Quality of life of patients with chronic lymphocytic leukaemia in the Netherlands: results of a longitudinal multicentre study

Nineteen hospitals in the Netherlands invited patients with CLL for participation in an observational study addressing the management of CLL, costs, and HRQoL [18]. Patients aged 18 years or older diagnosed with CLL could enter the study if he or she did not suffer from another serious malignant disease or previous malignancy, had a complete record, and gave informed consent. Patients who developed a non-CLL-related malignancy were censored at the time of its diagnosis.

Patients who participated in the HRQoL study received a HRQoL questionnaire at the start, halfway through, and at the end of therapy from their treating specialists. Additional questionnaires were sent every 6 months in the periods without treatment to get information about the HRQoL in the period before treatment and between treatments. Since chlorambucil was frequently administered continuously for a long and not predetermined period of time, we choose to send questionnaires during this treatment every 6 months as well, to get more information about the HRQoL during the whole period of treatment.

The instruments employed in the HRQoL assessment were the European Organisation for Research and Treatment of Cancer (EORTC) QLQ-C30 accompanying CLL-specific module [19] and a modified version of the EQ-5D in which five response levels replaced the original three levels [20] as suggested and investigated by Kind and Macran [21].

The EORTC QLQ-C30 has been developed by the EORTC Quality of Life Study Group to assess the QoL of patients with cancer [19]. The core instrument includes 30 questions covering many QoL issues related to cancer patients in general and can be supplemented by a diagnosis-specific module [22].

The questionnaire EORTC QLQ-C30 incorporates five functional scales, three symptom scales, a global quality of life scale (two items), and six single items. The functional scales are physical functioning, role functioning, emotional functioning, cognitive functioning, and social functioning. The symptom scales are fatigue, nausea and vomiting, and pain. Dyspnoea (shortness of breath), sleeping (disturbance), appetite loss, constipation, diarrhoea, and financial difficulties are the six single items. According to the EORTC scoring manual, scores were linearly transformed to a 0–100 scale [23]. A higher score on the functional scales and global quality of life scales meant better functioning and quality of life, whereas a higher score on the symptom scales meant more complaints. Differences in scale scores of 10 points or more were considered clinically meaningful [24].

In this study, the core questionnaire EORTC QLQ-C30 was supplemented by the CLL-specific module [25]. The module is used to describe aspects of CLL that are not included in the core questionnaire and provides information about several domains. There are three multi-item scales, i.e. fatigue, treatment side effects and disease symptoms, infections, and two single item scales on social activities and future health worries. However the module is not yet officially published, the score on the scales cannot be calculated [25], and the average score—ranging from 1 (not at all) to 4 (very much)—on the items can be described.

The EQ-5D measures the general HRQoL and is therefore not influenced by CLL only. At the time of start of the study, a five-level EQ-5D had been developed since the original three-level EQ-5D was not sensitive enough for smaller changes in HRQoL. Since patients with CLL in general experience a high level of HRQoL [15], at least until they reach the advanced stages, it was hypothesised that this expanded five-level classification might provide a more sensitive measure of change in health status than the original three-level EQ-5D (EQ-5D₃).

The modified version of the EQ-5D (EQ-5D₅) [21] comprised the same two items as the EQ-5D₃: a visual analogue scale (VAS) providing a single overall summary score of HRQoL and descriptive classification with five dimensions (mobility, self-care, usual activities, pain/discomfort, anxiety/depression). However, the descriptive classification of the EQ-5D₅ contained five levels, rather than the standard three levels. The two additional levels were unlabelled [21]. It can be seen as the predecessor of the labelled five-level version of the EQ-5D [26], which did not exist at the start of our study yet.

The responses on the descriptive classification can be translated to a utility score, which is a value that reflects an individual’s preference for a certain health outcome with zero reflecting states of health equivalent to death and one reflecting perfect health. Utility values for the EQ-5D₅ states have never been determined, as this instrument has been replaced by a five-level labelled version. We calculated utility values following the suggestion of the creator of the EQ-5D₅ [21]. The known utility values for the levels 1, 2, and 3 of the EQ-5D₃ were used for the levels 1, 3, and 5 of the EQ-5D₅, and the additional two levels were generated assuming the midpoint value between the standard two tariff values using an adaptation of the Dutch three-level tariff [27].

