Vol.:(0123456789)1 3
Quality of Life Research (2020) 29:1747–1766 https://doi.org/10.1007/s11136-020-02503-8
REVIEW
Communicating treatment risks and benefits to cancer patients: a systematic review of communication methods
L. F. van de Water1,2 · J. J. van Kleef1,2 · W. P. M. Dijksterhuis1 · I. Henselmans2 · H. G. van den Boorn1 · N. M. Vaarzon Morel1 · K. F. Schut1 · J. G. Daams3 · E. M. A. Smets2 · H. W. M. van Laarhoven1
Accepted: 8 April 2020 / Published online: 24 April 2020 © The Author(s) 2020
AbstractPurpose Cancer patients are increasingly involved in decision-making processes. Hence, clinicians need to inform patients about the risks and benefits of different treatment options in order for patients to make well informed decisions. The aim of this review is to determine the effects of methods of communicating prognostic information about (1) disease progression (survival, progression, recurrence and remission), (2) side effects and complications and (3) health-related quality of life (HRQL) on cognitive, affective and behavioral outcomes in cancer patients.Methods A literature search was performed to select articles that were published up to November 2019 and that examined verbal and/or visual risk communication interventions in an oncological clinical setting.Results The search yielded 14,875 studies; 28 studies were ultimately included. For disease progression information, we found that framing affects treatment choice. Furthermore, limiting the amount of progression information in a graphical display could benefit patients’ understanding of risks and benefits. For prognostic information about side effects and compli-cations, precise and defined risk information was better understood than information presented in words. When displaying HRQL data, no consensus was found on which graph type to use.Conclusion Great heterogeneity in the results and methodology and in the compared communication formats precluded us from drawing any further conclusions. Practical implications for clinicians are to consider the effects that different types of framing might have on the patient and to not rely exclusively on words to describe risks, but rather include at least some form of numbers or visualization.
Keywords Risk communication · Health-related quality of life · Side effects · Survival · Cancer
Introduction
In daily clinical practice, many decision making situations occur in which there is no ‘single best treatment option’, since either the medical scientific evidence on the ben-efit–harm ratio of the options is insufficient, or the ratio is dependent on patients’ values [1]. There is increasing con-sensus that in these situations, the patient and clinician need to work together and determine what is best for the patient, a process called shared decision-making (SDM) [2, 3].
The clinician however remains responsible for the trans-fer of probabilistic information to the patient, which is an important step in shared decision-making. If there are mul-tiple treatment options, patients should receive information about disease progression or survival, the risk of side effects and complications, and the impact on health-related qual-ity of life (HRQL) for each option. It is of importance that
This research was presented as a poster at the ARPH conference 2019, and the ISOQOL annual meeting 2018 (awarded with the Student Poster Award 2018).
Electronic supplementary material The online version of this article (https ://doi.org/10.1007/s1113 6-020-02503 -8) contains supplementary material, which is available to authorized users.
* H. W. M. van Laarhoven [email protected]
1 Department of Medical Oncology, Amsterdam University Medical Centers, Cancer Center Amsterdam, University of Amsterdam, P.O. Box 22700, 1100 DE Amsterdam, The Netherlands
2 Department of Medical Psychology, Amsterdam University Medical Centers, Amsterdam Public Health, University of Amsterdam, Amsterdam, The Netherlands
3 Amsterdam University Medical Centers, Medical Library, University of Amsterdam, Amsterdam, The Netherlands
http://orcid.org/0000-0003-3546-9709http://crossmark.crossref.org/dialog/?doi=10.1007/s11136-020-02503-8&domain=pdfhttps://doi.org/10.1007/s11136-020-02503-8
1748 Quality of Life Research (2020) 29:1747–1766
1 3
patients understand the risks and benefits [4], only then can the decision-making process result in the best choice for this particular patient.
However, research shows that correct understanding of relevant outcome information is by no means always achieved in cancer care. For example, studies have shown that breast cancer patients do not fully understand probabil-istic information as presented by the clinician [5, 6]. In the case of rectal cancer, only a few patients were found to be able to correctly estimate probabilistic information on treat-ment outcomes [7]. Clinicians can use different methods to communicate risks, for example, by using words or numbers, by using different types of graphic displays, or by framing the information negatively or positively (in terms of survival or mortality). Based on the review by Zipkin, et al., recom-mendations are available for risk communication in health care, such as avoiding the exclusive use of qualitative risk descriptors (such as ‘many’, or ‘high’), supplementing bar charts or icon arrays to numerical risks, and recognizing that framing and the use of relative risk reductions (RRRs) (see Table 1) influences decision making [8]. However, to date, we do not know to what extent these recommendations apply to the specific context of making decisions about can-cer treatment. These are difficult decisions because of the complexity of most treatments. Moreover, the life-threaten-ing nature of the disease induces many emotions [9], which might affect information processing and thus the decision-making process [10]. Furthermore, we do not know how these recommendations might differ for different types of probabilistic information: disease progression, side effects and complication or HRQL [8].
The aim of this systematic review is to evaluate the evidence on the effect of different ways of communicat-ing treatment-related disease progression, side effects and complications and HRQL risk information to cancer patients who have to decide about treatment. Therefore, the effects of different communication methods will be evaluated on three levels: cognitive outcomes (such as patients’ understand-ing), affective outcomes (such as preference for communi-cation method) and behavioral outcomes (such as treatment choice).
Methods
Search method
A literature search was conducted on the 28th of March 2018 in PubMed, Medline, Embase, PsycINFO and Web of Science. A search update was performed on the 14th of November 2019. Scanning the references of the articles that were initially retained, using citation analysis, helped us to create a set of potentially relevant publications. From this
reference set, key concepts for the systematic search were identified. All references from this search had to be retrieved by the final systematic search. Keywords related to the con-cepts of (surrogate) patients, communication methods (such as graphical, numerical and verbal information), outcomes (such as preference and understanding) and study designs (such as randomized controlled trial and observational stud-ies) were used to search the databases. A broad search strat-egy was applied (see Online Resource 1). Duplicates were removed and articles were screened on eligibility based on title and abstract using Rayyan [11] by six reviewers inde-pendently. Each article was screened in duplicate by two reviewers. The search in March 2018 was screened by three reviewer pairs (JJvK and LvdW, JJvK and IH, NVM and KS) and the search update in November 2019 was screened by LvdW and WD. The same reviewer pairs were used for title and abstract screening and for screening of the full-text articles.
