Validation of the Lymphoedema Quality of Life Questionnaire (LYMQOL) in Swedish cancer patients

Abstract

Background: The aim of this study was to validate a translated Swedish version of the lymphoedema-specific quality of life questionnaire (LYMQOL) in a cohort of Swedish cancer patients with secondary lymphoedema of the limbs after cancer treatment.

Material and methods: We recruited 102 patients with lymphoedema of the arms or legs after cancer treatment who were visiting lymphoedema therapists at the departments of oncology at the university hospitals in Linköping and Umeå. The LYMQOL questionnaires were translated forward and backward from English to Swedish. Content and face validity were evaluated. The construct validity was assessed by comparing the LYMQOL with the Short Form Health Survey (SF-36) and the perceived degree of lymphoedema of the limbs, respectively. Reliability was determined through test-retest. The internal consistency was assessed by determining Cronbach’s alpha and by factor analysis.

Results: The content and face validity assessments showed that LYMQOL was an easy, clear and not too long questionnaire to use for patients with lymphoedema. Construct validity was high in both versions when compared with the SF-36. The association between the degrees of perceived lymphoedema and the LYMQOL was only significant in the domains Function and Body Image in the arm version, whereas all domains in the leg version were significant. The reliability was good for the arm version (intra-class-correlation coefficients 0.53–0.87) and very good for the leg version (intra-class-correlation coefficients 0.78–0.90). The internal consistency was acceptable to excellent, with Cronbach’s alpha values between 0.79–0.93 (arm-version) and 0.87–0.94 (leg-version). The factor analysis confirmed the usefulness of the four domains in the LYMQOL versions.

Conclusions: This study confirmed the validity of the Swedish version of LYMQOL and demonstrated that LYMQOL may be a simple and useful tool for use in clinical practice and scientific contexts for evaluating QoL in patients with lymphoedema of the limbs.

Introduction

In the last 25 years, developments in cancer treatment have contributed to an increase in the number of long-term survivors [1]. This highlights the importance of quality of life (QoL) issues, and in particular the impact of adverse effects due to cancer treatment.

Lymphoedema is a troublesome adverse effect that occurs especially after lymphadenectomy in cancer treatment. It is a chronic condition that may progress and worsen over time if not treated and may in rare cases even be fatal. As lymphoedema can have a substantial impact on QoL [2,3] it is important to have simple validated tools to measure its occurrence and changes. Validation is an important issue in the assessment of QoL instruments to ensure that the instrument correctly is measuring what it is intended for. Validation comprises many different analyses. Although international guidelines exist for how to perform validation of questionnaires [4], no generally accepted guidelines have been adopted to ensure the methodological quality for measurement properties of health status questionnaires. Terwee et al. proposed quality criteria for components of a validation of health status questionnaires in order to ensure comprehensive validation [5] encompassing eight measurement properties: content validity, internal consistency, criterion validity, construct validity, reproducibility concerning agreement and reliability, responsiveness, floor and ceiling effects and interpretability. The criteria of each of the components in the validation were rated based on the outcome of the statistical analysis of the component.

Several QoL forms have been described in the literature for measuring lymphoedema-specific QoL [6]. Only a few have been validated under Swedish circumstances, and concerns about their usefulness have been raised as they were considered too non-specific or too generic (the European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire (EORTC-QLQ-C30)) [7] or contained too many questions (the Swedish version of Lymphedema Quality of Life Inventory (LyQLI)) in order to be useful in clinical practice [8].

The condition-specific Lymphoedema Quality of Life Questionnaire (LYMQOL) for the limbs were originally developed and validated by Keeley et al. in 2010 [7]. The intention with the LYMQOL was to develop a simple and useful instrument that could be used in routine clinical practice as an outcome measure and in scientific contexts. The LYMQOL is a self-report questionnaire with two versions; one adapted for oedema of the arms with 28 questions and one for oedema of the legs with 27 questions. The LYMQOL leg version has been translated into Dutch and validated [9] and the arm and leg versions have been translated into Turkish and validated by two research groups [10–13].

The LYMQOL covers various aspects categorised into four domains: function, body image/appearance, symptom and mood. Each domain is scored separately. A higher score indicates lower QoL. No normative values have been proposed. Although the original study by Keeley was comprehensive and the authors concluded that the LYMQOL was valid, none of the validated LYMQOL versions have been quality-criteria assessed.

