Comparison of two frailty screening tools in older women with early breast cancer

1. Introduction 

Over 25% of centenarians are autonomous in all 6 basic activities of daily living. However, for a large majority of older adults, the gradual deterioration in physical and functional status that occurs with aging may decrease independence and can result in premature death.

In Spanish populations, frail elderly individuals account for between 10% and 20% of patients over 65 years of age [1] and up to 46% of patients over 85 years of age [2], although these values are highly dependent on the criteria used to define frailty and the type of community being studied.

The term “frailty” is highly controversial. The concept is used in a broad sense to describe a state of high vulnerability to adverse health events [3]. The term was often used in medical literature in the 1980s, but its meaning is not well defined. The term frailty has the following connotations: being dependent on others, having substantial risk for developing or currently having multiple health problems, gradually losing “functional reserve”, having chronic diseases that are both medically and psychosocially complex and having “atypical” forms of disease presentation. This population could benefit from specialized geriatric programs. It is highly debated and not decided whether frailty is reversible.

Several studies have shown that frailty has serious prognostic implications, which are reflected by the higher incidence of disability, the higher rate of hospitalization and the greater number of prescription drugs used. Together, these prognostic factors result in the imposition of higher financial and caregiving burdens [4], [5].

To date, there is a lack of consensus concerning the definition of this concept among the elderly. For some authors, the coexistence of certain clinical conditions is the hallmark of frailty, while for others, it is the lack of independence in activities of daily living and the need for institutional care that define the term. Fried et al. hypothesized that frailty is a wasting syndrome characterized by weakness, lack of strength, low energy, physical sluggishness and a low activity level. In this work, the authors reported that the presence of these factors predict an early death [6]. Fried et al. developed an operational definition of frailty based on the presence of at least 3 of 5 criteria: unintentional weight loss (more than 4.5kg), exhaustion weakness (≤20th percentile by force dynamometer), low physical activity (≤20th percentile in kcal/week adjusted for gender) and slowness (≤20th percentile in the time taken to walk 5m). In this study, the cumulative mortality rates at 3 years were 3% and 18% in groups of patients without and with frailty, respectively. Further, the deterioration in ability to perform activities of daily living (ADL) 7 years after initial assessment were 23% and 63% and the rates of emergence of first decline at 3 years were 15% and 28%, respectively.

Recently, members of the Study of Osteoporotic Fractures Research Group have attempted to determine whether the CHS index devised by Fried and used thus far to define the syndrome of frailty in the elderly, could be replaced by a simple index with comparable predictive value, the SOF (Study of Osteoporotic Fractures). The SOF consists of only 3 components: intentional or unintentional weight loss of 5% or more, ability to lift oneself from a chair 5 times without using one's arms, and self-reported reduced energy level (answer to the question “Do you feel energetic?”). Older women who do not exhibit any of these 3 components are described as “not fragile”, those exhibiting 1 component are described as “pre-frail”, and those exhibiting all 3 components are considered “fragile”. The study concluded that the SOF index, which is a simplified method for determining frailty in the elderly due to its use of only three components, predicted the risk of falls, disability, fractures and death with the same efficiency as the CHS index, which is a more complex index using five components [7].

In oncology, many treatments are aimed at improving survival, autonomy and quality of life for patients, without a curative intent. Such treatments must be chosen appropriately, as their therapeutic indices (i.e., the window wherein people benefit without experiencing unacceptable side effects) may be low. Recently, research in oncology has begun to elucidate the factors that influence the ability of one's body to return to a state of homeostasis after the stress of chemotherapy treatment or aggressive surgical intervention [8].

Although there is no clear universal definition of frailty [9], [10], [11], [12], it is generally thought that the inability to engage independently in basic activities of daily living, which include bathing, continence, feeding, transferring, toileting and dressing is synonymous with frailty [13]. Indeed, research in this population suggests that loss of independent functioning in these areas signals the inability to accommodate circumstances that involve stress, even mild stress [13]. Dependence in instrumental activities of daily living, or core activities, is associated with increased mortality [14]. In addition, dependence on some instrumental activities predicts the development of dementia [15] and an increased risk of toxicity associated with chemotherapy [11].

From the literature on frailty, two points are evident. First, while frailty results in a dramatic change in functional status, this change progresses in a chronic fashion. Second, the median survival of a fragile patient is decreased by 2 years [16]. Frail elderly cancer patients need effective palliation, which may include low-dose chemotherapy [13].

Given the ambiguity of the definition of frailty, many have proposed a set of scales and screening items that aim to achieve a better definition and more accurately quantify this syndrome. In addition, because a full Comprehensive Geriatric Assessment (CGA) is time consuming, these screening tests permit the identification of patients who may benefit from a full CGA. It is unclear which index yields the best definition of the term ‘frailty’ and hence is preferable for use in research.

Of the screening tests currently in use, the Vulnerable Elderly Survey (VES-13), the Timed Up and Go Test, the 7-item physical performance test and the Barber Questionnaire are the most widely used. The VES-13 assesses age, functional status and activity. If the score is equal to or greater than 3, there is an increased risk of functional impairment and the patient would likely benefit from the implementation of a full CGA [17]. The VES-13 is a self-administered survey consisting of 1 item relating to age and 12 additional items relating to self-perception of health status, functional ability and physical fitness [17], [18]. The test “Timed Up and Go” consists of measuring the time required for an elderly person to get up, take a few steps, turn around and sit down again. The CGA is applied for patients who require more than 10s to perform the exercise, use their arms to get up or make a wrong turn [19], [20]. Of all of the described instruments, the Barber Questionnaire was created in England in the early 1980s, initially as a postal questionnaire, with the intent to identify elderly persons at risk for dependence [21], [22]. Other authors have subsequently modified the questionnaire with regard to both the content and the route of administration. It is currently the most widely used questionnaire in Spain for the identification of at-risk elderly [23]. It consists of 9 questions. The result is considered positive if a patient answers yes to one or more of the questions. The questionnaire is often used as a screen (by questionnaire) that identifies people at risk or as a further evaluation in order to inform intervention efforts. Its major applications are in the elderly.