Inclusion and exclusion criteria
All studies published in English up to November 14, 2019 with the following characteristics were included:
Population and context
Studies in which the participants (≥ 18 years) were diag-nosed with cancer were included, as were studies that involved healthy surrogate cancer patients (≥ 18 years), i.e., healthy participants answering as if they had cancer. Only studies focusing on risk communication regarding treatment decisions were included; studies on risk communication in cancer screening programs were excluded.
Study design
Randomized and nonrandomized controlled trials and cross-sectional and longitudinal observational studies were included. Qualitative/mixed method studies were only included if (1) quantitative data could be extracted in relation to an outcome and (2) a risk communication inter-vention was offered. Between- and within-subjects stud-ies were included. Case–control and case-series studies were excluded, as were reports, book chapters, conference abstracts and theses.
Interventions
Any method of communicating probabilistic information was included. If different communication methods were compared, the same data needed to have been presented in the compared formats. When studies used decision aids as part of an intervention, the (manipulated) characteristics
1749Quality of Life Research (2020) 29:1747–1766
1 3
Table 1 Examples of verbal and visual methods of communication
Method of communication Example
Verbal communication Positive framing …65 out of 100 patients are alive at 1 year… Negative framing …35 out of 100 patients die in 1 year time…
Mixed framing …65 out of 100 patients are alive at 1 year, 35 out of 100 patients die in 1 year time … Frequency …80 out of 100 patients… Percentage …80% of patients… Words …there is a high risk of… Absolute risk reduction (ARR) … the risk of death can be lowered by 3%, from 15 to 12% … (ARR = Event rate 1—Event rate 2) Relative risk reduction (RRR) …the risk of death can be lowered by 20%… (RRR = ARR/Event rate 1) Number needed to treat (NNT) …if 33 patients would be treated, 1 would survive because of the treatment… (NNT = 1/ARR) Absolute survival benefit (ASB) …the chance of survival can be increased by 5%, from 76 to 81%…
Visual communication Line graph
Bar chart
Pictograph
1750 Quality of Life Research (2020) 29:1747–1766
1 3
of the risk communication format needed to have been described in a detailed manner.
Outcomes
Cognitive outcomes (e.g., patients’ understanding), affec-tive outcomes (e.g., patients’ preference for the format) and behavioral outcomes (e.g., patients’ treatment choice) were included.
Data extraction
Data regarding study characteristics and outcomes were extracted by LvdW, using an extraction sheet. Outcomes were classified according to Zipkin et al. using the categories of cognitive, affective and behavioral outcomes [8].
Quality assessment
For the randomized, between-subjects trials, methodological quality was assessed with a combination of the criteria for-mulated by the adapted Cochrane Collaboration Consumer and Communication review group [12] and items from the Cochrane Collaboration Tool for Assessing Risk of Bias (Online Resource 2). The scale involves nine items, which were rated as either ‘not fulfilled’ (0), ‘fulfilled’ (1) or ‘not specified’ (0). Studies with positive scores on more than half of the items (> 4.5) were considered as ‘high quality’. Cross-sectional studies using a within-subjects comparison on the outcomes of interest were rated with an adapted version of the Newcastle–Ottawa Scale for Evaluating Cross-sectional/Survey Studies (Online Resource 3). This scale involves nine items, scored with a maximum of two points per item. Stud-ies with scores > 75% of total attainable points were consid-ered ‘high quality’, scores > 50% as ‘moderate quality’ and 50% or less as ‘low quality’ [13]. For qualitative studies, the Critical Appraisal Skills Programme (CASP) Qualitative Checklist [14] was used to assess methodological quality,
which consists of ten items scored with a maximum of 20 points in total [15]. The same cut-off scores as for cross-sectional studies were applied [13]. Quality assessment was performed by LvdW who, in case of uncertainty, discussed with a second reviewer (JJvK) until agreement was reached. To ensure a comprehensive review of the literature, assessed methodological quality was not set as an exclusion criterion.
Data analysis
Studies were subdivided according to the type of type of prognostic information: disease progression, side effects and complications or HRQL. Information on disease pro-gression includes information about survival, progression, remission and recurrence. Verbal and visual communication methods were defined, as shown in Table 1. Analysis was performed separately per type of prognostic information. For each information type, studies investigating the same com-munication methods were compared according to outcomes (cognitive, affective and behavioral).
Results
The search in PubMed, MEDLINE, PsychINFO, Embase and Web of Science yielded 20,102 articles; removing dupli-cates resulted in a total of 14,875 articles for screening. Of these, 181 were screened full text, and 28 were included. A summary of the search results can be found in Fig. 1.
Cognitive outcomes were categorized into interpretation accuracy, subjective understanding (including confusion) and cognitive effort. Affective outcomes were categorized into preference for format, satisfaction with format and per-ceived usefulness (including helpfulness for decision mak-ing). All behavioral outcomes (including endorsement of treatment and treatment preference or intentions) were cat-egorized as treatment choice. We found that this outcome
Table 1 (continued)
Method of communication Example
Pie chart
1751Quality of Life Research (2020) 29:1747–1766
1 3
was only assessed in studies with disease progression or side effects and complications information.
Seventeen studies investigated probabilistic informa-tion regarding disease progression, of which eleven stud-ies included cognitive outcomes (interpretation accuracy, subjective understanding and cognitive effort) [5, 16–25], ten studies included affective outcomes (preference and per-ceived usefulness) [5, 16, 18, 20, 22, 24–28] and seven stud-ies included behavioral outcomes (treatment choice) [17, 21, 23, 24, 29–31]. One study was not included in the behavioral outcome analysis, because it lacked a control group [26]. Of the seventeen studies on disease progression information, fourteen studies investigated information on survival, two studies on recurrence, one study on progression and one study on remission.
Seven studies investigated probabilistic information regarding side effects or complications, of which four studies included cognitive outcomes (interpretation accuracy) [28, 32–34], six studies included affective outcomes (preference and satisfaction) [28, 32–36] and one study included the behavioral outcome treatment choice [37].
Five studies investigated probabilistic information regard-ing HRQL, of which four studies included cognitive out-comes (interpretation accuracy and subjective understand-ing) [38–41] and four studies included affective outcomes (preference and perceived usefulness) [38, 39, 41, 42].