The aim of this study was to validate a Swedish version of the LYMQOL in patients with lymphoedema of the limbs after cancer treatment.

Material and methods

Study population

Patients with diagnosed lymphoedema of the arms or legs after cancer treatment who visited the lymphoedema therapists (EA, AF and ÅS) at the outpatient rehabilitation clinics in the departments of oncology at the university hospitals in Linköping and Umeå were eligible for the study. After receiving written and oral information, the participants gave written informed consent. On two occasions with an interval of approximately one week, the participants filled in the condition-specific QoL measure for lymphoedema, the LYMQOL, and the Short Form Health Survey (SF-36). The participants completed the forms at home and mailed them in a prepaid envelope to the lymphoedema therapist. After having received the first set of questionnaires, the lymphoedema therapist sent out the second set of questionnaires within a week.

On the first occasion, the participants also completed a six-item questionnaire concerning face and content validity. On the second occasion, the participants also answered questions concerning changes in lymphoedema and QoL compared to the previous occasion.

LYMQOL

The LYMQOL is a condition-specific QoL assessment tool for lymphoedema of the limbs [7]. The LYMQOL consists of two versions; a 28-item questionnaire concerning oedema of the arms and a 27-item version concerning oedema of the legs (Supplementary Table S1). The items cover various aspects categorised into four domains. The domain Function consists of 10 (arm version) or eight (leg version) items, the domain Body image/appearance consists of five (arm version) or seven (leg version) items, the domain Symptom consists of six (arm version) or five (leg version) items, and the domain Mood consists of six items, Each item is scored on a four-grade Likert-like scale as Not at all (= 1); A little (=2); Quite a bit (=3) or A lot (=4). The total score for each domain is made up by adding the individual item scores and dividing the total by the number of items answered. Consequently, a higher score indicates lower quality of life. If more than 50% of the items in a domain were not answered, the total score for that domain was set at nil. Thus, the total score ranged between 0 and 4. The last item in the forms is about ‘overall QoL’ (global score) and was scored on a scale from 0 (poor overall QoL) to 10 (excellent QoL).

The Swedish versions of the LYMQOL (Supplementary Table S2) also include a question about the patient’s perceived severity of the swelling of the limbs, graded as Not at all; A little; Quite a bit or A lot.

Short Form Health Survey (SF-36)

The SF-36 is a 36-item, patient-reported survey of patient health. The form is generic and widely used clinically and in research to evaluate the individual’s health status. The SF-36 assesses eight domains of health and has two summary component for physical (PCS) and mental health (MCS) calculated by summing the weighted subscale scores. The scores range from 0 to 100. A higher score represents a better QoL. The Swedish SF-36 has been validated [14].

Translation of the LYMQOL and development of the Swedish version

Initially the original LYMQOL was translated forward. Nine native Swedish-speaking people, all with academic education (three females and six males) independently translated the form from English to Swedish. The suggestions for translation were gathered and when agreement between the translators concerning concepts and semantics had been reached, a backward translation was carried out by a native English professional proof-reader. Eventually, the final version was tested on a few volunteers in order to evaluate the comprehensibility before the questionnaire was used in the validation study.

Validation of the Swedish LYMQOL

To validate the Swedish version of the LYMQOL we addressed its validity by testing face, content, and construct validity, and the reliability by conducting test-retest and internal consistency analyses.

Face and content validity

The extent to which the LYMQOL questionnaire was subjectively viewed as covering the concept it was intended to measure was evaluated by means of a questionnaire given to the participant on the first occasion of completing the LYMQOL. The questionnaire consisted of six questions and the options of answers were dichotomised or open (Supplementary Table S3).

Criterion and construct validity

Criterion and construct validity were tested by comparing the LYMQOL scores with the SF-36 MCS and PCS scores as ‘gold standards’ and with the patient’s reported perceived degree of swelling of the limbs. The hypotheses were that the MCS correlated with the LYMQOL domains Body image/appearance and Mood and that the PCS correlated with the LYMQOL domains Function and Symptom. To be acceptable the correlations coefficient values were required to be >0.50 (moderate to strong) and at least 75% of the results were in correspondence with the hypotheses.

