPS value were calculated according to earlier mentioned score system (Table II). Patients who did not have both CRP and albumin value within 3 months before liver surgery were excluded from the study.
Tumor Burden Score (TBS) was calculated according to [TBS
2= (maximum tumour
diameter)
2+ (number of liver lesions)
2] and the patients were divided into 3 groups according to published thresholds(18). Group 1: TBS < 3; Group 2: TBS ≥ 3 and < 9; Group 3: ≥ 9.
The total number of CRLM patients that had underwent resection was 1212. After excluding patients who did not have preoperative CRP- and albumin value, the final number was 849.
The primary end-point of the study was overall survival after first liver surgery.
The study was approved by the Ethics Committee.
Data collection
Many of the variables mentioned above had already been collected from journals to a local database. However, my own main data collection was to find CRP and albumin values within 3 months before liver surgery in the patients' journals. I also double-checked if already collected values were correct.
Statistics
Data analyses were performed using MedCalc Statistical Software.
Continuous variables are presented with median and interquartile range (IQR), while the categorical variables are presented with proportions and percentages. Categorical variables were compared using the Chi square test, and the Kruskal-Wallis test was used for the continual.
The Kaplan-Meier method was used to analyse survival, and a log-rank test was used to examine differences between the groups. Cox proportional hazard regression was used to calculate univariate hazard ratios of survival predictors and stepwise method was used for multivariate analysis. The univariate and multivariate hazard ratios are presented with 95%
confidence intervals (CI). The level of statistical significance was defined as P < 0.05.
Results
Clinicopathological characteristics
Patients clinicopathological characteristics are presented in Table III.
The final number of patients included was 849, of whom 296 patients from Uppsala and 553
from Stockholm. The overall survival was 54 months, and the 5 year-survival was 46 %. Four
patients died within the first 30 days after surgery, and 17 died within the first 90 days. The
male-to-female ratio was 1.73:1. The primary tumour was more often located in colon (57%),
and its embryonic origin was from hindgut in 669 of cases (79%). The majority of the patients
had synchronous metastases (61%). Median number of metastases was 2 (IQR 1-4), and
median size of largest metastasis was 2,5 cm (IQR 1,8-4), resulting in a median TBS of 4,12
(IQR 2,69-6,08).
CRP, albumin and GPS
Median serum albumin was 36 g/L (IQR 34-38), and 271 patients (32 %) had hypoalbumineria (serum albumin < 35 g/L).
Median CRP was 2,5 (IQR 1-6), and 128 patients (15%) had a CRP level over 10 mg/l.
The GPS and mGPS were calculated as described earlier (Table II). 519 patients (61%) had a GPS of 0, while 261 (31%) and 69 ( 8%) patients had GPS 1 or 2, respectively. Regarding mGPS, a big majority (85%) had a mGPS of 0. Only fifty-nine (7%) and sixty-nine (8%) patients had mGPS 1 or 2, respectively.
The relationship between preoperative GPS score and other clinicopathological characteristics is shown in Table III. In a few words, the GPS score was significantly higher (GPS 2 vs. 1/0;
1 vs. 0) in patients with older age, lower BMI, larger maximal size of metastases and higher TBS score. Notably, GPS was not significantly related with number of metastases.