Eleven studies included (cancer) patients [5, 18–22, 26, 28, 36, 38, 42], eight studies included healthy volunteers [16, 17, 23–25, 27, 31, 37] and nine studies included both
cancer patients and either healthy volunteers or health-care professionals or all of these [29, 30, 32–35, 39–41].
Methodological quality
For randomized, between-subjects studies, methodologi-cal quality was rated high for seven studies [16, 23, 25, 32, 34, 35, 37] and low for seven studies [17, 21, 24, 27, 30, 33, 40]. Specific scores can be found in Online Resource 4. Items that were often not sufficiently described encompassed whether the method of randomization was truly random, whether allocation was concealed, whether the data analyst was blinded and whether all prespecified outcomes were reported. Only four out of fourteen studies used validated outcome measures to assess one of the outcomes.
For nonrandomized within-subjects cross-sectional stud-ies, methodological quality was rated high for none of the studies, moderate for nine studies [19, 20, 22, 26, 36, 38, 39, 41, 42] and low for four studies [5, 28, 29, 31]. Statistical testing and nonresponse were often reported in insufficient detail. The representativeness and size of the sample were often not satisfactory. All studies used self-report measures, and only two out of eleven studies used validated measures to assess one of the outcomes.
Methodological quality was moderate for the one qualita-tive study that was included [18].
Because each study assessed and compared another set of communication methods, effects were only described in text if communication methods were compared in more than one study. An overview of the studies’ compared methods and results can be found in Tables 2, 3 and 4.
Probabilistic information on disease progression
Table 2 provides a summary of the results regarding the communication of information about disease progression. Eleven of these studies investigated the effect of different formats on cognitive outcomes. Based on three studies, of which two were of high methodological quality and one of low, graphs were better understood (interpretation accuracy and cognitive effort) when there was less information for the participant to process at one time (in one graph or in differ-ent graphs presented at once), compared to more information [23–25]. One study of low quality supported this conclusion for verbal formats, by showing higher confusion rates when different formats were presented all at once, compared to separate presentation [17]. Contradictory results were found when verbal information was compared to graphical infor-mation; two studies found higher subjective understanding for words compared to graphs (moderate quality) [20] or compared to graphs accompanied by words (low quality) [21]. However, one study of moderate quality found that
Fig. 1 Flow chart of the study inclusion process
1752 Quality of Life Research (2020) 29:1747–1766
1 3
graphs were objectively better understood than words [19]. No further conclusions could be drawn on cognitive out-comes because of heterogeneity in the compared commu-nication formats.
Ten studies included affective outcomes, of which nine investigated a preference for formats and two investigated the perceived usefulness of the formats. Although there was much variation in compared formats, two studies, both of high methodological quality, did compare pictographs (among other formats) with bar charts [16, 25]. However, these two studies found conflicting results: one found bar charts to be preferred over pictographs [16] and the other found pictographs to be preferred over bar charts [25]. Two studies of low and moderate quality found positive framing to be preferred over negative framing, although none per-formed statistical analysis on this effect [5, 18]. When com-paring words to numbers, three studies, of low and moderate quality, found conflicting results, of which none performed statistical analysis on this comparison [5, 26, 28].
Seven studies investigated the behavioral outcome treat-ment choice. Different treatment options were presented in these studies, ranging from very specific (radiotherapy) to more general descriptions of treatment (an unknown treat-ment that is more toxic but also has higher survival rates). In all but one study there were only two possible options [23]. Three studies investigated the differences in treatment choice when information was framed positively (in terms of survival), negatively (in terms of death) or both (mixed frame) [29–31]. Two of these, both of low methodologi-cal quality, found that surgery (higher chance of survival, but risk of perioperative death) was more frequently chosen (instead of radiotherapy—lower chance of survival and more side effects) when information was framed positively instead of negatively [29, 31]. When patients in another low-qual-ity study had to choose between a treatment that was more effective, but more toxic, and one that was less effective and less toxic, the first was less preferred in a negative frame (compared to positive and mixed) [30].
Probabilistic information on side effects and complications
Table 3 provides a summary of the results regarding infor-mation about side effects and complications. Seven stud-ies investigated the effects of communication methods on side effects and complications information [28, 32–37], of which four investigated the cognitive outcome interpretation accuracy, six the affective outcomes preference or satisfac-tion and one the behavioral outcome treatment choice. Two studies of high methodological quality compared accuracies for risk information in words to percentages and/or frequen-cies. These studies found that more precise risk information about side effects (percentage/frequency) was superior to
information in words [32, 34]. When comparing frequen-cies to percentages, no clear effect was found. One study of low quality only found a difference on one of six check questions, showing higher accuracies for percentages [33].
Concerning satisfaction with and preferences for commu-nication methods, most studies did not find a significant dif-ference between formats. However, one study of high quality found that communicating frequencies instead of (solely) verbal risk information can contribute significantly to patient satisfaction [34]. Two other studies of low and moderate quality found a description in both words and numbers and a (detailed) description in words only to be most preferred [28, 36]. These two studies did however not report statistical analysis on these preferences.
Probabilistic information on health‑related quality of life
Table 4 provides a summary of the results regarding HRQL information. Four studies compared different methods of communicating HRQL information on cognitive outcomes three looked at interpretation accuracy and four looked at subjective understanding. One study of moderate meth-odological quality found that basic line graphs were best understood (objectively and subjectively; compared to tex-tual descriptions, line graphs with ranges and several differ-ent bar chart formats) [38]. Another moderate-quality study found simple line graphs to score highest on ease of under-standing, but did not report any statistical testing on this out-come [39]. With respect to the directionality of line graphs, a study of low quality found lines going up meaning better outcomes, to be interpreted more accurately than lines going up meaning more of the outcome or normed lines [40]. A third study of moderate quality, however, found pie charts to be best understood (objectively and subjectively, compared to bar charts and icon arrays), but did not report statistical testing on this outcome for patients separately [41].
Affective outcomes for HRQL data were compared in four studies. Three studies measured perceived usefulness and one study measured preference for communication method. Three studies found that simple line graphs dis-playing mean scores were perceived as most useful [38, 39, 42], of which only one study, of moderate quality, reported a significant difference between formats: line graphs display-ing mean scores were perceived as most helpful (compared to line graphs with ranges, textual descriptions and various bar chart formats) [38]. Another study, however, reported pie charts to be most positively commented on, compared to bar charts and icon arrays [41].