Test-retest reliability

The participants filled in the LYMQOL questionnaire and the SF-36 on two occasions, with an interval of approximately one week. The association between the two occasions was assessed with intra-class correlation coefficient (ICC), standardised error of measurement (SEM) and the smallest real difference (SRD). An ICC above 0.70 was considered acceptable. The SEM specifies the value of the error between the measurements and the SRD represents the smallest difference between the measurements that is statistically significant.

Internal consistency

Internal consistency is a measure based on the correlations between different items on the same test or the same subscale, and measures whether several items that propose to measure the same general construct produce similar scores. The internal consistency is determined by the Cronbach’s alpha coefficient, where a value between 0.70–0.95 was acceptable. Factor analysis was used to check that the questions were ordered in the same manner as in the original LYMQOL, with the same number of components (Function, Body image/appearance, Symptom and Mood). Although no absolute threshold has been adopted a total variance of the model of at least 60% was considered satisfactory in order to represent a reliable model [15].

Ethical approval

The study was approved by the Regional Ethics Board of Linköping University (Dnr 2012/314-31 and Dnr 2014/471-32).

Statistics

The software IBM SPSS Statistics, Version 25 (IBM Corporation, Armonk, NY, USA) was used for the statistical analyses. Variants of measures of associations were applied. For comparison between the LYMQOL and SF-36 Spearman’s rank correlation was used, and between the two occasions of measurements of the LYMQOL we used intra-class correlation measured as a two-way random effects model. To analyse the association between the questions in the LYMQOL and the domains we used Cronbach’s alpha and factor analysis (principal component analysis) with Varimax rotation fixed to four dimensions. The associations between the patient’s perceived degree of limb lymphoedema and the score for the LYMQOL domains were analysed with a Kruskal–Wallis analysis of variance, and the post hoc tests by Mann–Whitney U-tests. Missing data were handled by multiple imputations using five imputed data sets. The level of significance was set at p < .05 for two-tailed tests.

Results

One hundred and six persons with diagnosed lymphoedema of the arms or legs after cancer treatment were enrolled. Of these, 102 filled in the LYMQOL questionnaires on both occasions and this group constituted the study population. The arms were affected in 52 persons and the legs in 50. Descriptive information about age, gender, duration of lymphoedema, affected limb and degree of lymphoedema is presented in Table 1. The lymphoedema of the arm was unilateral in all persons and the dominant arm was affected by lymphoedema in 54% of cases.

Table 1. Descriptive data of the study population in the validation of the Swedish versions of LYMQOL for arm and leg.

		LYMQOL version
		Arm (n = 52)	Leg (n = 50)
Age (years)		62 (34–75)	61 (34–82)
Gender	Female	51 (98%)	47 (94%)
	Male	1 (2%)	3 (6%)
Duration of lymphoedema (years)		2.0 (0.1–20)	1.5 (0.2–33)
Affected limbs (no. of persons)	Unilateral	52 (100%)	24 (48%)
	Bilateral	0 (0%)	26 (52%)
Degree of lymphoedema* (right side):	None	26 (50%)	9 (18%)
	A little	12 (23%)	20 (40%)
	Quite a bit	13 (25%)	17 (34%)
	A lot	1 (2%)	4 (8%)
Degree of lymphoedema* (left side):	None	27 (52%)	15 (30%)
	A little	11 (21%)	15 (30%)
	Quite a bit	13 (25%)	13 (26%)
	A lot	1 (2%)	7 (14%)

Figures denote median and range or number and per cent.

^* Patient reported severity of the lymphoedema.

Translation process

All nine translators agreed on the final version of the questionnaire. The backward translation from Swedish to English demonstrated nearly complete unanimity with the original questionnaire.

Completeness

The completeness of the LYMQOL and the SF-36 was very high. Of the 102 who completed the LYMQOL on both occasions, 101 completed the SF-36 on both occasions. Missing cells in the LYMQOL arm version comprised 0.55% and 3.23%, respectively, on the two occasions of filling in the questionnaire. The corresponding figures for the leg version were 2.30% on both occasions.

Missing cells in the SF-36 accounted in the arm version for 0.91% and 0.16%, respectively on the two occasions, and for the leg version the figures were, 0.68% on both occasions. Since the results from the imputed datasets were almost identical to the original, only the results from the original dataset are presented.