The HRQoL of a CLL patient over time was calculated by connecting all measurements per patient using area under the curve analysis over the entire study duration. To enable the comparison of patients, we presented area under the curve values corrected for the follow-up duration per patient. For each patient, an individual norm score was derived from age- and gender-matched scores of the general population on the EQ-5D [28] and EORTC QLQ-C30 [29]. These two studies, as reported in Refs. [28] and [29], used a panel consisting of more than 2000 Dutch households, representative of the Dutch-speaking non-institutionalised population in the Netherlands.

Patient scores were compared with norm scores using t test or nonparametric test for related samples (significant when p < 0.05). Patient scores of three patient groups (patients without any active treatment, patients treated with chlorambucil only, and other patients) were compared using the Kruskal–Wallis test.

Informed consent for participation was given by 173 CLL patients. Of these, 13 patients (6 %) were excluded from the analysis for the following reasons: eight patients did not meet the inclusion criteria after all; one patient chart was missing; and one patient withdrew himself from the study. Additionally, one hospital dropped out of the study, leaving three patients with incomplete follow-up data.

Of the 160 evaluable patients, 144 patients (90 %) participated in the HRQoL study. Table 1 presents patient characteristics of these 144 patients as a whole and per patient group: patients who did not receive any active treatment during the study period, patients who only received chlorambucil, and patients with other or more treatments. It also presents the characteristics of the patients who did not participate in the HRQoL part of the study.

The mean age at diagnosis of all patients was 62.6 years (SD = 10.5) of whom the majority were male (63 %). On average, male patients were younger at diagnosis (60.8 years, SD = 10.1) than female patients (65.5 year, SD = 10.5). Age at diagnosis did not differ significantly between the patient groups.

From diagnosis until the end of the HRQoL study, 85 patients received active treatment (59 %). Seventy-three patients started treatment before the start of the HRQoL study and 12 patients started their first-line treatment during the study period. Eighty-five per cent of all patients who received active treatment, were treated initially with chlorambucil with or without prednisone. Other initial treatments were chlorambucil–vincristine–prednisone (CVP) (7 %), fludarabine (2 %), fludarabine–cyclophosphamide (FC, 2 %), rituximab plus CVP (R-CVP, 1 %), cyclophosphamide (1 %), and cyclophosphamide–doxorubicin–teniposide–prednisone with bleomycin–vincristine + radiotherapy (1 %). Fifty-three patients also received subsequent line(s) of treatment. Second line treatment was fludarabine monotherapy in most patients (23 patients, 43 %). Other second line treatments were: CVP (17 %), FC (8 %), FCR (8 %), R-CVP (8 %), chlorambucil plus prednisone (6 %), rituximab (4 %), R-CHOP (4 %), CVPP (2 %) and fludarabine–rituximab (FR, 2 %).

Patients were diagnosed for on average 3.9 years at the time of their first questionnaire. Their last questionnaire was on average completed 2.6 years later, at 6.5 years since diagnosis. The mean number of questionnaires was 5.7 per patient, and 127 patients (88.2 %) completed three or more questionnaires. For 25 patients, we did not have information during the complete follow-up duration of the study (see Fig. 1).

Table 2 summarises the results on all instruments used for the total CLL population and for the three patient groups that were described before.

Taking into account the total group of CLL patients, the score on both the EQ-5D and the VAS was lower than the norm score corrected for age and gender [28]. This also applies for the subgroups of patients treated with chlorambucil only or with more/other treatments than chlorambucil. Patients who received no active treatment at all, scored lower on the VAS than the general population, but not on the utility score of the EQ-5D₅.

The patients’ mean score and the mean norm scores per EORTC QLQ-C30 item/scale are also shown in Table 2. It identifies the significant differences of p < 0.05 from the norm score. Statistically significant differences are, however, not always clinically meaningful. Meaningful differences (of more than 10 points [24]) between the norm score and patients’ score were observed for role functioning and fatigue in the total group of CLL patients. This was also applicable to the subgroups of patients treated with chlorambucil only or with more/other treatments than chlorambucil. Other differences were observed for emotional and cognitive functioning, appetite loss, and sleeping in patients who only received chlorambucil, for physical and social functioning, and for dyspnoea in patients who received more or other treatments than chlorambucil. None of the significant differences for patients who did not receive any active treatment were clinically meaningful.