2. Materials and methods 

2.4. Basic design 

In this study, patients older than 65 years of age with an indication for chemotherapy in the first visit were given two frailty screening tests, the Barber Questionnaire and the Vulnerable Elderly Survey. They were also given a standard CGA, which was developed previously in our center. The CGA includes a number of fundamental domains: functional assessment scales, comorbidity, number of prescription drugs, social support and nutritional screening. Deficits in any of these domains can negatively impact the condition of elderly individuals. The cut-off points that define the deficits in each of these assessment tools have been prospectively associated with an increased risk of disability or death in elderly persons who are not admitted to a hospital (Table 1).

Table 1. Glossary of the different tests used.

	Test	Dominio geriátrico	Number of items	Range	Cut-off score of adverse events
	Barber Questionnaire	Screening tool	9	0–9	≥1
	VES-13	Screening tool	4	0–10	≥3
	Barthel scale	Function	10	0–100	≤60
	Lawton–Brody scale	Function	7	0–14	≥12
	Charlson index	Comorbidity	18	0–54	>10
	Gijón Social scale	Social support	3	0–12	≥8
	Pfeiffer test	Cognitive level	10	0–10	>3 (falls)
	NSI scale	Risk of desnutrition	10	0–21	≥2
	Number of medications	Comorbidity/toxicity potential from medication interactions	1	0–∞	>4
	ECOG	Baseline status	5	0–5	≥2

The diagnostic accuracies of the Barber Questionnaire and the VES-13 were determined by comparing scores on these assessments with those obtained by administering the CGA to the study sample. All surveys were administered by one person, the principal investigator. The results of each of the scales and the CGA screening were recorded. In addition, each test was scored on a dichotomous scale, based on whether there was or not impairment in any of the parameters. Based on previous studies, the detection of deficits in ≥2 of the CGA scales indicates an increased risk of disability or death [14], [25], [26], [27], [28], [29], [30]. A score above 0 on the Barber Questionnaire [22] or a score at or above 3 on the VES-13 [17] also indicates an increased risk of disability or death. Therefore, in this study, the risk of frailty was defined as the existence of deficits on ≥2 of the individual questionnaires that comprise the CGA [26], [31], [32], [33], [34], a score >0 on the Barber Questionnaire or a score of ≥3 on the VES-13 questionnaire, with the CGA being the gold standard for subsequent comparisons. Participation in the study was completely voluntary.

3. Results 

3.1. Characteristics of patients enrolled in the study 

The group consisted of 41 patients, all of whom were female. The average age at diagnosis was 74.5 years, with a range of 66.5–87.5 years. Half of the population was between 75.32 and 82.18 years of age. The majority of the patients (78%) were over 70 years of age.

Of the sample, 78% were married and 17.2% were widowed. Regarding education, 68.3% did not surpass elementary educational level, only one of them had a school diploma and there were nobody with university studies (Table 2).

Table 2. Patient and tumor characteristics.

		Nu	%
	Age
	Average/standard deviation	74.5 (5.1)
	Median/range	74.5/(66.5–87.5)
	<70 years	9	22.0%
	>/70 years	32	78.0%
	Civil status
	Alone	1	2.4%
	Married	32	78.0%
	Separated	1	2.4%
	Widow	7	17.2%
	Educational level
	Illiterate	28	68.3%
	Primary studies	12	29.3%
	School diploma	1	2.4%
	Histological subtype
	Ductal	33	80.5%
	Lobulillar	7	17.1%
	Medullar	1	2.4%
	Tumoral stage
	pT
	T1	26	53.7%
	T2	17	41.5%
	T3	1	2.4%
	T4	1	2.4%
	pN
	N0	12	29.4%
	N1a	26	63.3%
	N2	2	4.9%
	Estrogen receptors
	<5%	6	14.6%
	>5%	35	85.4%
	Progesterone receptors
	<10%	8	19.5%
	>10%	33	80.5%
	Her-2 status
	Negative	32	78.1%
	Positive	8	20.5%
	Unknown	1	2.4%

3.2. Descriptive study of tumor characteristics and treatment 

The most frequent histological diagnosis was infiltrating ductal carcinoma (80.5%). There was only 1 case of medullary cancer. T1–T2 tumors (n=39) predominated; the tumor affected the chest wall or skin (T4) in only one case. Twelve women had no axillary involvement and 24 women had only 1–3 affected lymph nodes (N1a). The majority of the tumors were hormone receptor-positive (35 cases with estrogen receptor positivity and 33 cases with progesterone receptor positivity) and most did not overexpress HER2 (n=32, 78.1%) (Table 2).

In 18 cases, surgical treatment was radical (n=18). The vast majority of patients were treated with an anthracycline-based scheme (n=31), while 9 were treated with the traditional regimen of cyclophosphamide, methotrexate and fluorouracil (CMF). In one case, a taxane was used alone (Table 3).

Table 3. Treatment of the tumor.