Face and content validity

The results of the face and content analysis are presented in Table 2. More than 90% of the participants found the questionnaires satisfactory as regards ease of answering; the questions were clear, were of a sufficient number, and none were unnecessary. However, in the leg version 28% found that there were important areas of lymphoedema that might have an impact on QoL that had not been covered in the LYMQOL form. The comments on the open questions are summarised in a synthesis of themes (Supplementary Table S4). The number of themes in the leg version was higher than in the arm version, but none of the items in the themes were reported by more than one or two of the participants.

Table 2. Summary results of the face - and content validity survey with six items of the Swedish arm and leg versions of the LYMQOL.

		LYMQOL version
		Arm (n = 52)		Leg (n = 50)
Item	Question in the survey	Yes	No	Yes	No
A	Was the form easy to answer?	49 (94%)	3 (6%)	50 (100%)	–
B	Too many questions in the form?	–	52 (100%)	2 (4%)	48 (96%)
C	Were the questions clear?	48 (92%)	4 (8%)	43 (86%)	7 (14%)
D	Was there any important areas of lymphoedema that might have an impact on QOL that had not been covered in the LYMQOL form?	7 (13%)	45 (87%)	14 (28%)	36 (72%)
E	Did you think any item in the form was unnecessary?	–	52 (100%)	2 (4%)	48 (96%)
F	Comments to the questionnaire?	8 (15%)	44 (85%)	6 (12%)	44 (88%)

Construct validity

The associations between the domains in the LYMQOL and the SF-36 summary component subscales PCS and MCS are presented in Table 3. The analysis revealed that the LYMQOL domains and the SF-36 subscales PCS and MCS correlated significantly (r_s = −0.29 to −0.82) except for the association between MCS and Body image/appearance in the arm version on the first occasion of measurement.

Table 3. Construct validity of the Swedish version of the LYMQOL.

	LYMQOL version
	Arm								Leg
	SF-36 domain
	PCS				MCS				PCS					MCS
	Timing of filling in the questionnaire
	First occasion		Second occasion		First occasion		Second occasion		First occasion		Second occasion		First occasion			Second occasion
LYMQOL domain	rs	p-value	rs	p-value	rs	p-value	rs	p-value	rs	p-value	rs	p-value	rs		p-value	rs	p-value
Function	−0.34	.01	−0.37	<.01					−0.79	<.01	−0.82	<.01
Body image/appearance					−0.28	.05	−0.26	.02					−0.41		<.01	−0.49	<.01
Symptom	−0.50	<.01	−0.29	.04					−0.70	<.01	−0.75	<.01
Mood					−0.66	<.01	−0.60	<.01					−0.53		<.01	−0.52	<.01

r_s: Spearman’s correlation coefficient (rho). Correlations between LYMQOL domains Function and Symptom and the SF-36 physical component summary (PCS) and the LYMQOL domains Body image/appearance and Mood and the SF-36 mental component summary (MCS), respectively.

The associations between the domains in the arm version of the LYMQOL and the patient-reported perceived swelling of the arms were not consistently statistically significant whereas all domains in the leg version were significantly associated with the degree of the perceived swelling of the legs (Supplementary Table S5). Higher scores indicated a higher degree of swelling. For the arm version, only the domains Function and Body image/appearance were significantly associated with the degree of swelling. However, the distribution of the various groups of swelling was unbalanced for the arm, with only two persons having swelling classified as A lot.

No statistically significant differences were found in the scores of the LYMQOL domains when comparing the individuals with their dominant arm affected by lymphoedema and those with lymphoedema in the non-dominant arm (Supplementary Table S6).

Test-retest reliability

The lymphoedema was reported as stable between the two occasions of assessment in 94% (49/52) of the arm group and in 84% (42/50) of the leg group. Likewise, 83% (46/52) of the arm group and 86% (43/50) of the leg group claimed that the QoL did not differ between the two occasions for reasons other than lymphoedema changes (Supplementary Table S7). The agreement in answering the forms between the two occasions showed that the global agreement was classified as good [16] for the arm form (ICC, range 0.53 to 0.85) and very good for the leg form (ICC, range 0.78 to 0.91) (Table 4).

Table 4. Test-retest reliability.