1753Quality of Life Research (2020) 29:1747–1766
1 3
Tabl
e 2
Sum
mar
y of
find
ings
for d
isea
se p
rogr
essi
on
Stud
yIn
form
atio
nO
utco
me
Des
ign
Com
pare
d m
etho
dsFi
ndin
gs o
f int
eres
tM
eth-
odol
ogic
al
qual
ity
Cog
nitiv
e ou
tcom
es C
hao
2003
a
#17
Surv
ival
Inte
rpre
tatio
n ac
cura
cy +
Sub
-je
ctiv
e un
ders
tand
ing
Ran
dom
ized
cro
ss-s
ectio
nal,
betw
een-
-sub
ject
sFi
rst:
RR
R, A
RR
, ASB
, NN
TTh
en: a
ll fo
rmat
s tog
ethe
r (R
RR
+ A
RR
+ A
SB +
NN
T)
Parti
cipa
nts w
ere
sign
ifica
ntly
le
ss a
ccur
ate
if th
ey h
ad
chos
en c
hem
othe
rapy
inste
ad
of su
rger
y an
d w
ere
pres
ente
d w
ith R
RR
in c
ompa
rison
to
the
othe
r thr
ee fo
rmat
s. C
onfu
-si
on ra
tings
wer
e si
gnifi
cant
ly
high
er w
hen
pres
ente
d w
ith a
ll fo
rmat
s tog
ethe
r the
n w
ith o
ne
form
at o
nly.
Low
Dav
ey 2
003
#18
Surv
ival
Inte
rpre
tatio
n ac
cura
cyQ
ualit
ativ
eU
nder
stan
ding
: med
ian
surv
ival
, re
mai
ning
life
span
, abs
olut
e su
rviv
al, r
elat
ive
surv
ival
, co
nditi
onal
surv
ival
, gra
ph
Rem
aini
ng li
fesp
an a
nd a
bsol
ute
surv
ival
wer
e co
rrec
tly u
nder
-sto
od b
y m
ost p
artic
ipan
ts
(res
pect
ivel
y by
18
and
16
out o
f 26)
. Med
ian
surv
ival
w
as c
orre
ctly
und
ersto
od b
y 8
parti
cipa
nts a
nd se
emed
to
hav
e of
ten
been
mis
inte
r-pr
eted
as t
he m
ean.
The
gra
ph
was
cor
rect
ly u
nder
stood
by
6 of
26
and
cond
ition
al a
nd
rela
tive
surv
ival
dat
a w
ere
poor
est u
nder
stood
(2 a
nd 1
pa
rtici
pant
(s),
resp
ectiv
ely)
.
Mod
erat
e
Dav
is 2
010a
,e
#19
Surv
ival
Inte
rpre
tatio
n ac
cura
cyO
bser
vatio
nal c
ross
-sec
tiona
l, w
ithin
-sub
ject
sB
ar c
hart,
pic
togr
aph,
sim
pli-
fied
Kap
lan–
Mei
er su
rviv
al
curv
e, te
xtua
l des
crip
tion
(with
nu
mbe
rs)
Acc
urac
y ra
tes w
ere
sign
ifi-
cant
ly lo
wer
for t
he te
xtua
l de
scrip
tion
than
for t
he b
ar
char
t, pi
ctog
raph
and
Kap
lan–
Mei
er c
urve
.
Low
Hag
erty
200
4b#2
0Su
rviv
alSu
bjec
tive
unde
rsta
ndin
gO
bser
vatio
nal c
ross
-sec
tiona
l, w
ithin
-sub
ject
sW
ords
, per
cent
age,
frac
tion,
pi
ctog
raph
, pie
cha
rt, su
rviv
al
(line
) gra
ph
Parti
cipa
nts f
ound
wor
ds a
nd
num
bers
eas
ier t
o un
ders
tand
th
an v
isua
l pre
sent
atio
ns su
ch
as p
ie c
harts
or g
raph
s.
Mod
erat
e
1754 Quality of Life Research (2020) 29:1747–1766
1 3
Tabl
e 2
(con
tinue
d)
Stud
yIn
form
atio
nO
utco
me
Des
ign
Com
pare
d m
etho
dsFi
ndin
gs o
f int
eres
tM
eth-
odol
ogic
al
qual
ity
Ham
stra
2015
a
#16
Recu
rren
ceIn
terp
reta
tion
accu
racy
+ su
bjec
-tiv
e un
ders
tand
ing
Ran
dom
ized
cro
ss-s
ectio
nal,
betw
een-
subj
ects
Pict
ogra
ph w
ith n
o nu
mbe
rs
labe
led
in th
e gr
aph,
line
gr
aph,
pie
cha
rt, p
icto
grap
h an
d ba
r cha
rt w
ith o
nly
the
num
ber o
f affe
cted
indi
vidu
als
labe
led,
pie
cha
rt, p
icto
grap
h,
bar c
hart
with
bot
h th
e nu
mbe
r of
affe
cted
and
the
num
ber o
f un
affec
ted
indi
vidu
als l
abel
ed
Gra
phs w
ith b
oth
affec
ted
and
unaff
ecte
d in
divi
dual
s lab
eled
, re
sulte
d in
mar
gina
lly h
ighe
r ac
cura
cy sc
ores
than
gra
phs
with
onl
y th
e aff
ecte
d in
di-
vidu
als l
abel
ed (n
on-s
igni
fi-ca
nt).
Ther
e w
as n
o si
gnifi
cant
di
ffere
nce
with
in g
raph
s with
2
num
bers
(pie
cha
rt, p
icto
-gr
aph,
bar
cha
rt) n
or w
ithin
gr
aphs
with
1 n
umbe
r (lin
e gr
aph,
pie
gra
ph, p
icto
grap
h,
bar c
hart)
.Si
gnifi
cant
ly le
ss p
artic
ipan
ts
answ
ered
they
wou
ld b
ette
r un
ders
tand
pro
babi
listic
in
form
atio
n in
a p
icto
grap
h,
com
pare
d to
bar
cha
rt, p
ie
grap
h or
line
gra
ph.
