A health economic evaluation of follow-up after breast cancer surgery: Results of an rct study

Abstract

Introduction. Breast cancer follow-up programmes consume large resources and despite the indications that several alternative approaches could be used effectively, there is no coherent discussion about costs and/or cost-effectiveness of follow-up programmes. Patient and methods. In a prospective trial there were 264 breast cancer patients, stage I and II, randomised to two different follow-up programmes- PG (physician group) and NG (nurse group). The trial period was 5 years. The women in the two intervention groups did not differ in anxiety and depression, their satisfaction with care, their experienced accessibility to the medical centre or their medical outcome as measured by recurrence or death. The analyses were done from different lists representing costs at three hospitals in Sweden according to the principles of a cost minimization study. Results: The cost per person year of follow-up differed between the groups, with €630 per person year in PG compared to €495 per person year in NG. Thus, specialist nurse intervention with check-ups on demand was 20% less expensive than routine follow-up visits to the physician. The main difference in cost between the groups was explained by the numbers of visits to the physician in the respective study arms. There were 21% more primary contacts in PG than NG. Discussion. The difference in cost per year and patient by study arm is modest, but transforms to nearly €900 per patient and 5-year period, offering a substantial opportunity for reallocating resources since breast cancer is the most prevalent tumour worldwide.

Due to achievements in early diagnosis and treatment, breast cancer prognosis has improved over the last decades and today is very favourable compared to many other tumour forms. This, together with a high incidence, makes it the most prevalent cancer today, with an estimated 4.4 million women worldwide alive after a breast cancer diagnosis during the last five years [1]. The majority of women treated within the last five years attend a follow-up programme [2], [3]. Thus, in total, these follow-up programmes consume large resources even if the individual consultation entails only limited expense [4], [5]. Since there is no generally accepted strategy for a follow-up programme, these vary in their design from being conducted by cancer specialists or general practioners, from being on a routine basis to be patient initiated or to be based on follow-up conducted by registered nurses [2], [6], [7].

Recent research has indicated that more intensive follow-up does not increase medical safety [8]. Furthermore, there are indications that good quality programmes can be run with trained nurses [9–11]. Despite the high total amount of resources spent on follow-up programmes and the indications that several alternative approaches could be used effectively, there is no coherent discussion about costs and/or cost-effectiveness of follow-up programmes [4], [12].

We studied the costs incurred during the five years of follow-up in both study arms in a randomised clinical trial, comparing nurse-led follow-up on demand after breast cancer surgery with routine follow-up visits to a physician. Since it is shown in the previous study that the two follow-up strategies where equal in terms of anxiety and depression, their satisfaction with care, their experienced accessibility to the medical centre or their medical outcome as measured by recurrence or death [10], our study reduces to a cost minimization study [13]. We investigated which of the two follow-up programmes is the least cost alternative.

Patients and methods

The women in this study were all participants in a longitudinal, randomized, multicenter study performed from 1991 to 2001 at two hospitals in Sweden [10]. Two hundred and sixty-four consecutively selected women with newly diagnosed breast cancer, classified as UICC stage I or stage II [14], were randomized after full informed consent, either to routine medical follow-up by a physician - the physician group (PG, n = 131) – or to on demand visits to a specialist nurse – the nurse group (NG, n = 133). The patients were equally distributed between the two centers. The participants were asked to answer the Hospitality and Depression Scale (HAD) and Satisfaction and Accessibility (SaaC) scales at baseline and twice a year over a period of five years. The number of contacts with the health care services, the number of diagnostic procedures relating to breast cancer, and the time to recurrence or death were monitored. The study, which was approved by the Ethics Committee, Uppsala University, Sweden, is described in detail elsewhere [10].

Interventions

Routine follow-up visits to the physician

A specialist in oncology or surgery examined the patients four times per year during the first two years after surgery, followed by bi-annual examinations for up to five years, and yearly after five years. At the follow-up visits, the patient was interviewed regarding symptoms that could signal a loco-regional relapse or distant metastases, and a clinical examination of the breasts, chest wall and regional lymph nodes was carried out. Mammography was carried out at one-year intervals. Blood tests, chest x-ray or other imaging techniques were only performed on clinical indication.