		Nu	%
	Kind of surgery
	Radical surgery (mastectomy)	18	43.9%
	Conservative surgery	23	56.1%
	Adjuvant chemotherapy
	EC×4–TXT×4	17	41.5%
	FEC90×6	10	24.4%
	CMF×6	9	21.9%
	AC×4	4	9.8%
	Paclitaxel×12	1	2.4%

EC: Epirubicin–Cyclophosphamide; TXT: docetaxel; FEC: 5 Fluorouracil–Epirubicin–Cyclophosphamide; CMF: Cyclophosphamide–Metotrexate–5 Fluorouracil; and AC: Doxorubicin–Cyclophosphamide.

3.3. Analysis of geriatric assessment scales 

Of the total sample, 26 patients were independent in basic activities of daily living (63.4%) and 22 were independent in instrumental activities (53.7%). Only one patient was in social risk and none had a social problem, as assessed by the Gijón socio-family scale. Nineteen of the patients (46.3%) had no comorbidity and 17 patients scored only one point on the Charlson comorbidity scale. The majority (80.5%) of the sample had no cognitive impairment, as measured by the Pfeiffer scale, while the remaining older adults with cognitive impairment were accompanied by their families, thus allowing for administration of the comprehensive geriatric assessment. Of the sample, 65.9% (n=27) had either no or slight nutritional risk. More than half of the patients (56.1%; n=23) did not take any medication and, of those who did, the number of drugs taken did not exceed 3. Baseline status corresponded to an ECOG of 0 for 46.3% of patients and an ECOG of 1 in 41.5% of patients (Table 4).

Table 4. Elements of the CGA.

		Nu	%
	Activities of daily living (Barthel)
	Independent in ADL	26	63.4%
	Dependent in ADL	15	36.6%
	Average/standard deviation	95.7/10.6
	Instrumental activities of daily living (Lawton–Brody)
	Independent in IADL	22	53.7%
	Dependent in IADL	19	46.3%
	Average/standard deviation	10.0/3.9
	Social valoration (Gijón)
	Normal (<8)	40	97.6%
	Social risk (8–11)	1	2.4%
	Social trouble (12+)	–
	Average/standard deviation	3.5/2.2
	Comorbidity (Charlson index)
	0	19	46.3%
	1	17	41.5%
	2	4	9.8%
	3	1	2.4%
	Average/standard deviation	0.7/0.8
	Cognitive evaluation (Pfeiffer test)
	No intellectual deficit	33	80.5%
	Cognitive deficit	8	19.5%
	Average/standard deviation	8.6/1.9
	Risk of desnutrition (NSI scale)
	No risk of low risk	27	65.9%
	Medium–high risk	14	34.1%
	Average/standard deviation	2.0/1.8
	Number of medications
	No medication or low consumer (0–3)	23	56.1%
	Moderate–high consumer (>4)	16	39.0%
	Average/standard deviation	3.4/2.8

Thirteen patients showed no deficits in any of the scales employed by the CGA and 4 patients had more than 4 scales indicating pathology. Nineteen patients (46.3%) had deficits in two of the CGA scales, while the remaining 22 had deficits in more than two of the scales.

3.4. Analysis of screening questionnaires of frailty 

The mean score on the Barber Questionnaire was 0.73 (range 0–5). Of the patients, 58.24% (n=24) scored a zero on that scale and only one had a score of 5. By contrast, the average score achieved on the VES-13 scale was 1.95 (range 0–10), with 14 patients (34.1%) who scored zero on this one (Table 5).

Table 5. Analysis of the frailty screening tests.

	Barber Questionnaire
	Score 0	24	58.5%
	Score 1	11	26.8%
	Score 2	3	7.3%
	Score 4	2	4.9%
	Score 5	1	2.4%
	Average/standard deviation	0.7/1.2
	VES-13 scale
	Score 0	14	34.1%
	Score 1	9	22.0%
	Score 2	6	14.6%
	Score 3	5	12.2%
	Score 4	3	7.3%
	Score 5	1	2.4%
	Score 8	2	4.9%
	Score 10	1	2.4%
	Average/standard deviation	1.9/2.5

The item that rated the highest was item 4 on the Barber Questionnaire (“Are there any days when you are unable to have a hot meal?”), with 26.8% of participants responding affirmatively, followed by item 6 (“Is there anything about your health causing you concern or difficulty?”), which was indicated by 6 patients (14.6%), followed by item 5 (“Are you confined to your home through ill health?”) and item 1 (“Do you live on your own?”), to which 4 patients responded affirmatively. On the contrary, item 2 (“Are you without a relative you could call on for help?”) and item 9 (“Have you been in hospital during the past year?”) did not receive any positive responses (Table 6).

Table 6. Items in the Barber Questionnaire.

	Item	Positive answer	Prevalence
	1. Do you live on your own?	4	9.8%
	2. Are you without a relative you could call on for help?	0	0
	3. Do you depend on someone for regular help?	3	7.3%
	4. Are there any days when you are unable to have a hot meal?	11	26.8%
	5. Are you confined to your home through ill health?	4	9.8%
	6. Is there anything about your health causing you concern or difficulty?	6	14.6%
	7. Do you have difficulty with vision?	2	4.9%
	8. Do you have difficulty with hearing?	1	2.4%
	9. Have you been in hospital during the past year?	0	0%

The most endorsed block of the VES-13 scale was obtained for age (41.5% over 74 years), as there were 15 patients between the ages of 75 and 84 and 2 patients 85 years old or older. The next most frequent scores occurred in block 4 (“Difficulties in other activities”), for which 15 patients had a positive score (36.6%). These 15 patients reported difficulty kneeling, stooping or bending (n=5), doing heavy housework (n=5), walking 400 meters (n=2) and climbing while carrying a weight of 4.5kg (n=2). There was only one item for which no patients showed a deficit (“Extend and raise your arms”). For the other four items, only one patient showed deficits (“Managing Money”, “Using transportation”, “Doing light housework”, “Writing or other fine hand movements”) (Table 7).