When looking at the total population of CLL patients that reported “a little”, “quite a bit”, or “very much” problems on the EORTC QLQ-CLL16 questionnaire, most patients reported problems on future health concern (62 % of the questionnaires), feeling to have not enough energy (50 %), and having night sweats (48 %). For all patient groups, most problems were reported on future health concern and night sweats. The subgroup of patients who were treated with more or different therapies than chlorambucil also reported many problems with respiratory infections and worries about getting infections. The subgroup with patients who only received chlorambucil had the highest (worst) total mean score over all items.

Figure 2 shows that on almost all single items and scales, the group without any active treatment (watch and wait approach only) scored best of all patient groups. Patients who were treated with chlorambucil only scored worse on HRQoL than patients who were treated with more or different treatments with regard to emotional functioning, cognitive functioning, bruises, uncomfortable stomach, and apathy (data not shown).

Being currently treated or not did influence the HRQoL. The 41 patients who filled in questionnaires during treatment and before/after treatment had a significantly lower utility score during treatment (data not shown). This pattern was also observed in the total study sample as presented elsewhere when the data were analysed per treatment line [30].

In the total population of CLL patients, scores on the VAS and EQ-5D₅ differed significantly between the categories of WHO performance status and the presence/absence of comorbidities (see Supplemental Table 1).

HRQoL results during the watch and wait phase are based on all questionnaires completed before the start of active treatment. This covers not only patients who did not receive any active treatment at all, but also the patients who received a treatment after being in the watch and wait phase. During the watch and wait phase, HRQoL can be compromised due to the illness and its related insecurity as well as by other causes like comorbidities or life events.

The HRQoL during treatment with chlorambucil covers only those questionnaires which were filled in during active treatment with chlorambucil.

Both patients in the watch and wait phase and those during treatment with chlorambucil scored lower on the VAS than the general population corrected for age and gender distribution (Table 3). The difference in utility was only significant for patients treated with chlorambucil (0.81 vs. 0.90).

Supplemental Table 1 shows that scores on the EQ-5D₅ were significantly different between the categories of gender, age at diagnosis, WHO performance status, and the presence/absence of comorbidities for patients during the watch and wait phase. During treatment with chlorambucil, none of the collected patient characteristics influenced the score on the EQ-5D and VAS significantly (e.g. males vs. females).

When comparing the individual patient scores on the EORTC QLQ-C30 with the individual age and gender-adjusted norms scores, the patients’ scores were meaningfully different from the norm score on emotional and role functioning and on sleeping and dyspnoea during treatment with chlorambucil. Differences were also found for physical, cognitive, and social functioning and sleeping scales, but although statistically significant, they were not clinically meaningful. In the watch and wait phase, differences from the norm score for cognitive, role and physical functioning, fatigue, and sleeping were statistically significant, but not clinically meaningful.

With regard to the items of EORTC QLQ-CLL16 module, patients in the watch and wait phase suffered most from worries about their future health (55 % of the questionnaires), night sweats (44 %), and having not enough energy (40 %). Patients during treatment with chlorambucil suffered most from worries about their future health (78 % of the questionnaires), having not enough energy (61 %), and infection risk (56 %).

Patient characteristics of the patients who completed questionnaires during the watch and wait phase were comparable with those of the patients who completed questionnaires during treatment with chlorambucil. Age at diagnosis was 68.7 versus 67.2 years (p = 0.426), WHO performance status was 0 in 83.1 versus 82.9 % of the patients (p = 0.981), and co-morbidity was present in 26.8 versus 37.2 % of the patients (p = 0.195), respectively.

The HRQoL was significantly worse during treatment with chlorambucil than during the watch and wait phase for the following outcomes: utility, VAS, emotional functioning, social functioning, fatigue, dyspnoea, losing weight, changes in temperature, feeling apathetic, lack of energy, respiratory infections, and risk of infections.

Norm scores were available for the EQ-5D [28] and the EORTC QLQ-C30 [29]. The mean difference between the patients’ score and the norm score for that patient was significantly higher during treatment with chlorambucil than during the watch and wait phase for the following scales and items: emotional functioning (p = 0.004), fatigue (p = 0.021), dyspnoea (p = 0.003), VAS (p = 0.002), and utility (p = 0.004).

