The G8 screening tool detects relevant geriatric impairments and predicts survival in elderly patients with a haematological malignancy
- First Online:
- Cite this article as:
- Hamaker, M.E., Mitrovic, M. & Stauder, R. Ann Hematol (2014) 93: 1031. doi:10.1007/s00277-013-2001-0
- 864 Views
The G8 screening tool was developed to separate fit older cancer patients who were able to receive standard treatment from those that should undergo a geriatric assessment to guide tailoring of therapy. We set out to determine the discriminative power and prognostic value of the G8 in older patients with a haematological malignancy. Between September 2009 and May 2013, a multi-dimensional geriatric assessment was performed in consecutive patients aged ≥67 years diagnosed with blood cancer at the Innsbruck University Hospital. The assessment included (instrumental) activities of daily living, cognition, mood, nutritional status, mobility, polypharmacy and social support. In parallel, the G8 was also administered (cut-off ≤ 14). Using a cut-off of ≥2 impaired domains, 70 % of the 108 included patients were considered as having an impaired geriatric assessment while 61 % had an impaired G8. The G8 lacked discriminative power for impairments on full geriatric assessment: sensitivity 69, specificity 79, positive predictive value 89 and negative predictive value 50 %. However, G8 was an independent predictor of mortality within the first year after inclusion (hazard ratio 3.93; 95 % confidence interval 1.67–9.22, p < 0.001). Remarkably, patients with impaired G8 fared poorly, irrespective of treatment choices (p < 0.001). This is the first report on the clinical and prognostic relevance of G8 in elderly patients with haematological malignancies. Although the G8 lacked discriminative power for outcome of multi-dimensional geriatric assessment, this score appears to be a powerful prognosticator and could potentially represent a useful tool in treatment decisions. This novel finding certainly deserves further exploration.
KeywordsElderlyGeriatric assessmentPrognosticationHaematological malignancy
Haematological neoplasms in elderly patients represent a major challenge. The median age at diagnosis for frequently diagnosed and relevant diseases such as acute myeloid leukaemia, multiple myeloma and chronic lymphocytic leukaemia is around 70 years while for myelodysplastic syndromes, the median age is another 5 years higher . With the imminent ageing of society, haematologists will be faced with increasing numbers of older patients. Treatment guidelines are based on clinical studies from which the elderly and those with comorbidity have often been excluded , making it uncertain if their findings are also applicable to the older patient. As a result, there is a great need for individualised treatment algorithms and programmes designed specifically for the elderly .
In oncology, geriatric assessments are used to detect disabilities and comorbid conditions that potentially contribute to an older patient’s vulnerability, predisposing to poor outcome and treatment complications [4, 5]. However, such assessments are time-consuming, leading cancer specialists to seek a short screening tool that can separate fit older cancer patients, who are able to receive standard cancer treatment, from vulnerable patients that should subsequently receive a full assessment to guide tailoring of their treatment regimen .
The G8 questionnaire
Possible responses (score)
Has food intake declined over the past 3 months due to loss of appetite, digestive problems, chewing, or swallowing difficulties?
0 = Severe decrease in food intake
1 = Moderate decrease in food intake
2 = No decrease in food intake
Weight loss during the last 3 months?
0 = Weight loss >3 kg
1 = Does not know
2 = Weight loss between 1 and 3 kg
3 = No weight loss
0 = Bed or chair bound
1 = Able to get out of bed/chair but does not go out
2 = Goes out
0 = Severe dementia or depression
1 = Mild dementia
2 = No psychological problems
Body mass index (BMI)? (weight in kilograms) / (height in square metres)
0 = BMI <19
1 = BMI 19 to <21
2 = BMI 21 to <23
3 = BMI ≥23
Takes more than three prescription drugs per day?
0 = Yes
1 = No
In comparison with other people of the same age, how does the patient consider his/her health status?
0.0 = Not as good
0.5 = Does not know
1.0 = As good
2.0 = Better
0 = >85
1 = 80–85
2 = <80
Total score 0–17
Cut-off ≤ 14
Most studies on the discriminative power of screening tools were performed in patients with solid tumours or in heterogeneous populations including both solid and haematological malignancies. Data specifically addressing the use of G8 in elderly patients with blood cancer are lacking. Furthermore, no study has thus far addressed the association between G8 and patient outcomes. Therefore, we set out to assess the value of G8 for elderly patients with a haematological malignancy, both in identifying patients with impairments in their multi-dimensional geriatric assessment and in predicting 1-year mortality.