Table 7. Items in VES-13 scale.

	Items	Number of patients with this score	Prevalence
	1. Age
	a. 75–84 years (1 point)	15	36.6%
	b. ≥85 years (3 points)	2	4.9%
	2. Self-reported health
	a. Good or excellent (1 point)	6	14.6%
	b. Fair or poor (1 point)	7	17.1%
	3. ADLs/IADLs. Needs helps in
	a. Bathing (1 point)	3	7.3%
	b. Shopping (1 point)	2	4.9%
	c. Managing money (1 point)	1	2.4%
	d. Transferring (1 point)	1	2.4%
	e. Doing light housework (1 point)	1	2.4%
	4. Activities. Needs help in
	a. Stooping, crouching or kneeling (1 point)	5	12.2%
	b. Doing heavy housework (1 point)	5	12.2%
	c. Reaching or extending arm above shoulder (1 point)	0	0%
	d. Writing or handling small objects (1 point)	1	2.4%
	e. Walking 1/4 mille (1 point)	2	4.9%
	f. Lifting or carrying 10lbs (1 point)	2	4.9%

Twenty-four patients scored a zero on the Barber scale and 17 patients scored ≥1 (risk of frailty). A total of 29.3% of patients (n=12) had a score ≥3 on the VES-13 scale (risk of frailty).

3.5. Relationship between frailty screening questionnaires and the comprehensive geriatric assessment 

In the analysis of the CGA components that are associated with a score indicative of frailty risk on the Barber Questionnaire, patients with dependence in ADLs had a score greater than 0 on the questionnaire more frequently than patients who were independent (66.7% vs. 26.9%, p=0.013). A similar finding was found for patients who were dependent in IADLs when compared with those who were independent with respect to IADLs (68.4% vs. 18.2%, p=0.001). On the contrary, being over 70 years of age compared with being younger than 70 years of age (46.7% vs. 27.3%, p=0.309) or being a high consumer of drugs as opposed to a mild–moderate consumer (33.3% vs. 40%, p=1.000) were not associated with an increased risk of frailty as measured by the Barber Questionnaire. We were unable to analyze the social value (Gijón scale) due to the absence of significant social risk in our sample. Patients with an ECOG of 2 also had an increased risk of frailty when compared to patients with an ECOG of 0–1 (100% vs. 40%, p=0.415) (Table 8).

Table 8. Barber Questionnaire and VES-13 related to the diverse scales of the CGA.

		Barber score>0
		N	Yes	%	p (both-sized)
	Age
	<70 years	11	3	27.3%	0.309
	>70 years	30	14	46.7%
	ECOG
	0–1	40	16	40.0%	0.415
	2	1	1	100%
	Activities of daily living (Barthel)
	Independent	26	7	26.9%	0.013
	Dependent	15	10	66.7%
	Instrumental activities of daily living (Lawton–Brody)
	Independent	22	4	18.2%	0.001
	Dependent	19	13	68.4%
	Cognitive status (Pfeiffer test)
	No deficit	33	12	36.4%	0.241
	Intellectual deficit	8	5	62.5%
	Risk of desnutrition (NSI)
	No risk or low risk	27	9	33.3%	0.142
	Moderate–high risk	14	8	57.1%
	Comorbidity (Charlson index)
	0–3	41	17	41.5%	–
	>3	–	–	–
	Social risk (Gijón)
	No risk (<8)	40	16	40.0%	0.415
	Risk (>8)	1	1	100%
	Number of medications
	No medication or low consumer (0–3)	30	12	40.0%	1.000
	Moderate–severe consumer (>4)	9	3	33.3%

		VES-13 Score ≥3
		N	Yes	%	p
	Age
	<70 years	11	0	0%	0.018
	>70 years	30	12	40%
	ECOG
	0–1	40	11	27.5%	0.293
	2	1	1	100%
	Activities of daily living (Barthel)
	Independent	26	4	15.4%	0.015
	Dependent	15	8	53.3%
	Instrumental activities of daily living (Lawton–Brody)
	Independent	22	1	4.5%	0.000
	Dependent	19	11	57.9%
	Cognitive status (Pfeiffer test)
	No cognitive deficit	33	7	21.2%	0.034
	Cognitive deficit	8	5	62.5%
	Risk of desnutrition (NSI)
	No risk or low risk	27	3	11.1%	0.001
	Risk of desnutrition	14	9	64.3%
	Comorbidity (Charlson index)
	0–3	41	12	29.3%	–
	>3	–	–	–
	Social risk (Gijón scale)
	No risk (<8)	40	11	27.5%	0.293
	Risk (>8)	1	1	100%
	Number of medication
	No medication or low consumer (0–3)	30	7	23.3%	0.238
	Moderate–severe consumer (>4)	9	4	44.4%

We calculated the sensitivity, specificity and predictive values of each of the items in the Barber Questionnaire (dichotomized into risk of frailty vs. no risk) with respect to the various components of the CGA, which were also dichotomized.

The sensitivity of the Barber Questionnaire to predict deterioration or frailty of the ADL was 66.7%, while specificity was 73.1%. The sensitivity and specificity of the questionnaire to predict deterioration in IADLs were 68.4% and 81.8%, respectively.