The specialist nurse intervention with check-ups on demand

In a visit to the physician that took place following radiotherapy and after randomization, patients were given an appointment to meet with an experienced nurse approximately three months after surgery. In the course of this meeting, the patient received information about how to recognise a recurrence in breast, skin, axilla and scar. The nurse arranged mammography at one-year intervals and informed the patient of the result of the mammography by telephone or letter. After three years, the patients were referred back to the routine mammography screening programme. The nurse gave advice on aspects of self-care, such as medication and breast self-examination, and took time to talk to the patient about her psychosocial situation. The patient was instructed to contact the nurse at any time if she had any questions or symptoms that she perceived could be related to breast cancer. The nurse, working in a setting where she had free access to specialists in surgery and/or oncology within her own hospital, co-ordinated the healthcare resources and consulted a physician or a physiotherapist when needed.

Data collection

In monitoring all medical records breast cancer specialist nurses collected data about medical examinations (mammography, pulmonary x-ray, scintigraphy, CT-scans, ultra sound, other imaging, laboratory evaluations, cytolologies, biopsies), visits (Nurse-, Physician-, Social worker-, Physiotherapist- and Breast prosthetic technician visits) and telephone contacts. We also measured the numbers of contacts with other physicians, there were only a few women making visits to other health care institutions.

The unit cost of these events was taken from lists of unit costs from three different hospitals in Sweden: Lund, Varberg and Örebro. Lund and Örebro are University hospitals and Varberg is a central hospital for one county. The unit costs from Lund and Varberg hospitals reflect prices for the year 2006. To reflect year 2006 prices, the year 2005 unit costs from Örebro hospital were adjusted using a 3% discount rate, as recommended by for example Gold et al, [15]. The recourses (and thus costs) considered are those seen in Table I. There is a difference in how the 18 county councils’ in Sweden calculate health care costs. Each county finances the health care and there is no uniformity down to the detail between counties in how costs are estimated. In some counties costing may include rent of space, technical support, consumables etc, in some counties only the salary is included. Thus we used three different counties as information sources of costs, since using only one principle of costing would give a result that would have been difficult to generalize.

Table I.  Unit costs of resources at three different Swedish hospitals for the year 2006. Costs are given in Euros.

Resource	Cost at Lund Hospital	Cost at Varberg Hospital	Cost at Örebro Hospital
Nurse visits	33	118	34
Physician visits	67	294	329
Physiotherapist	44^a	51	37
Social worker	33	95	37
Breast prosthetic technician	35	31	30
Mammography	95	139	140
Pulmonary x-ray	31	46	49
Scintigraphy, CT scans	255	225	201
Ultrasound, other imaging	58	92	90
Laboratory evaluations^b	18	45	24
Telephone contact	27	16	6
Cytolologies, biopsies	176	118	75

Notes: ^aMissing value was imputed with the average unit cost from the other hospitals.

^bLaboratory evaluations, full blood count (haemoglobin, white blood cells, platelet count), liver function test, serum calcium.

For resources that lacked a cost estimate from a specific hospital, we used the average cost of that resource in the other hospitals (Table I).

Economic evaluation-statistical analysis

When alternative follow-up programs are equally efficient in all relevant aspects, a cost-minimization study is sufficient [13]. This means searching for the least cost alternative. Since in our previous study we could not detect any differences between the groups in terms of anxiety, depression, and the patient's notion of accessibility to and satisfaction with health care, recurrence or death [10], the analyses were done according to the principles of a cost minimization study.

We compared cost of follow up between the PG and the NG, by summing costs for all events. The cost per patient of these events was estimated using the unit costs from the three different Swedish hospitals. In order to see if there was a difference between cost of follow-up in PG compared to NG, variation between individuals and variation between costs was accounted for using a bootstrap analysis with 1 000 replicates, posing a probabilistic sensitivity analysis [16], [17]. This is a type of systematic multi-way sensitivity analysis, since it systematically evaluates the effect (on the group difference between PG and NG) of multi-way variation in the input data and assumptions (all resource use and all costs). Each replicate of the two groups, PG and NG, used a list of unit costs that was drawn at random from the three different lists representing costs at Lund, Varberg and Örebro hospitals. In each replicated group, the use of each resource (visits, examinations, etc) was combined with the corresponding cost. The total cost for the group was divided by the total observation time in the group to provide an estimate of the mean cost per year of follow-up.