As of September 2009, all consecutive patients aged 67 years and older newly diagnosed with a haematological malignancy at the Department of Internal Medicine V (Haematology and Oncology), Innsbruck Medical University, Innsbruck routinely received a multi-dimensional geriatric assessment. This assessment was performed by specifically trained research assistants. This study was approved by the local ethics committee. Written informed consent was obtained from all patients included.
Content of multi-dimensional geriatric assessment including cut-off value for impairments
Barthel index 
Lawton and Brody 
Number of medications
Body mass index and/or
Recent weight loss
≥1 kg in the last 3 months
In addition to the geriatric assessment data, the following factors were included in this analysis: age at diagnosis, sex, World Health Organisation (WHO) performance status, comorbidity, haematological diagnosis based on the WHO-classification  and tumour characteristics. Comorbidity was scored with the cumulative illness rating scale-geriatric version (CIRS-G) , which scores the severity of disease on a scale of 0 (no problems ) to 4 (extremely severe problem requiring immediate treatment or causing severe functional impairment) across 14 organ systems, resulting in a score between 0 and 56. Tumour prognosis was classified as favourable or unfavourable based on standardised guidelines; an overview of this classification per tumour type can be found in the Appendix. An initial treatment decision was made by a minimum of two haematologists based on treatment guidelines combined with clinical judgement regarding the patient’s ability to tolerate treatment; complex cases were also discussed at an internal multi-disciplinary haematological tumour board. For this study, treatment was classified as standard treatment, attenuated treatment, no treatment despite being required or no treatment needed at the time of inclusion. Criteria for this classification are also listed in Appendix. Last-known date to be alive or date of death were extracted from the official mortality file for Tirol/Austria applying a well-established record linkage method .
The sensitivity, specificity, positive predictive value and negative predictive value of the G8 screening tool for detecting the presence of an impaired multi-dimensional geriatric assessment (defined as impairments in two or more geriatric domains) were calculated from a 2 × 2 table.
For comparisons between patients with a normal G8 score and those with an impaired G8, the chi-square test and risk analysis were used for nominal and ordinal variables, as well as for continuous variables with a non-normal distribution; for continuous variables with a normal distribution, the Student’s t test was used. A p value of 0.05 was considered as statistically significant.
This analysis focuses on 1-year mortality. To determine which factors were associated with mortality in the 12 months following assessment, a Cox regression analysis was performed. For each variable, the Cox proportional hazards assumption was tested using the log minus log plot. Next, a univariate Cox-regression analysis was performed to determine which variables were associated with mortality in the 12 months following admission. To assess whether factors which are significantly associated in the univariate analysis retained their significance after correction for confounders, a subsequent multi-variable analysis was performed, including age, sex, haematological diagnosis, unfavourable tumour characteristics, comorbidity, initial treatment decision and performance status.
The SPSS (Statistical Package for the Social Sciences) version 21.0 was used for the analyses.
Normal G8 score
Impaired G8 score
n = 108
n = 42
n = 66
Median age (range)
Percentage of female
25 (23 %)
11 (26 %)
14 (21 %)
Acute myeloid leukaemia
31 (29 %)
11 (26 %)
20 (30 %)
5 (5 %)
2 (5 %)
3 (5 %)
13 (12 %)
6 (14 %)
7 (11 %)
31 (29 %)
11 (26 %)
20 (30 %)
3 (3 %)
1 (2 %)
2 (3 %)
None given despite indication
Not required at inclusion
Median total CIRS-G score (range)
WHO performance status ≥2
The prevalence of geriatric conditions was high: polypharmacy was seen in 65 %, IADL impairment in 45 %, risk of malnourishment in 45 %, impaired mobility in 24 %, symptoms of depression in 24 %, ADL impairments in 20 %, limited social support in 21 % and cognitive dysfunction in 17 %. Overall, 92 % of patients had impairments in at least one geriatric domain: one impairment was seen in 23 %, two in 23 %, three in 17 %, and 29 % of patients had four or more impairments. In addition, G8 was impaired in 61 % of patients.