Furthermore, in the analysis of the CGA components that were associated with a score of ≥3 on the VES-13 scale (score indicative of risk of frailty), it was found that patients with dependence in ADLs reached that score more often than patients who were independent in their ADLs (53.3% vs. 15.4%, p=0.015). A similar result was obtained for patients who were dependent in IADLs when compared with those who were independent (57.9% vs. 4.5%, p<0.001). Further, intellectual impairment (62.5% vs. 21.2%, p=0.034) and risk of malnutrition (64.3% vs. 11.1%, p=0.001) were related to increased vulnerability as assessed by the VES-13 scale. Being older than 70 years, having an ECOG of 2, being of high social risk and having a high degree of drug consumption were important, but were not statistically related to an increased risk of frailty as determined by the VES-13 scale (Table 8).

The sensitivity of the VES-13 scale to predict deterioration in ADLs was 53.6% with a specificity of 84.6% in terms of the IADL; the sensitivity was 57.9% with a specificity of 95.5%. In terms of predicting cognitive impairment, the sensitivity was 62.5% and specificity was 78.8%. Finally, in predicting risk of malnutrition, the sensitivity of the scale was 64.3% and the specificity was 88.9%. The VES-13 scale was also predictive of deterioration in other specific geriatric individual domains of the CGA with the exception of drug use, the Gijón Social scale (PPV very low) and the ECOG (PPV very low).

3.6. Analysis of diagnostic performance or diagnostic accuracy and the predictive power of two screening questionnaires 

In the study of the association between the Barber Questionnaire and the CGA, it was noted that the questionnaire had a sensitivity of 59.1% and a specificity of 78.9% for detecting risk of frailty. The positive predictive value (PPV) of the questionnaire was 76.5%, while the negative predictive value (NPV) was 62.5% and the validity was 68.3%. The likelihood ratio of a positive result with this questionnaire was 2.8.

In the study of the correlation between the questionnaire and the CGA, the Barber Questionnaire indicated that the correlation was fair/good based on the criteria of Fleiss et al., with an ICC for consistency of 0.67 (95% confidence interval [CI]=0.46–0.81; F=5.1; p<0.001) and an absolute agreement ICC 0.53 (95% CI=0.03–0.77). The average number of scales of the CGA that scored as having impairment (2.00) was higher than that for the average Barber (0.7). There was a significant bias between the means of the two ratings (F=35.6, p<0.001).

A total of 59.1% of patients with more than one domain of deterioration in the CGA had a pathological Barber score (Barber>0) compared with 21.1% of patients who had 0–1 domains with deterioration according to the complete CGA (p=0.014).

The predictive capacity of the Barber Questionnaire regarding frailty risk was intermediate (ROC area=0.72; 95% CI=0.56–0.88; standard error=0.080).

For a cut-off point of 0.50, the sensitivity of the Barber Questionnaire to detect frailty was 59.1% and its specificity was 78.9%. By contrast, for a cut-off point of 1.50, the sensitivity was 27.3% with a specificity of 100%.

In the study of the association between the VES-13 scale and the CGA, it was noted that this questionnaire had a sensitivity to detect the risk of fragility of 54.6% and a specificity of 100%. The PPV, NPV and validity were 100%, 65.5% and 75.6%, respectively.

The study of the correlation between the VES-13 scale and the CGA indicated that the correlation between the two was very good, according to the criteria of Fleiss et al., as reflected by an ICC for consistency of 0.81 (95% CI=0.68–0.90; F=9.8; p<0.001) and an absolute agreement ICC of 0.82 (95% CI=0.68–0.90). The average number of scales of the CGA that scored as having impairment (2.0) did not exceed the average level of the VES-13 (1.9). There was no bias between the means of the two ratings (F=0.01; p=0.909).

A total of 54.5% of patients with more than one domain of the CGA indicating impairment had a pathological score on the VES-13 scale (VES-13≥3) compared with patients who scored as having deterioration on 0–1 domains on the CGA but had a pathological score on the VES-13 scale (p<0.001).

The predictive ability of the VES-13 scale with regard to risk of frailty was high (ROC area=0.88; 95% CI=0.76–0.99; standard error=0.058). Overall, the VES-13 scale was highly predictive of the risk of frailty compared with the CGA.

For a cut-off score of 0.50, the sensitivity of the VES-13 scale for the detection of frailty was 86.4% with a specificity of 57.9%. By contrast, for a cut-off score of 1.50, the sensitivity was 77.3% with a specificity of 94.7%; for a cut-off point of 2.50, the sensitivity was 54.5% with a specificity of 100%.

For two 65-year-old patients with breast cancer who will be receiving chemotherapy, one of whom is at risk for frailty, the assessment indicates a 0.71 probability of having a Barber score of greater than zero and a 0.88 probability of having a VES-13≥3. Put differently, a patient selected at random from this sample who was at risk for frailty would have a Barber score greater than zero or score ≥3 on the VES-13 scale 71.9% and 87.6% of the time, respectively, in comparison to a patient who was not fragile.

Fig. 1 shows a comparison of both scales for frailty screening. At first glance, it seems that the VES-13 scale is better than the Barber Questionnaire at detecting frailty in this sample. Additionally, the Hamley–McNeil statistics has a z-value=7.1; thus, the performance of the tests in the detection of frailty is significantly different. Therefore, the VES-13 scale has better performance than the Barber Questionnaire as a screening tool for frailty in this population.