TABLE III. Clinicopathological characteristics of CRLM patients undergoing resection, and their relationships with the GPS
Variable Total GPS 0 GPS 1 GPS 2
n, if not given
other in () median (Q1-Q3) median (Q1-Q3) median (Q1-Q3) median (Q1-Q3) p- value
n (%) n (%) n (%) n (%)
Centre Stockholm 553 (65) 311 (60) 191 (73) 51 (74) ≤0.001
Uppsala 296 (35) 208 (40) 70 (27) 18 (26)
Sex Man 538 (63) 340 (66) 158 (61) 40 (58) 0.247
Age at liver
surgery (years) 66.9 (59.3-72.6) 66.3 (58.4-72.3) 67.0 (61.9-73.1) 69.4 (61.6-75.3) 0.026
Age >70 291 (34) 165 (32) 94 (36) 32 (46) 0.044
BMI (kg/m2) 25.7 (23.4-28.4) 26.0 (23.8-28.9) 25.0 (23.1-28.1) 24.7 (22.4-27.2) 0.002
Obese (BMI ≥ 30) 142 (17) 100 (19) 36 (14) 6 (9) 0.029
ASA 1 52 (6) 43 (8) 9 (3) 0 (0) <0.001
2 519 (61) 329 (63) 152 (58) 38 (55)
3 267 (31) 143 (28) 96 (37) 28 (41)
4 10 (1) 3 (1) 4 (2) 3 (4)
Primary tumour Colon 482 (57) 295 (57) 149 (57) 38 (55) 0.993
localization Rectum 358 (42) 219 (42) 109 (42) 30 (43)
Both 9 (1) 5 (1) 3 (1) 1 (1)
Embryonal
origin Midgut 173 (20) 93 (18) 64 (25) 16 (23) 0.196
Hindgut 669 (79) 422 (81) 194 (74) 53 (77)
Unclear 7 (1) 4 (1) 3 (1) 0 (0)
Metachronous
metastases 333 (39) 219 (42) 84 (32) 30 (43) 0.020
Metastases
number (#) 2 (1-4) 2 (1-4) 2 (1-4) 2 (1-4) 0.326
Maximal
metastasis size (cm) 2.5 (1.8-4.0) 2.3 (1.5-3.8) 2.8 (2.0-4.0) 4.0 (2.5-6.0) <0.001 Tumor Burden
Score
(#) 4.1 (2.7-6.1) 3.6 (2.5-5.4) 4.3 (2.9-6.1) 5.1 (3.3-9.3) <0.001
<3 264 (31) 185 (36) 67 (26) 9 (13) <0.001
≥3 and <9 498 (59) 291 (56) 168 (64) 41 (59)
≥9 87 (10) 43 (8) 26 (10) 19 (28)
Albumin (g/l) 36 (34-38) 38 (36-40) 33 (32-34) 31 (28-33) <0.001
Albumin <35 271 (32) 0 (0) 202 (77) 69 (100) <0.001
CRP (mg/l) 2.5 (1.0-6.0) 2.0 (1.0-3.6) 4.0 (2.0-9.0) 25.0 (15.0-49.3) <0.001
CRP >10 128 (15) 0 (0) 59 (23) 69 (100) <0.001
GPS 0 519 (61)
1 261 (31)
2 69 (8)
mGPS 0 721 (85)
1 59 (7)
2 69 (8)
Resection type Minor 484 (57) 290 (56) 153 (59) 41 (59) 0.692
Major 282 (33) 173 (33) 88 (34) 21 (30)
Extended 83 (10) 56 (11) 20 (8) 7 (10)
Two-stage surgery 17 (2) 8 (2) 7 (3) 2 (3) 0.445
PVE 36 (4) 22 (4) 11 (4) 3 (4) 0.986
Peroperative ablation 29 (3) 17 (3) 9 (3) 3 (4) 0.899
Operation time (min) 211 (149-280) 204 (145-272) 224 (170-286) 228 (170-303) 0.009 Blood lost (ml) 700 (350-1300) 700 (350-1300) 600 (325-1250) 650 (325-1400) 0.894 Clavien-Dindo
score None or 1 394 (46) 248 (48) 117 (45) 29 (42) 0.890
2 233 (27) 139 (27) 73 (28) 21 (30)
3a 128 (15) 76 (15) 42 (16) 10 (14)
3b 44 (5) 24 (5) 15 (6) 5 (7)
4a 24 (3) 16 (3) 7 (3) 1 (1)
4b 18 (2) 11 (2) 5 (2) 2 (3)
5 6 (1) 3 (1) 2 (1) 1 (1)
Clavien-Dindo ≥3a 220 (26) 130 (25) 71 (27) 19 (28) 0.770
Clavien-Dindo ≥3b 92 (11) 54 (10) 29 (11) 9 (13) 0.791
30-day mortality 4 (0) 2 (0) 2 (1) 0 (0) 0.640
90-day mortality 17 (2) 7 (1) 8 (3) 2 (3) 0.233
Length of
hospital stay (days) 10.0 (8.0-13.0) 10.0 (8.0-13.0) 10.0 (8.0-13.0) 10.0 (9.0-17.0) 0.011
Continued care 278 (33) 147 (28) 97 (37) 34 (49) <0.001
Overall survival
(KM) (months) 53,8 61,7 46,9 26,6 <0.001
1 year survival % 89 93 85 73
3 year survival % 64 69 59 43
5 year survival % 46 50 43 29
Prognostic impact of clinicopathological factors
The following factors were significantly associated with poorer survival in univariate analysis (Table IV): age over 70 (and as a continuous variable), embryonic origin from midgut,
metachronous metastases, total number of metastases (continuous variable), largest size of metastasis (continuous variable), TBS (both as a continuous variable and with cut-offs), length of hospital stay postoperatively (continuous variable), if the patients received
continued care at another hospital after surgery, extended resection type, two-stage surgery, peroperative ablation and Clavien-Dindo score >2.