At the time of inclusion, 7 % required no treatment and 67 % received standard treatment regimens. Two patients died before the treatment could be started. In 7 % of patients, the decision was made not to initiate anti-cancer therapy despite having an indication for treatment, and 17 % received an attenuated haematological treatment. Reasons for not receiving standard treatment were comorbidities or poor performance status in 22 patients, including two patients with severe cognitive impairment, advanced age in one patient and at patient’s request in two cases.
Initial treatment decisions did not appear to be associated with G8 outcome (p = 0.17); of 97 patients having an indication for treatment at the time of inclusion, 34 had normal G8 and received standard treatment and 39 patients had an impaired G8 and received normal treatment. Of the 25 patients receiving adjusted treatment, six had a normal G8, five of these had serious comorbidity and one refused standard treatment.
Discriminative power of G8
The sensitivity of G8 for an impaired multi-dimensional geriatric assessment (≥2 impairments) was 69 % (95 % confidence interval (CI) 58–78 %) with a specificity of 79 % (95 % CI 62–91 %). The negative predictive value was 50 % (95 % CI 55–79 %), meaning that almost half of the patients with normal G8 actually had impairments in at least two geriatric domains, thus being incorrectly identified as fit (Fig. 1). Reversely, the positive predictive value was 89 % (95 % CI 79–98 %), meaning that 11 % of the patients with an impaired G8 did not have any geriatric impairments in the full assessment.
Overall, G8 demonstrated the best sensitivity for ADL impairment, depressive symptoms, and nutritional status, with, respectively, 9, 8 and 12 % of patients with issues in these domains having a normal G8 score. By comparison, 22 % of the patients with IADL care dependence had a normal G8, as did 19 % with impaired mobility, 34 % with social issues, 17 % with cognitive impairment and 34 % with polypharmacy.
Prognostic value of G8 and other geriatric assessments
Factors associated with 1-year mortality
95 % Confidence interval
95 % Confidence interval
Unfavourable tumour prognosis
WHO performance status
Impaired geriatric assessment
In this study, we set out to determine the value of G8 for elderly patients with a haematological malignancy. We found that this tool lacked the discriminative power for impairments on a multi-dimensional geriatric assessment, with a sensitivity of 69 %, a specificity of 79 % and a negative predictive value of 50 %. However, G8 was an independent predictor of mortality within the first year after inclusion, with a hazard ratio of 3.93 (95 % CI 1.67–9.22). Remarkably, patients with impaired G8 fared poorly, irrespective of treatment choice. Of geriatric domains, impaired mobility and risk of malnutrition were also found to be associated with mortality.
The patients in this study had a high prevalence of geriatric impairments, with 92 % of patients having issues in one or more geriatric domains. It is likely that this fact negatively affected the discriminative power of the G8 screening tool, as it has been demonstrated that a higher prevalence of the target condition will lower the sensitivity of a diagnostic tool while at the same time increasing the specificity . Previous studies addressing the G8 in relation to a more elaborate geriatric assessment used varying numbers of domains and varying definitions of when the outcome was labelled as impaired . These factors will impact the prevalence of an impaired assessment and, subsequently, the discriminative power of screening tools aimed at identifying this impairment. This could explain why these studies generally found a good sensitivity and good negative predictive value of G8, while we found that half of patients with normal G8 still had significant geriatric impairments (Fig. 1).
The G8 was developed to avoid a time-consuming elaborate assessment in patients for whom it is unlikely to yield new information . However, in our study it is questionable whether using a two-stepped approach with a screening tool preceding the full assessment would actually have been beneficial: only 8 % of patients did not have any geriatric impairments and almost 70 % of patients had at least two impairments. With so many issues being uncovered, many of which provide opportunities for interventions aimed at optimizing the patient’s health status or quality of life [20-22], perhaps all elderly patients deserve a thorough assessment prior to initiation of treatment.