Since new treatments tend to prolong the overall survival of CLL patients [32], the quality of life during and after treatment becomes more important. Although our study provides insight into the problems that patients with CLL are likely to have, the relatively small number of patients did not allow for comparisons between therapies. This would be very informative for clinicians, but to enable these comparisons in a real-world setting, many patients need to be enrolled, given the high number of available treatments. Due to a low incidence rate of CLL, this would require a long inclusion period or an international approach. Changes in management of CLL over time make it difficult to interpret results of a study with a long inclusion period, and an international study also carries difficulties to the interpretation of the results. Fortunately, the HRQoL results of clinical trials can provide important information on this issue.

A second limitation of our study was that due to the observational character of the study, we were dependent on the health practitioners involved in the study for the timely administration of questionnaires, specifically the questionnaire at the start of a new treatment. Despite our efforts to remind them, they forgot to hand over the questionnaire to the patients before the start of the treatment in the majority of the patients who started a new treatment during our study period. We did not have sufficient information about the HRQoL at the start of treatment to compare the HRQoL before and after treatment.

Another limitation is the uncertainty around the utility scores of the EQ-5D₅ instrument. To decrease this uncertainty, we also showed the mean utility over the study period using two other methods to generate utility values. The first additional method using a predictive model has been developed in multiple myeloma and validated in non-Hodgkin lymphoma patients. The predicted values appeared to follow a similar pattern to the observed EQ-5D values [33]. The second additional method used the “crosswalk” obtained from an international study of the EuroQol group that administered both the three-level and five-level versions of the EQ-5D (see their website: www.​euroqol.​org).

The mean utility score of the midpoint estimation and the two additional methods for the total CLL group—based on only those questionnaires without missing values necessary to derive all three estimations—give the following utility scores: 0.854, 0.847, and 0.844. Since these three methods give quite similar results, we can conclude that our calculation is quite reliable.

Since WHO performance status and the presence of comorbidities influence HRQoL, they are potential confounders in our study. We were not able to correct for these potential confounders due to the heterogeneity in treatment patterns resulting in too small patient groups to apply for example propensity score matching. Patients with measurements during the watch and wait phase and those with measurements during treatment with chlorambucil did, however, not differ statistically in WHO performance status and the presence of comorbidities.

The patient characteristics in our study seem to be reasonably representative for the entire Dutch CLL population since the distribution of gender and the average age at diagnosis agree reasonably well with those of the national registration of CLL and indolent lymphomas (63 vs. 56 % males and 63 vs. 66 years of age) [34]. The slightly lower mean age at diagnosis may be caused by the tendency of haematologists not to bother older patients with the study, or the higher refusal rate to participate by the older patients. The distribution of the disease stages, however, also corresponds with the published distribution in The Netherlands: Binet stage A: 71 versus 60 %, Binet stage B: 16 versus 30 %, and Binet stage C: 11 versus 10 % [35].

In contrast to most RCTs, we also included patients with severe co-morbidity. Co-morbidity (severe heart failure, severe pulmonary disease, severe neurologic disease, severe metabolic disease, inadequate liver function, inadequate renal function, or other co-morbidity) was present in 28 % of the patients. RCTs which aim to study the efficacy of treatments and their influence on HRQoL, often exclude these patients. The outcome of treatments in daily practice could therefore differ from the results found in the RCT. We showed that HRQoL is indeed negatively influenced by having comorbidities and the WHO stage at diagnosis. In our study, the patient group “chlorambucil only” had the highest percentage of patients with co-morbidity. This may explain the relatively worse HRQoL of the patients in this group compared with the patients receiving other treatments.

The percentage of patients with comorbidities was even higher in the group with non-participants. They were also significantly older at diagnosis than participants. This might be related to their choice not to participate in the quality of life study. The percentage of patients willing to participate in the HRQoL study was, however, very high (90 %) so that we do not expect that inclusion of these patients would significantly affect the results.

Since the group of patients with co-morbidity is growing steadily due to an ageing population and an improved overall survival, future research should also focus on the effectiveness of treatments in these patients and the effect of treatments on their HRQoL.