	LYMQOL version
	Arm						Leg
	Occasion							Occasion
	First	Second						First	Second
Domain	Mean score	Mean score	ICC	p-value	Difference means	SEM	SRD	Mean score	Mean score	ICC	p-value	Difference means	SEM	SRD
Function	1.33	1.31	0.85	<.01	0.02	1.04	2.89	1.67	1.64	0.78	<.01	0.03	2.00	5.54
Body image/appearance	1.52	1.56	0.82	<.01	−0.04	1.69	4.68	1.89	1.84	0.91	<.01	0.05	1.45	4.02
Symptom	1.58	1.70	0.53	<.01	−0.12	3.26	9.04	1.82	1.77	0.88	<.01	0.05	1.64	4.56
Mood	1.42	1.44	0.87	<.01	−0.02	1.43	3.96	1.59	1.55	0.81	<.01	0.04	2.17	6.02
Global score	7.88	7.82	0.65	<.01	0.06	8.09	22.4	7.52	7.48	0.83	<.01	0.04	4.47	12.4

ICC: intra-class correlation coefficient; LYMQOL: Lymphoedema Quality of Life Questionnaire; SEM: standardised error of measurement; SRD: smallest real difference. Intra-class correlations, standardised error of measurement and smallest real difference between LYMQOL domains and occasion of filling in the validation of the Swedish LYMQOL questionnaire versions for arm and leg.

Internal consistency

The analysis of the internal consistency of the LYMQOL for the arm version showed Cronbach’s alpha values between 0.79 to 0.93 (categorised as acceptable to excellent [17]) and for the leg version the values were 0.87 to 0.94 (categorised as good to excellent) (Table 5).

Table 5. Internal consistency in the validation of the Swedish version of the LYMQOL.

	LYMQOL version
	Arm				Leg
	Timing of filling in the questionnaires
	First occasion		Second occasion		First occasion		Second occasion
Domain	Cronbach’s alpha	Number of items	Cronbach’s alpha	Number of items	Cronbach’s alpha	Number of items	Cronbach’s alpha	Number of items
Function	0.79	10	0.86	10	0.91	8	0.92	8
Body image/appearance	0.85	5	0.89	5	0.88	7	0.88	7
Symptom	0.83	6	0.85	6	0.87	5	0.88	5
Mood	0.87	6	0.93	6	0.92	6	0.94	6

The factor analysis revealed that most questions had loadings above 0.4 and the cumulative explained variance was 61.8% and 72.5% for the arm and leg versions, respectively, both with four components (Supplementary Table S8). Thus, the factor structure in the four domains was found to be acceptable.

Floor and ceiling effects and quality criteria

The proportion of participants that had either a floor or ceiling score in the LYMQOL domains was lower than 15% in all domains on both occasions of measurement (Supplementary Table S9).

Both forms of the LYMQOL met the quality criteria for measurement properties in the validation of health status questionnaires. The overview is presented in Supplementary Table S10.

Discussion

We translated the arm and leg versions of the LYMQOL into a Swedish context. The Swedish LYMQOL showed high face validity, good content validity, acceptable to excellent internal consistency, and high construct validity in both the arm and leg versions. The reliability was good for the arm version and very good for the leg version. The factor analysis confirmed the usefulness of the four domains in the LYMQOL.

In accordance with the findings of Keeley et al., our validation confirmed that the participants felt the instrument was easy to use and not too long. The comments on what the participants found was missed out in the questionnaire in our study differed somewhat from those reported by Keeley et al. This probably reflects different social, health care and cultural circumstances between the countries. Although 28% of the participants in our study found that there were important areas of lymphoedema that might have an impact on QoL in the leg version that had not been covered in the LYMQOL form the stated causes were widespread and not possible to summarise to a single item. Keeley et al. argued against increasing the number of items in the questionnaire. Based on practical issues and statistical considerations they found it important to keep the size of the LYMQOL small to make it acceptable for users. None of the other translated versions of the LYMQOL have conducted a face validity evaluation of the questionnaire [7,9–13].

The construct validity was evaluated by comparing the domains Function and Symptom in the LYMQOL with the generic QoL questionnaire SF-36’s PCS and the LYMQOL domains Body image/appearance and Mood with MCS. The leg version showed statistically significant correlations in all domains with PCS and MCS. This is in accordance with the findings of van de Pas et al. [9] and Borman et al. [13]. The arm version showed that Body image/appearance was not consistently significantly correlated with MCS and the correlation coefficient was relatively low. These distinctions in the arm and leg versions probably reflect the impact of the meaning of Body image/appearance for the upper and lower limbs.