However, although using the G8 in a two-stepped approach to geriatric assessment may not be optimal, this does not render this tool redundant. The significant association of impaired G8 with poor prognosis, also found in a recent publication by Kenis et al.  and the lack of predictive value of traditional prognosticators such as performance status, demonstrates that the G8 uncovers vulnerabilities within the patient that are not revealed with standard care. Moreover, as demonstrated by Fig. 2, G8 was a more powerful predictor of prognosis than treatment decision based on clinical judgement. These are very similar to what was found by Tucci et al. in a study of 84 elderly patients with diffuse large B cell lymphoma, which demonstrated that a simplified geriatric assessment was more effective than clinical judgement in identifying which patients would benefit from aggressive (i.e. standard) chemotherapy regimens, which were as identified as unfit based on geriatric parameters and were similar irrespective of receiving standard or attenuated treatment . This study also found that toxicity-related deaths and response to treatment were more strongly associated with these parameters than with clinical judgement.
These results suggest that the combination of clinical judgement with geriatric parameters such as the G8 score is able to identify patients with a good prognosis under standard treatment, while the intrinsic risk of poor outcome for vulnerable patients is not altered significantly by treatment choice. However, as treatment assignment in our study was not at random, we do not know how the outcome would have been affected if all patients with impaired G8 had received attenuated therapy. Potentially vulnerable patients would have benefited from more cautious treatment regimens, as their risk of treatment-related morbidity and mortality is greater . On the other hand, perhaps these patients would have incurred an even worse prognosis if suboptimal dosing or withholding of treatment was added to their intrinsic vulnerability. Reversely, many valid reasons exist for offering patients with attenuated treatment instead of standard care; in our study, patient numbers were too small to determine if G8 also has prognostic value in that patient subgroup. These questions are highly relevant to future treatment guidelines and, thus, deserve further exploration.
In our study, the geriatric domains with the strongest predictive value for one-year mortality were nutritional status and mobility. Both factors are closely related to the frailty phenotype as defined by Fried et al. . Several prior studies have found nutritional status to be a strong predictor of prognosis [27-32] and although results for mobility are a little more variable [32-39], the majority of studies also mirror our findings of poorer outcome with poorer mobility. Further support comes from a large pooled analysis including over 34,000 community-dwelling adults aged 65 years and older, which found that survival increased across the full range of gait speeds, with significant increments per 0.1 m/s . While nutritional status appears to reflect the severity of disease , gait speed could be considered a summary indicator of a patient’s vitality because it integrates known as well as unrecognised disturbances across multiple organs systems such as the heart, lungs, circulatory, nervous and musculoskeletal system . Slowing gait could thus reflect damage in these systems and indicate a vicious cycle of reduced physical activity and deconditioning that has a direct effect on health and survival .
Our study is the first to examine the prognostic value of the G8 screening tool in haematological malignancies. However, it also has some limitations. First of all, we determined the prognostic value of G8 in patients deemed fit for standard treatment by haematologists. As there are as no golden standard for fitness yet, this is now in part based on the clinical judgement of the haematologists involved. If patients are selected differently, this may also affect the value of G8. As this study was performed at a single centre, the value in other centres with different methods of stratifying patients as fit or unfit may be different. Thus, further confirmation is needed. Secondly, although we included only patients with a haematological malignancy, the patient population was still heterogeneous in terms of disease subtypes and prognostic subgroups. Despite incorporating these factors in the multi-variate analyses, this could affect the possibility of extrapolating these results to the individual patient. Furthermore, not all haematologists were blinded to the outcome of geriatric assessment at the time of treatment decision-making. However, although the proportion of patients receiving attenuated treatment was somewhat higher in patients with impaired G8 (23 vs. 7 % in patients with normal G8), this difference was not statistically significant (p = 0.17) and similar results were found for the number of geriatric deficits (p = 0.26). In addition, the associations between geriatric domains and outcome were all corrected for treatment choice. Finally, due to limited duration of follow-up, this study focused only on short-term (i.e. 1 year) survival. We do not yet know whether the associations we found are also predictive of longer-term prognosis. However, given the great differences found in median survival times (Fig. 2), longer follow-up is unlikely to alter the prognostic value of G8.
In conclusion, in a patient population with such a high prevalence of geriatric impairments, using the G8 in a two-stepped approach to the geriatric assessment does not appear beneficial. However, G8 appears to be a powerful prognosticator and could potentially be a useful tool in treatment decisions. This is a novel finding that certainly deserves further exploration.
This study was financially supported by Verein Senioren-Krebshilfe.
Conflicts of interest
No potential conflicts of interest related to this manuscript.
Informed consent was obtained from all patients for being included in the study.