Hig
h
Kan
g 20
18a,
c,c
#21
Surv
ival
Subj
ectiv
e un
ders
tand
ing
Ran
dom
ized
cro
ss-s
ectio
nal,
betw
een-
subj
ects
Wor
ds v
s. w
ords
+ pi
ctog
raph
Parti
cipa
nts i
n th
e w
ords
-onl
y co
nditi
on, e
ndor
sed
a si
gnifi
-ca
ntly
hig
her d
egre
e of
com
-pr
ehen
sion
than
par
ticip
ants
in
the
wor
ds a
nd n
umbe
rs
cond
ition
.
Low
Kie
ly 2
013a
#22
Surv
ival
Subj
ectiv
e un
ders
tand
ing
Obs
erva
tiona
l cro
ss-s
ectio
nal,
with
in-s
ubje
cts
3 ris
k sc
enar
ios (
in w
ords
and
in
a b
ar p
lot)—
best
case
, mos
t lik
ely,
wor
st ca
se—
vs. m
edia
n su
rviv
al (i
n w
ords
)
Sign
ifica
ntly
mor
e pa
rtici
pant
s fo
und
that
the
thre
e sc
enar
ios
wer
e ea
sy to
und
erst
and
and
impr
oved
und
erst
andi
ng o
f su
rviv
al ti
me
com
pare
d to
m
edia
n su
rviv
al.
Mod
erat
e
1755Quality of Life Research (2020) 29:1747–1766
1 3
Tabl
e 2
(con
tinue
d)
Stud
yIn
form
atio
nO
utco
me
Des
ign
Com
pare
d m
etho
dsFi
ndin
gs o
f int
eres
tM
eth-
odol
ogic
al
qual
ity
Lob
b 19
99b
#5Re
curr
ence
Inte
rpre
tatio
n ac
cura
cyO
bser
vatio
nal c
ross
-sec
tiona
l, w
ithin
-sub
ject
sA
bsol
ute
risk
of re
laps
e, R
RR
, m
edia
n 5-
year
surv
ival
, gra
ph-
ical
pre
sent
atio
n, p
erce
ntag
es
vs. f
requ
enci
es, n
umer
ical
vs
. ver
bal d
escr
iptio
ns o
f ris
k, p
ositi
vely
vs.
nega
tivel
y fr
amed
stat
emen
ts
Of p
artic
ipan
ts, 7
3% in
terp
rete
d th
e m
edia
n co
rrec
tly. 8
0%
of p
artic
ipan
ts in
terp
rete
d th
e gr
aphi
cal p
rese
ntat
ion
corr
ectly
. 47%
cou
ld c
alcu
late
th
e R
RR
cor
rect
ly b
ased
on
the
AR
R a
nd 8
6% in
terp
rete
d th
e ab
solu
te ri
sk o
f rel
apse
co
rrec
tly. N
o co
nsist
ency
was
fo
und
in in
terp
reta
tions
of
verb
ally
and
num
eric
ally
com
-m
unic
ated
risk
s.
Low
Zik
mun
d 20
08a
#25
Surv
ival
Inte
rpre
tatio
n ac
cura
cy +
cogn
i-tiv
e eff
ort
Ran
dom
ized
cro
ss-s
ectio
nal,
betw
een-
subj
ects
4-op
tion
pict
ogra
ph, 4
-opt
ion
horiz
onta
l bar
cha
rt, 2
-opt
ion
pict
ogra
ph, 2
-opt
ion
horiz
onta
l ba
r cha
rt
Parti
cipa
nts w
ho v
iew
ed th
e 2-
optio
n pi
ctog
raph
or t
he
2-op
tion
bar c
hart
vers
ion
wer
e si
gnifi
cant
ly m
ore
accu
rate
on
a ris
k re
duct
ion
ques
tion
than
par
ticip
ants
w
ho v
iew
ed th
e 4-
optio
n pi
ctog
raph
or t
he 4
-opt
ion
bar
char
t.Pa
rtici
pant
s ans
wer
ed m
ore
quic
kly
(cog
nitiv
e eff
ort)
whe
n pr
esen
ted
with
a
2-op
tion
grap
h th
an w
hen
pres
ente
d w
ith a
4-o
ptio
n gr
aph.
4-o
ptio
n pi
ctog
raph
s or
4-o
ptio
n ba
r cha
rts d
id n
ot
diffe
r sig
nific
antly
in ti
me
need
ed.
Hig
h
Zik
mun
d 20
10a
#24
Surv
ival
Inte
rpre
tatio
n ac
cura
cy +
cogn
i-tiv
e eff
ort
Ran
dom
ized
cro
ss-s
ectio
nal,
betw
een-
subj
ects
Mul
ti-ou
tcom
e pi
ctog
raph
(sur
-vi
val a
nd m
orta
lity +
caus
e),
surv
ival
-onl
y pi
ctog
raph
(bot
h pi
ctog
raph
s sho
wed
2 tr
eat-
men
t opt
ions
)
Parti
cipa
nts w
ho v
iew
ed
surv
ival
-onl
y pi
ctog
raph
s w
ere
sign
ifica
ntly
mor
e ac
cura
te w
hen
repo
rting
the
tota
l cha
nce
of su
rviv
al w
ith
both
com
bine
d an
d ho
rmon
al
ther
apy,
but
not
whe
n re
port-
ing
the
incr
emen
tal c
hanc
e of
su
rviv
al. C
ogni
tive
effor
t did
no
t diff
er si
gnifi
cant
ly b
etw
een
the
two
cond
ition
s.
Low
1756 Quality of Life Research (2020) 29:1747–1766
1 3
Tabl
e 2
(con
tinue
d)
Stud
yIn
form
atio
nO
utco
me
Des
ign
Com
pare
d m
etho
dsFi
ndin
gs o
f int
eres
tM
eth-
odol
ogic
al
qual
ity
Zik
mun
d 20
11a
#23
Surv
ival
Inte
rpre
tatio
n ac
cura
cyR
ando
miz
ed c
ross
-sec
tiona
l, be
twee
n-su
bjec
ts1 ×
4 pi
ctog
raph
s at o
nce,
2 ×
2 pi
ctog
raph
s seq
uent
ially
Parti
cipa
nts i
n th
e se
quen
tial
cond
ition
wer
e si
gnifi
cant
ly
mor
e ac
cura
te th
an th
ose
in
the
‘all
at o
nce’
con
ditio
n.