View Large Image Download to PowerPoint

Fig. 1. ROC analysis of VES-13 scale and BQ scores and impairment on the CGA (Se: sensitivity; Sp: specificity).

When we removed the age item from the latter and proceeded to compare the VES-13 with the CGA, we found that the AUC decreased to the value obtained with the full VES-13 scale (AUC 0.79 vs. AUC 0.88, respectively), which was a statistically significant change (Hamley and McNeil statistic, z=5.9). Overall, the VES-13 remained predictive in identifying impairment-risk of frailty in comparison with the CGA, even after excluding the item “age”, with an AUC of 0.79 (95% CI=0.65–0.93; standard error=0.07) (Fig. 2). In this case, the optimal cut-off was >1 (sensitivity=63.6% and specificity=94.7%).

View Large Image Download to PowerPoint

Fig. 2. ROC analysis of the VES-13 scale score, without including the item “age”, and impairment on the CGA (Se: sensitivity; Sp: specificity).

4. Discussion 

The elderly are a heterogeneous group. The spectrum of impairment can range from those who are independent to those who have a moderate risk of health deterioration (vulnerable) to those at risk of functional impairment or death (fragile) [17], [38]. The frail elderly and those at greatest risk of frailty have reduced functionality and physiologic reserve, which lead to an increased likelihood of progressive deterioration, loss of function and adverse health events.

In a recent study that explored the diagnosis of non-cutaneous cancer in the elderly, cancer was associated with higher levels of disability, the coexistence of geriatric syndromes and the presence of vulnerability, as measured by the VES-13 scale, and frailty, as measured by the criteria of Balducci [39]. The goal of identifying frail patients in the field of medical oncology is to prevent chemotherapy from causing functional dependency or institutionalization. According to Balducci, patients with dependence for one or more of the basic activities of daily living (ADL) and with severe comorbidity in one or more of the geriatric syndromes are eligible only for symptomatic management. The underlying reason for this limitation is that the functional reserve of these patients is so limited that they would not be able to tolerate any stress, let alone treatment with cytotoxic drugs [40]. Based on the idea of Lachs et al. [41], the National Comprehensive Cancer Network (NCCN) has proposed a screening test for each of the components of the CGA and then a full assessment of only those domains that received a positive score [8].

There is still no official consensus on the definition of frailty. Linda Fried's group defined a clinical syndrome of frailty [6], in which presentation with 3 of the 5 criteria, listed before, indicates frailty. Several studies have used these criteria to diagnose the states of frailty and pre-frailty [42], [43], [44], [45]. However, a number of problems arise from implementation of some of the criteria [46]. For example, assessment of muscle weakness requires a dynamometer, which is not available in all consultations for medical oncology. In recent studies, researchers have changed the criteria for muscular strength [47], gait [47] and “low level of physical activity” [48]. Recently, other criteria have been added. These factors include cognitive impairment and depressive symptoms. In recent studies, a slow gait, low physical activity, weight loss and cognitive impairment were predictors of institutionalization, death, serious falls and chronic disability; however, there are doubts concerning the value of muscle weakness and self-ratings of fatigue [49]. For these reasons, we chose not to include the Fried criteria for assessment of the risk of frailty in our sample.

Ensrud et al. proposed another definition of frailty, which could be operational for patient consultations. The three criteria of the frailty phenotype were weight loss, inability to rise from a chair 5 times without using one's arms and a low energy level. The presence of these 3 criteria is indicative of weakness; the presence of 1 or 2 of the criteria classifies the patient as pre-frail and those who do not meet any of the criteria are classified as non-fragile. This definition, like that of Fried, is able to predict adverse events; however, it has only been validated in the elderly and is not in widespread use [7]. For this reason, we have not used the Ensrud criteria in our work.

In our cancer population, we used the VES-13 scale, a self-administered test consisting of age-related items and 12 items relating to self-perception of health status, functional ability and physical condition. In a national sample of the elderly, the Medicare Current Beneficiary Survey was used to validate the VES-13 questionnaire, and a score ≥ 3 identified 32% of the patients tested as vulnerable. This group had a fourfold higher risk of death or functional decline over 2 years compared with seniors who scored less than 3 [17]. Much higher scores predicted an increased risk of functional impairment and/or death [34]. The average time to complete this assessment was less than 5min [50]. It is now included in the guidelines for the NCCN. The usefulness of the VES-13 questionnaire compared with the CGA has been studied in patients over 70 years of age with prostate cancer and androgen blockade [51]. It has also been used in oncology to help in the identification of patients for risk stratification and the assessment of toxicity [52], [53].

We agree with Overcash et al. that the presence or absence of vulnerability should not be assessed in all patients older than 70 who will receive treatment [54]. We also agree with the NCCN guidelines that not all elders need to be given the CGA [55]. The NCCN guidelines suggest that both the Fried et al. criteria of frailty [56] and the VES-13 could be used as screening tools for impairment in the elderly [17]. However, there are few data to support a specific recommendation regarding screening tools or suggestions as to what should be the sequence of evaluations. In our study, we chose the Barber and the VES-13 scale and then compared both to identify the tools that would be more useful in a very specific population: women over 65 years of age with non-metastatic breast cancer prior to adjuvant chemotherapy.

It has been argued that the Barber Questionnaire may not have good utility for identifying elderly persons who are at risk. Few studies have examined the validity of this questionnaire, and the results have been inconclusive [23], [57], [58]. In this study, we directly evaluated the relationship between this questionnaire and the CGA.