In multivariate analysis, age over 70 (HR = 1.49, 95% CI 1.22-1.81), extended resection type (HR = 2.10, 95% CI 1.57-2.80), two-stage surgery (HR = 1.98, 95% CI 1.18-3.32) and peroperative ablation (HR = 1.90, 95% CI 1.23-2.93) remained significant factors of survival (Table V). Figure 2A-2D shows Kaplan-Meier survival curves for respective factor. Clavien- Dindo score, length of hospital stay and if the patients received continued care after surgery were not analysed in the multivariate analysis as predictors because of their nature when can be evaluated only after surgery and therefore not useful as prognostic factors before surgery.
Prognostic impact of CRP, albumin and GPS
In univariate analyses, CRP significantly correlated with reduced survival both analysed as a continuous variable (HR = 1.01, 95% CI 1.00-1.01) and at cut off > 10 mg/l (HR = 1.56, 95%
CI 1.24-1.98).
Similarly, albumin significantly correlated with survival as a continuous variable (HR = 0.98, 95% CI 0.96-0.996) and at cut-off < 35 g/l (HR = 1.40, 95% CI 1.16-1.70).
However, CRP and albumin together in the GPS were identified as the strongest prognostic factor in univariate analysis when patients with GPS 2 were compared to patients with GPS 0 (HR = 2.07, 95% CI 1.52-2.82, P < 0.001). The mGPS did not add prognostic value compared to the GPS.
In multivariate analysis, CRP and albumin by themselves were not statistically significant
correlated with survival. GPS on the other hand not just remained but improved its capacity as
a prognostic factor for survival, both when patients with GPS 1 were compared with GPS 0
(HR = 1.35, 95% CI 1.10-1.66, P 0.005) and when GPS 2 was compared to GPS 0 (HR =
2.16, 95% CI 1.1.55-3.00, P < 0.001). (Table IV)
TABLE IV. Cox proportional hazard regression univariate and multivariate analyses of possible prognostic factors in CRLM patients
variable Univariate Multivariate
HR (95% CI) p-value HR (95% CI) p-value
Centre Uppsala 1.10 (0.91-1.33) 0.341
Sex Man 1.15 (0.95-1.39) 0.152
Age at liver surgery
1.01 (1.01-1.02) 0.003 Age >70
1.47 (1.22-1.77) <0.001 1.49 (1.22-1.81) <0.001
BMI 1.00 (0.98-1.01) 0.544
Obese
1.15 (0.91-1.46) 0.242
ASA 1 0.73 (0.48-1.11) 0.139
3 1.09 (0.89-1.33) 0.403
4 1.57 (0.75-3.32) 0.236
Primary tumour
localization Colon
1.02 (0.85-1.23) 0.796 Multiple primary tumour 1.00 (0.42-2.41) 0.998 Embryonal origin Midgut 1.29 (1.04-1.61) 0.022
Metachronous 0.81 (0.67-0.98) 0.031
Metastases number
1.06 (1.03-1.08) <0.001 Maximal metastasis size
1.04 (1.00-1.07) 0.037 Tumor Burden Score 1.05 (1.03-1.08) <0.001
<3 0.70 (0.56-0.86) 0.001 ≥9 1.34 (1.01-1.79) 0.047
Albumin 0.98 (0.96-1.00) 0.021
Albumin <35 1.40 (1.16-1.70) <0.001
CRP 1.01 (1.00-1.01) 0.002
CRP >10 1.56 (1.24-1.98) <0.001
GPS 1 1.31 (1.07-1.60) 0.009 1.35 (1.10-1.66) 0.005
2 2.07 (1.52-2.82) <0.001 2.16 (1.55-3.00) <0.001
mGPS 1 1.27 (0.91-1.77) 0.164
2 1.93 (1.42-2.61) <0.001 Resection type Major 1.03 (0.84-1.26) 0.765
Extended 1.72 (1.29-2.29) <0.001 2.10 (1.57-2.80) <0.001 Two-stage surgery 2.03 (1.21-3.39) 0.007 1.98 (1.18-3.32) 0.010
PVE 1.41 (0.95-2.09) 0.088
Peroperative ablation 2.23 (1.45-3.42) <0.001 1.90 (1.23-2.93) 0.004 Operation time
1.00 (1.00-1.00) 0.038 Blood lost
1.00 (1.00-1.00) <0.001 Clavien-Dindo ≥3a 1.34 (1.10-1.63) 0.004 Clavien-Dindo ≥3b 2.04 (1.58-2.64) <0.001 Length of hospital stay