Hig
h
Affe
ctiv
e ou
tcom
es C
arey
201
8b#2
8Su
rviv
al +
rem
issi
onPr
efer
ence
Obs
erva
tiona
l cro
ss-s
ectio
nal,
with
in-s
ubje
cts
Wor
ds, n
umbe
rs, b
oth
wor
ds a
nd
num
bers
Rem
issi
on: m
ost p
atie
nts p
re-
ferr
ed b
oth
wor
ds a
nd n
umbe
rs
(42%
) ove
r onl
y w
ords
(30%
) or
onl
y nu
mbe
rs (1
0%).
16%
ha
d no
pre
fere
nce;
05%
gav
e no
info
rmat
ion;
(5 y
ear)
sur-
viva
l: m
ost p
atie
nts p
refe
rred
bo
th w
ords
and
num
bers
(4
3%) o
ver o
nly
wor
ds (2
8%),
or o
nly
num
bers
(8%
). 15
%
had
no p
refe
renc
e; 3
% d
id n
ot
wan
t any
info
rmat
ion.
Low
Dav
ey 2
003b
#18
Surv
ival
Pref
eren
ceQ
ualit
ativ
eFr
amin
g: p
ositi
ve, n
egat
ive,
m
ixed
Pre
sent
atio
n: 1
00 fa
ces
pict
ogra
ph, n
umbe
rs a
nd
perc
enta
ges,
text
onl
y, g
raph
St
age:
one
stag
e al
one,
thre
e st
ages
toge
ther
25 o
ut o
f 26
parti
cipa
nts p
re-
ferr
ed p
ositi
ve fr
amin
g an
d 1
parti
cipa
nt p
refe
rred
mix
ed
fram
ing.
9 o
ut o
f 26
parti
ci-
pant
s pre
ferr
ed in
form
atio
n on
al
l thr
ee st
ages
, 6 p
refe
rred
to
rece
ive
info
rmat
ion
only
for
thei
r sta
ge.
Mod
erat
e
Hag
erty
200
4b#2
0Su
rviv
alPr
efer
ence
Obs
erva
tiona
l cro
ss-s
ectio
nal,
with
in-s
ubje
cts
Wor
ds, p
erce
ntag
e, fr
actio
n,
pict
ogra
ph, p
ie c
hart,
surv
ival
(li
ne) g
raph
Parti
cipa
nts p
refe
rred
wor
ds a
nd
num
bers
ove
r vis
ual p
rese
nta-
tions
such
as p
ie c
harts
or
grap
hs.
Mod
erat
e
Ham
stra
2015
a
#16
Recu
rren
cePr
efer
ence
Ran
dom
ized
cro
ss-s
ectio
nal,
betw
een-
subj
ects
Pict
ogra
ph w
ith n
o nu
mbe
rs
labe
led
in th
e gr
aph,
line
gr
aph,
pie
cha
rt, p
icto
grap
h an
d ba
r cha
rt w
ith o
nly
the
num
ber o
f affe
cted
indi
vidu
als
labe
led,
pie
cha
rt, p
icto
grap
h,
bar c
hart
with
bot
h th
e nu
mbe
r of
affe
cted
and
the
num
ber o
f un
affec
ted
indi
vidu
als l
abel
ed
Sign
ifica
ntly
less
par
ticip
ants
w
ould
pre
fer p
roba
bilis
tic
info
rmat
ion
in a
pic
togr
aph,
ra
ther
than
in a
bar
cha
rt, p
ie
grap
h or
line
gra
ph.
Hig
h
1757Quality of Life Research (2020) 29:1747–1766
1 3
Tabl
e 2
(con
tinue
d)
Stud
yIn
form
atio
nO
utco
me
Des
ign
Com
pare
d m
etho
dsFi
ndin
gs o
f int
eres
tM
eth-
odol
ogic
al
qual
ity
Kie
ly 2
013a
#22
Surv
ival
Pref
eren
ce +
perc
eive
d us
eful
ness
Obs
erva
tiona
l cro
ss-s
ectio
nal,
with
in-s
ubje
cts
3 ris
k sc
enar
ios (
in a
bar
plo
t)—be
st ca
se, m
ost l
ikel
y, w
orst
case
—on
ly m
edia
n su
rviv
al (i
n w
ords
)
48%
of p
artic
ipan
ts p
refe
rred
to
rece
ive
both
thre
e sc
enar
ios
and
med
ian
surv
ival
, 40%
pr
efer
red
thre
e sc
enar
ios o
nly
and
5% p
refe
rred
med
ian
only
. Pa
rtici
pant
s fou
nd th
e th
ree
scen
ario
s sig
nific
antly
mor
e he
lpfu
l tha
n m
edia
n su
rviv
al.
Mod
erat
e
Lob
b 19
99b
#5Re
curr
ence
Pref
eren
ceO
bser
vatio
nal c
ross
-sec
tiona
l, w
ithin
-sub
ject
sA
bsol
ute
risk
of re
laps
e, R
RR
, m
edia
n 5-
year
surv
ival
, gra
ph-
ical
pre
sent
atio
n, p
erce
ntag
es
vs. f
requ
enci
es, n
umer
ical
vs
. ver
bal d
escr
iptio
ns o
f ris
k, p
ositi
vely
vs.
nega
tivel
y fr
amed
stat
emen
ts
43%
of p
artic
ipan
ts p
refe
rred
po
sitiv
ely
fram
ed ri
sks
(eg,
‘cha
nce
of c
ure’
), 33
%
pref
erre
d ne
gativ
ely
fram
ed
mes
sage
s (eg
, ‘ch
ance
of
rela
pse’
) and
25%
had
no
pref
eren
ce. W
hen
com
parin
g fr
eque
ncie
s to
perc
enta
ges,
44%
of r
espo
nden
ts p
refe
rred
a
perc
enta
ge, 1
3% p
refe
rred
a
freq
uenc
y, a
nd th
e re
mai
n-de
r had
no
pref
eren
ce. W
hen
com
parin
g nu
mer
ical
to v
erba
l ris
ks, 5
3% o
f par
ticip
ants
pre
-fe
rred
a n
umer
ical
des
crip
tion
(per
cent
age)
, 38%
pre
ferr
ed
a ve
rbal
des
crip
tion
and
the
rem
aind
er h
ad n
o pr
efer
ence
.
Low
Maz
ur 1
999b
,c
#26
Surv
ival
Pref
eren
ceO
bser
vatio
nal c
ross
-sec
tiona
l, w
ithin
-sub
ject
sW
ords
onl
y, n
umbe
rs o
nly
Of p
artic
ipan
ts h
avin
g a
pref
er-
ence
for e
ither
num
bers
or
wor
ds, 4
4% h
ad a
pre
fere
nce
for w
ords
onl
y an
d 56
% fo
r nu
mbe
rs o
nly.