Previous studies have specifically examined the validity of the Barber Questionnaire and have analyzed the characteristics of the Barber Questionnaire as a diagnostic test [21], [22], [57], [59], [60]. Importantly, no studies have specifically evaluated the cancer population.

In our sample, 41.5% of women scored above 0 on the Barber Questionnaire. It means that 41.5% of patients were at risk for frailty. Various studies analyzing the Barber Questionnaire noted that the risk of frailty was between 55% and 80% [21], [57], [60]. The results of a study conducted in Spain, which serves as a benchmark, were similar to ours (47%) [23]. In the study by Martín-Lesende, 62.9% of patients had a positive Barber Questionnaire [58]. Therefore, the data obtained in our study are consistent with those published thus far, although, as noted, none of these studies evaluated a cancer population.

Furthermore, the risk of frailty as assessed by the VES-13 scale (≥3 points) was 32% in the series by Saliba et al. [17], [18] and 50% in the study by Mohile et al. [51]. In our sample, the risk of frailty was 29.3%. Although this proportion is very similar to that reported by Saliba et al., it is significantly lower than the prevalence in the sample of Mohile et al. It is important to note that Mohile et al. included only patients with prostate cancer who were receiving androgen deprivation therapy and this treatment has been associated with a higher prevalence of disability [61].

Finally, the risk of frailty, as measured by the set of questionnaires in the CGA, was 53.7%, higher than indicated by the two screening tools.

A considerable proportion of patients presented with dependence in ADLs (36.6%) compared with 27.6% in Martín-Lesende's cancer sample [58] or 24% in the Mohile et al. cancer study (this author used the Katz index to measure this parameter) [61]. A considerable proportion exhibited dependence in IADLs (46.3%), similar to the Mohile series (42%) (measured by the Lawton–Brody scale). The similarity between the percentage of patients who were dependent in activities of daily living and the percentage of positive results on the Barber Questionnaire (41.5%) favors the use of such a test as a screening tool. Furthermore, in our sample, a score greater than 0 on the Barber Questionnaire was significantly associated with dependence in ADLs (p=0.013) and in IADLs (p=0.001). Thus, the Barber Questionnaire was able to identify the aspect that most closely relates to frailty. Similarly, a score of ≥3 on the VES-13 scale was significantly associated with dependence in ADLs (p=0.015) and in IADLs (p<0.001). Moreover, this scale was associated with the presence of intellectual impairment (p=0.034) and the risk of malnutrition (p=0.001). These results indicate that the VES-13 scale is associated with more traditional parameters of the CGA and, therefore, may be more accurate than Barber Questionnaire when assessing older adults with breast cancer. Studies that aim to identify other markers of frailty are under development. Most recently, they have been identified a series of markers that detect the risk of fragility/vulnerability in 42% of the elderly. Such studies would have remained obscure if they had exclusively assessed ADL and IADL [62]. The VES-13 scale is another possible alternative to detect frailty as it is related to many scales of the CGA.

In a study by Barber et al., the authors found a sensitivity of 95% and a specificity of 68%, but identifying 80% of the population [21], [22]. In our sample, the sensitivity of the Barber Questionnaire was 59.1% and the specificity was 78.9%, although this study assessed women with breast cancer and evaluated the ability of the questionnaire to be used as a screening tool for frailty. The PPV of the Barber Questionnaire is 76.5%, similar to that described by Larizgoitia and Larizgoitia in their sample, which was 81% in the rural population and 49% in urban areas. Although Elche is an urban area, 68.3% of women in our sample were illiterate and 29.3% had only attended primary school, which is more reflective of a rural sample than an urban one [23]. In Martín-Lesende's sample, the PPV was 27.7%, but this value referred to the ability of the Barber Questionnaire to predict institutionalization, dementia and death and not the risk of frailty compared with the CGA.

The VES-13 scale has a very high PPV (100%), indicating that a positive result on this scale (score ≥3) could predict the presence of the risk of frailty. It is noticed that, although this scale has a good fit with the CGA, this scale has a NPV of 65.5% (or 77.8% at a threshold >2), and some frail patients might be missed by it if VES-13 score is<3.

If we consider the questionnaire items, the Barber Questionnaire shows a defect in its content validity, as several items had few affirmative answers. For example, only one person endorsed item 8 (“It is very difficult conversation because of difficulties with hearing”), only two people endorsed item 7 (“Vision difficulties make common tasks difficult”), and no one responded positively to question 2 (“Not having anyone to turn to if you need help”) or 9 (“Not being admitted to hospital in the last year”). Among the various studies of non-cancer populations, the results are highly variable and differ from those observed in our study. Martín-Lesende found lower scores for items 2 and 4 in their sample [58]. Other authors reported very high scores on item 2 (44.4%) [59], whereas no patient responded affirmatively to that item in our study. This distinction may have been due to the fact that our sample had very low social risk (n=40 patients, i.e., 97.6% of the sample had no social risk, and only one patient had social risk). The questions that make up the Barber Questionnaire are quite specific, making it unlikely that the differences are due to variations in interpretation and, therefore, defects in their reliability. Thus, the discrepancy between the studies likely reflects cultural and social differences. This questionnaire was created and used in the context of the social, health and historical concreteness of the English society of the 1980s, and its current use likely requires cultural adaptation. Additionally, the scale should be adapted to consider transgenerational changes that have occurred since its inception.

In this sample, a cut-off score >0 had a low sensitivity (59.1%) and a specificity of 78.9%, while a cut-off score >1 had an even lower sensitivity (27.3%) and a specificity of 100%. These data confirm that the best cut-off point for this scale is 0, which was used in this study [21], [22], [57].