Keeley et al. used the EORTC QLQ-C30 in the construct validation and found significant correlation and interclass correlation coefficients in both limb versions. Similar comparisons and results were presented in both Turkish studies concerning the arm version. A concern can be raised about using the EORTC QLQ-C30 as a reference since the LYMQOL is made up of questions derived from the EORTC QLQ-C30 [7]. For obvious reasons such a construction would probably facilitate a positive and significant correlation. However, the Turkish studies [10–13] also used other QoL instruments to analyse the construct validity and found good to very good correlations with the LYMQOL in these.

Another element of establishing construct validity was to analyse the correlation between objective or subjective measures of degree of swelling and the outcome of the LYMQOL. Keeley et al. compared limb volume with LYMQOL scores but found no significant correlations and could not confirm the usefulness of the LYMQOL as a clinical tool [7]. However, the results in studies of the correlation between objective limb volume and QoL are contradictory [18, 19]. We compared the subjectively perceived degree of lymphoedema and the domains of the LYMQOL and confirmed that the LYMQOL, especially the leg version, reflected the degree of subjectively perceived lymphoedema. A higher degree of perceived lymphoedema corresponded to a lower lymphoedema-specific QoL. We believe that this finding is very important as it suggests that the LYMQOL might be a useful tool in a clinical context and in evaluating the fluctuation of lymphoedema over time. In addition, it also indicates that self-reported degree of lymphoedema might be more strongly linked to QoL than limb volume. Consequently the LYMQOL can be seen as a better condition-specific instrument as the severity of the lymphoedema has impact on the QoL. No evaluations have been carried out in the previous publications of validation of the LYMQOL concerning the association between objective or subjective measures of lymphoedema volume and QoL aspects [9–13].

The correlation analyses between the test-retests revealed a high degree of consensus, with correlation coefficients around 0.90 (ICC) in both versions. The internal consistency was acceptable to excellent in the arm version and good to excellent in the leg version. These accord with the results in the other studies of the arm and leg versions [9–13].

A strength of this study is that it addressed both the arm and leg versions of the LYMQOL, equivalent to the original LYMQOL validation. Another important strength is that the participants were enrolled consecutively with around 75% agreeing to participate. Except for the request for proficiency in Swedish no selection of participants was carried out. All participants had clinically confirmed lymphoedema and were being treated by professional lymphoedema therapists in outpatient rehabilitation clinics. Moreover, they had varying degrees of lymphoedema, which enabled us to evaluate the association between the degrees of lymphoedema and the outcomes of the LYMQOL. Thus, we were able to show that a higher degree of lymphoedema was associated with a higher score in the LYMQOL. The low number of participants with perceived severe lymphoedema in the arm group might be seen as a limitation of the study. Other limitations, which might reduce the external generalisation, were the uneven gender distribution. Breast and gynaecological cancer constituted the vast majority of the cancer diagnoses and consequently women constituted more than 94% of the participants. The LYMQOL does not reflect the impact of lymphoedema of the feet, hands, lower trunk and genital area or the neck and head. This might be considered as a limitation and was also recognised by Keeley et al. [7].

Conclusion

This study confirmed the validity of the Swedish LYMQOL forms and met the quality criteria for measurement properties of the components of validation of health status questionnaires. The study indicated that the LYMQOL may be a useful simple tool for measuring lymphoedema and lymphoedema-related QoL aspects in the limbs in clinical practice as well as in scientific contexts.

Acknowledgments

The original LYMQOL questionnaires were used with kind permission from the copyright holder, professor Vaughan Keeley, MD. We thank the professors Göran Berg, MD, Mats Hammar, MD and Lawrence Lundgren, MSci and the associate professors Torsten Johansson, MD, Elizabeth Nedstrand, MD and Lotta Lindh-Åstrand, RN for their valuable work and discussions with the translation of the questionnaires. We are indebted to the women and men, suffering from lymphoedema after cancer treatment, who unselfishly participated in the study.

Disclosure statement

The authors report no conflicts of interest. The granting institutions had no roles in the design of the trial, data collection, analysis, interpretation or writing of the manuscript.

Additional information

Funding

The study was supported financially by grants from the Swedish Cancer Society (CAN 2013/620), Linköping University, the County Council of Östergötland.