Mod
erat
e
Stu
dts 2
005a
#27
Surv
ival
Perc
eive
d us
eful
ness
Ran
dom
ized
cro
ss-s
ectio
nal,
betw
een-
subj
ects
RR
R, A
RR
, ASB
, NN
TPa
rtici
pant
s rat
ed th
e A
SB si
g-ni
fican
tly m
ore
ofte
n as
mos
t he
lpfu
l and
mos
t infl
uent
ial
than
the
othe
r thr
ee m
etho
ds
(AR
R, R
RR
and
NN
T).
Low
Zik
mun
d 20
08a
#25
Surv
ival
Pref
eren
ceR
ando
miz
ed c
ross
-sec
tiona
l, be
twee
n-su
bjec
ts4-
optio
n pi
ctog
raph
, 4-o
ptio
n ho
rizon
tal b
ar c
hart,
2-o
ptio
n pi
ctog
raph
, 2-o
ptio
n ho
rizon
tal
bar c
hart
The
4-op
tion
and
the
2-op
tion
pict
ogra
phs r
ecei
ved
sign
ifi-
cant
ly h
ighe
r pre
fere
nce
scor
es
than
the
4-op
tion
bar c
hart.
Hig
h
1758 Quality of Life Research (2020) 29:1747–1766
1 3
Tabl
e 2
(con
tinue
d)
Stud
yIn
form
atio
nO
utco
me
Des
ign
Com
pare
d m
etho
dsFi
ndin
gs o
f int
eres
tM
eth-
odol
ogic
al
qual
ity
Zik
mun
d 20
10a
#24
Surv
ival
Pref
eren
ceR
ando
miz
ed c
ross
-sec
tiona
l, be
twee
n-su
bjec
tsM
ulti-
outc
ome
pict
ogra
ph (s
ur-
viva
l and
mor
talit
y + ca
use)
, su
rviv
al-o
nly
pict
ogra
ph (b
oth
pict
ogra
phs s
how
ed 2
trea
t-m
ent o
ptio
ns)
Surv
ival
-onl
y pi
ctog
raph
s wer
e ra
ted
sign
ifica
ntly
bet
ter t
han
mul
ti-ou
tcom
e pi
ctog
raph
s.
Low
Beh
avio
ral o
utco
mes
Cha
o 20
03a
#17
Surv
ival
Trea
tmen
t cho
ice:
Adju
vant
che
mot
hera
py v
s. N
o ad
juva
nt c
hem
othe
rapy
(in
add
ition
to su
rger
y an
d ta
mox
ifen)
Ran
dom
ized
cro
ss-s
ectio
nal,
betw
een-
subj
ects
Firs
t: R
RR
, AR
R, A
SB, N
NT
Then
: all
form
ats t
oget
her
(RR
R +
AR
R +
ASB
+ N
NT)
Parti
cipa
nts w
ho re
ceiv
ed a
R
RR
wer
e si
gnifi
cant
ly m
ore
likel
y to
end
orse
adj
u-va
nt c
hem
othe
rapy
. Whe
n pa
rtici
pant
s rec
eive
d al
l fou
r m
etho
ds o
f com
mun
icat
ing
surv
ival
ben
efits
of c
hem
othe
r-ap
y, th
ere
wer
e no
sign
ifica
nt
treat
men
t dec
isio
n di
ffere
nces
. D
ecis
ion
confi
denc
e di
d al
so
not d
iffer
sign
ifica
ntly
bet
wee
n co
nditi
ons.
Low
Kan
g 20
18a,
c,d
#21
Prog
ress
ion
Trea
tmen
t cho
ice:
Surg
ical
con
sulta
tion
vs. S
ur-
veill
ance
usi
ng C
T
Ran
dom
ized
cro
ss-s
ectio
nal,
betw
een-
subj
ects
Wor
ds, w
ords
+ pi
ctog
raph
Whe
n nu
mer
ic a
nd g
raph
ical
in
form
atio
n (p
icto
grap
h) w
as
adde
d to
des
crip
tive
info
rma-
tion
(wor
ds) a
bout
a 2
-cm
re
nal t
umor
, par
ticip
ants
fa
vore
d su
rgic
al c
onsu
ltatio
n si
gnifi
cant
ly le
ss o
ften
com
-pa
red
to w
hen
risk
info
rmat
ion
was
pro
vide
d us
ing
wor
ds
only
. Thi
s effe
ct w
as n
ot
appa
rent
whe
n pa
rtici
pant
s w
ere
pres
ente
d w
ith th
e sc
e-na
rio o
f hav
ing
a 5-
cm re
nal
tum
or.
Low
McN
eil 1
982a
#29
Surv
ival
Trea
tmen
t cho
ice:
Sur
gery
vs.
Radi
atio
n th
erap
yR
ando
miz
ed c
ross
-sec
tiona
l, be
twee
n-su
bjec
tsPo
sitiv
e an
d ne
gativ
e fr
amin
gIn
the
posi
tive
fram
ing
cond
ition
(p
roba
bilit
y of
livi
ng),
radi
a-tio
n th
erap
y w
as si
gnifi
cant
ly
less
ofte
n pr
efer
red
to su
rger
y th
an in
the
nega
tive
fram
-in
g co
nditi
on (p
roba
bilit
y of
dy
ing)
.
Low
1759Quality of Life Research (2020) 29:1747–1766
1 3
Tabl
e 2
(con
tinue
d)
Stud
yIn
form
atio
nO
utco
me
Des
ign
Com
pare
d m
etho
dsFi
ndin
gs o
f int
eres
tM
eth-
odol
ogic
al
qual
ity
O’C
onno
r 198
9a#3
0Su
rviv
alTr
eatm
ent c
hoic
e:M
ore
toxi
c tre
atm
ent w
ith h
ighe
r su
rviv
al v
s. Le
ss to
xic
treat
-m
ent w
ith lo
wer s
urvi
val
Ran
dom
ized
cro
ss-s
ectio
nal,
betw
een-
subj
ects
Posi
tive
fram
ing,
neg
ativ
e fr
amin
g, m
ixed
fram
ing
(bot
h po
sitiv
e an
d ne
gativ
e)
Parti
cipa
nts i
n th
e ne
gativ
e fr
amin
g co
nditi
on c
onsi
dere
d th
e m
ore
toxi
c, m
ore
effec
tive
treat
men
t sig
nific
antly
less
de
sira
ble
(com
pare
d to
the
less
eff
ectiv
e, le
ss to
xic
treat
men
t) th
an th
ose
resp
ondi
ng in
the
posi
tive
or m
ixed
fram
es.