Confidence intervals were generated using bootstrap intervals.

Also an “Extremes Analysis” was performed to ascertain how sensitive the results were to more extreme values of the most substantial costs [18]. Also the telephone cost was included in this extremes analysis since it could be questioned whether the low cost levels of telephone contacts are generally realistic. Each of the costs for a telephone contact, a visit to a physician and a visit to a nurse were then assigned a higher level than in the original probabilistic sensitivity analysis. The high telephone contact cost was chosen arbitrarily, while the costs for visits to the physician or nurse were taken from Uppsala University Hospital's fees for special referral patients from outside the hospital's catchment area. Patients in breast cancer follow-up programmes are, in the standard situation, patients belonging to the hospital's primary catchment area and therefore the special referral costs may here represent a somewhat extreme level. As an approach for handling missing data, for each resource we analyzed the cost based on all individuals with a non-missing value. The analyses were carried out using the R software [19], [20].

Results

The resource utilization in PG and NG is presented in Table II, along with the corresponding costs. The most prominent difference between the groups in terms of resource use is the number of visits to a physician, i.e. 6.9 during the whole follow-up period in PG versus 3.5 in NG. This carries over to the cost estimate, where we see a difference of €886 in favor of NG. The other differences are all small, and only contribute marginally to the total cost (Table III). Costs per patient and year were €630 in PG and €495 in NG within the first five years of observation. The estimate of confidence intervals is seen in Table III. The difference in cost per patient year and patient by study arm is modest, but transforms to nearly €900 per patient for the entire follow-up period.

Table II.  Estimated mean number of visits and procedures per patient and estimated mean cost per patient for each visit/procedure during the entire follow up in Physician group (PG), Nurse group (NG). Costs are given in Euros.

Resource	Visits and no. procedures, PG	Visits and no. procedures, NG	Costs, PG	Costs, NG
Nurse visits	0.3	0.8	32	94
Physician visits	6.9	3.5	2 035	1 035
Physiotherapist	1.2	1.5	61	77
Social worker	0.2	0.0	23	2
Breast prosthetic technician	0.2	0.2	6	5
Mammography	2.6	2.7	358	379
Pulmonary x-ray	0.3	0.7	16	35
Scintigraphy, CT scans	0.3	0.3	66	59
Ultrasound, other imaging	0.1	0.1	10	10
Full blood count	1.2	1.4	53	61
Liver function test	0.3	0.5	15	23
Serum calcium	0.3	0.4	13	18
Telephone contact	0.7	2.0	11	33
Cytologies	0.4	0.2	45	26
Biopsies*	0.0	0.0	4	3

*Due to figures being rounded, the number of visits for biopsies can be zero even though the costs are not zero.

Table III.  Estimated total cost with 95% confidence intervals (CI) per patient during the whole follow-up, observation time and mean cost per patient per year, in the Physician group (PG) and Nurse group (NG). Costs are given in Euros.

	PG	NG	Difference (PG minus NG)
	Mean (95% CIl)	Mean (95% CI)	Mean (95% CI)
Cost per patient	2 748(2 391 to 4 590)	1 862(1 580 to 2 978)	886(429 to 1 725)
Years in study	4.28(4.01 to 4.57)	3.67(3.41 to 3.97)	0.60(0.18 to 1.01)
Cost per year	630(557 to 1 055)	495(410 to 797)	136(22 to 298)

In the extremes analysis (Table IV) there was a difference between PG cost and NG cost (per patient for the whole follow-up period) for all price levels. For all price levels the lower confidence bound exceeded zero. When the physician visit and nurse visit costs where set to an extreme level the difference between PG and NG cost and the difference between PG and NG cost per year increased. The difference decreased when the telephone contact cost was set to a high level. The difference between PG cost per year and NG cost per year was not significant in the case where only the cost of telephone contact was high.

Table IV.  Results of a sensitivity analysis, showing the difference between the study arms in total costs per patient during the whole follow-up and difference between the study arms in cost per patient per year. Estimates are shown with 95% confidence intervals (CI). Costs are given in Euros.