Given that previous publications have indicated a cut-off point of 3 on the VES-13 scale, our study has also chosen this one. In this sample, a cut-off score >0 had a high sensitivity (86.4%) and specificity of 57.9%, while a cut-off score >1 had a sensitivity of 77.3% and a specificity of 94.7%. We chose this cut-off point because, in large part, this score was due to age. For example, being in the age range of 75–84 years, coupled with a score of +1 but an age ≥85 years, results in a score of +2. Patients over 74 years of age represented a substantial proportion (41.5%) of the sample. Patients with a cut-off between 4 and 7 had similar characteristics with regard to sensitivity and specificity. However, the cut-off point of >2 had the best sensitivity and specificity (77.3% and 94.7%, respectively). Overall, these results highlight the need for more studies to better assess the characteristics of the VES-13 scale using different cut-off points and study populations.

The AUC was significantly greater for the VES-13 scale than for the Barber Questionnaire. No other study on this questionnaire has been reported. The study of Mohile et al., which compares the ability of the VES-13 scale to screen for frailty compared with the CGA, found an AUC of 0.90 with a standard error of 0.05 and a 95% CI of 0.80–0.01. These values are very similar to those obtained in our study, despite being from very different samples [51].

The results of this study should be interpreted with caution due to limitations in the design. First, the definition of a gold standard for detecting deterioration in geriatric domains is arbitrary, since numerous tools are available to identify geriatric deficits. In addition, the CGA model used in this study was designed and used at our center. However, not all studies use the same scales, although the various components of the CGA have been validated in the geriatric population. The choice of a pathological score according to 2 or more scales of our model of CGA as an indicator of frailty is conservative and may underestimate the sensitivity of the screening questionnaires (Barber and VES-13) for detecting the risk of frailty. Evaluation of such a small sample cannot establish causal links among patient characteristics, tumor characteristics and elements of the CGA. If we had increased the period of inclusion of patients in our study, we might have achieved greater statistical power in our analysis. However, the selection of patients for the same kind of cancer and tumoral stage results in greater importance in terms of usefulness of the Barber and the VES-13 scale by oncologists in daily consultation and allows clinicians to avoid bias if patients have been assessed by different professionals.

These results cannot be generalized to all elderly patients with cancer. Several studies have evaluated the CGA and VES-13 scales in patients with cancer receiving chemotherapy [33], [62] or androgen blockade for prostate cancer [51]. However, our study is the first to evaluate them in a population of women with breast cancer.

Based on the results of this study, we propose the use of the VES-13 scale in this population, as well as implementation of a Comprehensive Geriatric Assessment in those patients with a VES-13 score of ≥3.

5. Conclusions 

The usefulness of the Barber Questionnaire to screen the elderly population for risk has been questioned. Few studies have analyzed the validity of frailty screens [21], [22], [57]. Only two have been assessed in the Spanish population [23], [58] and none of them have been validated for populations with cancer.

From the results of our study, the Barber Questionnaire does not seem to be useful in identifying frailty in women over 65 years of age with non-metastatic breast cancer before chemotherapy, given that it has a sensitivity of 59.1% and an index of validity of 68.3%, with weaknesses in the area of content validity.

The Barber Questionnaire was significantly associated with the presence of dependence in ADLs, as measured by the Barthel scale, and dependence in IADLs, as measured by the rate of Lawton–Brody. This association is reasonable, given that many of its items are related to functionality.

The VES-13 scale (AUC 0.88) has more diagnostic accuracy for screening of frailty in the population of women over 65 years of age with breast cancer than the Barber Questionnaire (AUC 0.72) (z>1.9). The scale is also associated with ADL and IADL dependence, as well as risk of malnutrition and cognitive impairment.

The VES-13 scale has many advantages over the Barber Questionnaire in this population:

-The VES-13 is associated with additional scales included in the CGA: functional status (ADL and IADL, as part of the Barber), nutritional risk (NSI scale) and cognitive impairment (Pfeiffer's questionnaire).

-If a patient scores greater than zero on the Barber Questionnaire, it is 2.8 times more likely that the patient is fragile (likelihood ratio 2.8). However, when the VES-13 scale is positive in this population (VES-13≥3), there is a 100% risk of frailty (PPV 100%).

-The VES-13 scale has higher diagnostic accuracy for the screening of frailty in this specific population than Barber Questionnaire.

-The VES-13 scale has high concordance with the CGA, as assessed by the intraclass correlation coefficient (consistency ICC=0.81; absolute agreement ICC=0.82; p<0.001). The Barber Questionnaire has regular/good agreement (consistency ICC=0.67; absolute agreement ICC=0.523; p<0.001). A significant bias was observed between the mean of the scores obtained with the CGA and the score on the Barber Questionnaire (F=35.6, p<0.001).

-The VES-13 scale should be administered to all women over the age of 65 who have been diagnosed with breast cancer and who will receive chemotherapy; for those who score ≥3 on that scale, a full CGA should be administered.

Conflict of interest 

The authors declare that they have no conflicts of interest.

Reviewers 

Miriam B. Rodin, MD, PhD, The University of Chicago, Section of Geriatrics, MC 6098, 5841 S Maryland Ave W-700, Chicago, IL 60637, United States.

Catherine Terret, MD, PhD, Centre Léon Bérard, Department of Medical Oncology, 28, rue Laënnec, F-69373 Lyon cedex 08, France.

Martine Extermann, MD, PhD, H Lee Moffitt Cancer Center USF, 12902 Magnolia Drive, Tampa, FL 33612, United States.