1.02 (1.01-1.02) <0.001
Continued care 1.21 (1.00-1.47) 0.048
To better understand the prognostic impact of the GPS, a survival curve with the Kaplan- Meier method was analysed (Figure 1). The log rank test demonstrated that survival time significantly differed between the three GPS groups (P < 0.0001). The median survival of patients with GPS 0 was 61.7 months compared to 46.9 months in patients with GPS 1, and 26.6 months in patients with GPS 2. The 5 year-survival for GPS 0 was 50% compared to 43% and 29% for GPS 1 and GPS 2, respectively.
FIGURE 1. Kaplan-Meier curves showing the relationship between preoperative GPS score and overall survival. (log rank test, p < 0.0001)
Discussion
Preoperative elevated CRP and hypoalbuminemia, merged in the GPS, were found to be a strong independent prognostic factor for long-term survival in patients with colorectal liver metastases undergoing resection of liver metastases. Patients with a preoperative score of GPS 2 had a median survival of 26.6 months compared to 61.7 months in patients with preoperative GPS 0. Our results suggest that the presence of systemic inflammation, expressed by GPS preoperatively, is an important factor of disease outcome.
The results are in line with previous studies investigating the prognostic role of systemic inflammation markers in several cancer types including CRC - and partly CRLM (20–24, 27–
29).
In the present study, the GPS had greater prognostic value than CRP and albumin by
themselves. This in contrast to the earlier mentioned study by Køstner and co-workers where
(months)
CRP itself had the strongest prognostic impact (27). In that study, 25% of patients had hypoalbuminemia in contrast to 32 % in our study. That might point to a wider selection of patients suitable for liver surgery in our study. Hypoalbuminemia is, in addition of being a marker of systemic inflammation, also an indicator of malnutrition(31). Our data showed a significant connection between lower BMI and a higher GPS score, which may support that malnutrition is a part of the GPS risk assessment.
Patients with a preoperative GPS 2 had the highest hazard ratio of all variables in multivariate analysis, including age >70 and patients operated with advanced surgical techniques:
extended resection, two-stage surgery or peroperative ablation. That is remarkable since patients requiring these combined operation strategies usually present with higher TBS and due to complex metastasis situation are considered to be most challenging. GPS was superior to size of largest metastasis, number of metastases and TBS as a prognostic factor. At the same time combined operation strategies needed to deal with complex situation has higher risk of postoperative complications and has worse survival prognosis. In other words, complexity of the liver metastasis localization and appropriate complex surgical procedure needed to get tumour free situation is more important than the tumour number, maximal size or TBS for prognosis of the patient.
FIGURE 2. Kaplan-Meier curves showing the relationship between overall survival age
> 70 (A), resection type (B), two-stage resection (C) and peroperative ablation (D). (log rank test, for all with P < 0.0001)
Time (months) Time (months)
Time (months) Time (months)