	PG total cost per patient minus NG total cost per patient	PG cost per year and patient minus NG cost per year and patient
	Mean (95% CI)	Mean (95% CI)
No costs put to high level	886 (429 to 1 725)	136 (22 to 298)
Telephone contact cost put to high level (€106)	762 (289 to 1 604)	100 (−20 to 263)
Physician and nurse visit costs put to high level (€863 and €300, respectively)	2 762 (1 804 to 3 699)	495 (252 to 750)
Telephone contact cost, physician and nurse visit costs put to high level (€106, €863 and €300, respectively)	2 638 (1 659 to 3 621)	459 (203 to 717)

Discussion

We found follow-up by specialist nurse of patients after primary breast cancer treatment to be approximately 20% less expensive per person year than specialist nurse follow-up with contacts on demand. The major determinant of the cost difference was the number of visits to a physician.

The economic analysis is based on a randomized clinical trial with prospective data collection by specialist nurses who scrutinized all medical records. The analyses were done according to the principles of a cost minimization study, since, in our earlier analysis, we found anxiety, depression, and the patient's notion of accessibility to and satisfaction with health care were similar in the study arms [10].

A drawback of the study is the relatively small sample size and the contrast between the study arms is modest in terms of management. Conclusions should be carefully weighted in this perspective. However, this caveat mainly influences power and thus type 2 errors and despite this we find a clear difference in cost. To estimate reliable confidence intervals we have re-sampled the patient data set, using Bootstrap Sampling, which provides confidence intervals reflecting as well variation between individuals as variation between costs.

There is a disadvantage that our data collection could not separate the costs from the different years of follow-up since the first years after primary breast cancer or recurrence are associated with high medical costs, including more frequent visits to a physician for medical interventions [21]. For three resource variables (Physiotherapist, Social worker and Breast prosthesis technician) approximately 39% the patients in both study arms had missing values. It is however unlikely that the main result of the study would be altered even if the missing data from NG would contain more use of these resources than the PG, judging from the amount of use of these resources in the observed (non-missing) part of the data. For all the other resource variables there were very few missing values.

The perspective of the evaluation is mainly that of public health and concerns alternative use of resources, e.g. can physician time be redirected to areas such as new consultations where their competence is absolutely necessary.

The costs varied widely between centres and the explanation could be that health care in Sweden is financed with taxes, which makes it difficult to determine the real costs of each resource. However, it is likely that the relative difference in costs between the two groups is realistic.

The result of the extremes analysis indicates that the study is sensitive for the very high unit cost of a telephone contact. In a setting where telephone costs are much higher than in our setting the difference between the groups would diminish. On the other hand, the difference between the groups increases with increasing physician and nurse visit costs. This is true even if the telephone cost is high. Thus, the difference between the groups disappears only when the telephone cost alone is substantially higher than in our study, which is an unlikely scenario.

Since breast cancer is a common cancer and survival times are long, there are a substantial number of breast cancer survivors in the Western world. For example, in Sweden, with the base population of around 9 million people, the number of women with a breast cancer diagnosis and treated within the last ten years is around 60 000 [22]. Furthermore, modern treatment strategies such as breast conservation and prolonged hormonal treatments demand long follow-up times. Thus, even moderate savings per patient and year of follow-up that can be made without lowering the quality of the follow-up, will be important. If we estimate that 50% of the 60 000 women with a prevalent breast cancer in Sweden are under active follow-up, the savings for Sweden alone would amount to €4.1 million per year applying the findings from this study. Even if the total sum of this saving is a comparatively small part of the cost of breast cancer treatment, in times where opportunities for new diagnostic procedures and treatments are increasing much more rapidly than health care resources, we need to constantly consider alternative uses of resources. Our study points to one such possibility to allocate money from follow-up to other interventions. Physician competence is also a limited resource which can be re-allocated to the area of medical interventions, where such competence is essential, since follow-up can be adequately and effectively managed by nurses [9–11].

Few studies so far have considered costs of follow-up specifically [3], [4], [21], [23], and, in two recent reviews [24], [25], it remains unclear what specific values and to what extent follow-up procedures are included in the total cost. As stated by Sikora and Bosanquet [25] more information is needed on how to use staff time most effectively. Follow-up for breast cancer – and cancer follow-up in general – is an important area for developing new, less costly alternatives, so that scarce resources can be used more effectively [2]. Our study indicates that follow-up after primary cancer treatment is an area where economic evaluation can be worthwhile.

Acknowledgements

The authors thank Sixten Borg at the Swedish Institute for Health Economics, Lund.