Low
Woo
dhea
d 20
11a
#31
Surv
ival
Trea
tmen
t cho
ice:
Surg
ery
vs. R
adia
tion
ther
apy
Obs
erva
tiona
l cro
ss-s
ectio
nal,
with
in-s
ubje
cts
Posi
tive
and
nega
tive
fram
ing
Parti
cipa
nts w
ere
cate
goriz
ed o
n de
cisio
nal s
trate
gy; s
trate
gies
th
at w
ere
mai
nly
data
-driv
en o
r dr
iven
by
pers
onal
exp
erie
nce.
W
hen
parti
cipa
nts u
sed
sce-
nario
dat
a (in
stead
of e
xper
i-en
ce) t
o in
form
thei
r dec
ision
s an
d w
ere
pres
ente
d w
ith a
po
sitiv
e fra
me,
they
wer
e sig
-ni
fican
tly m
ore
likel
y to
cho
ose
surg
ery
over
radi
atio
n th
erap
y co
mpa
red
to w
hen
pres
ente
d w
ith a
neg
ativ
e fra
me.
Low
Zik
mun
d 20
10a
#24
Surv
ival
Trea
tmen
t cho
ice:
Hor
mon
al th
erap
y vs
. Com
bine
d th
erap
y (h
orm
onal
+ ch
emo-
ther
apy)
Ran
dom
ized
cro
ss-s
ectio
nal,
betw
een-
subj
ects
Mul
ti-ou
tcom
e pi
ctog
raph
(sur
-vi
val a
nd m
orta
lity +
caus
e),
surv
ival
-onl
y pi
ctog
raph
Parti
cipa
nts i
n th
e su
rviv
al-o
nly
cond
ition
wer
e si
gnifi
cant
ly
less
like
ly to
say
that
they
pr
efer
red
com
bine
d th
erap
y (h
orm
onal
+ ch
emot
hera
py) t
o ho
rmon
al th
erap
y.
Low
Zik
mun
d 20
11a
#23
Surv
ival
Trea
tmen
t cho
ice:
No
adju
vant
th
erap
y vs
. Hor
mon
al th
erap
y vs
. Che
mot
hera
py v
s. Bo
th
chem
othe
rapy
and
hor
mon
al
ther
apy
Ran
dom
ized
cro
ss-s
ectio
nal,
betw
een-
subj
ects
1 × 4
pict
ogra
phs a
t onc
e, 2
× 2
pict
ogra
phs s
eque
ntia
llyH
ighe
r-num
erac
y pa
rtici
pant
s w
ere
sign
ifica
ntly
less
like
ly to
pr
efer
che
mot
hera
py w
hen
in
sequ
entia
l con
ditio
n vs
. in
the
‘all
at o
nce’
con
ditio
n. L
ower
-nu
mer
acy
parti
cipa
nts d
id n
ot
show
a d
iffer
ence
influ
ence
d by
com
mun
icat
ion
met
hod.
Hig
h
a Fou
nd a
t lea
st on
e si
gnifi
cant
resu
lt be
twee
n fo
rmat
s, on
the
disc
usse
d ou
tcom
eb D
id n
ot re
port
any
stat
istic
al a
naly
sis o
n th
e di
scus
sed
outc
ome
c Als
o co
mm
unic
ate
risk
info
rmat
ion
on c
ompl
icat
ions
on
certa
in tr
eatm
ents
, bes
ides
info
rmat
ion
on su
rviv
al/re
curr
ence
d Com
mun
icat
es ri
sk in
form
atio
n on
can
cer p
rogr
essi
on, w
hich
cou
ld b
e se
en a
s the
opp
osite
of p
rogr
essi
on-f
ree
surv
ival
e Did
not
repo
rt st
atist
ical
testi
ng o
n th
is o
utco
me,
but
test
for q
ualit
y w
as p
erfo
rmed
by
the
revi
ewer
s. Po
st ho
c te
sts re
veal
ed a
sig
nific
ant d
iffer
ence
bet
wee
n th
e ve
rbal
con
ditio
n an
d th
e ot
her
thre
e co
nditi
ons
1760 Quality of Life Research (2020) 29:1747–1766
1 3
Tabl
e 3
Sum
mar
y of
find
ings
for s
ide
effec
ts a
nd c
ompl
icat
ions
Stud
yO
utco
me
Des
ign
Com
pare
d m
etho
dsFi
ndin
gs o
f int
eres
tM
eth-
odol
ogic
al
qual
ity
Cog
nitiv
e ou
tcom
es C
arey
201
8b#2
8In
terp
reta
tion
accu
racy
Obs
erva
tiona
l cro
ss-s
ectio
nal,
with
in-s
ubje
cts
Perc
enta
ge, f
requ
ency
61%
of p
atie
nts i
nter
pret
ed a
per
cent
-ag
e on
the
risk
of si
de e
ffect
s cor
-re
ctly
, 17%
inte
rpre
ted
a pe
rcen
t-ag
e on
the
risk
of c
ompl
icat
ions
co
rrec
tly, a
nd 6
5% in
terp
rete
d a
freq
uenc
y on
the
risk
of si
de e
ffect
s co
rrec
tly.
Low
Kna
pp 2
009
(exp
. 1)a
#32
Inte
rpre
tatio
n ac
cura
cyR
ando
miz
ed c
ross
-sec
tiona
l, be
twee
n-su
bjec
tsVe
rbal
des
crip
tors
, per
cent
age,
fr
eque
ncy
Parti
cipa
nts i
n th
e pe
rcen
tage
and
fr
eque
ncy
cond
ition
s wer
e m
ore
accu
rate
than
thos
e in
the
verb
al
cond
ition
s, th
is p
atte
rn w
as si
g-ni
fican
tly d
iffer
ent f
rom
cha
nce
for
two
of th
ree
accu
racy
que
stion
s.
Hig
h
Kna
pp 2
009I
Ia#3
4In
terp
reta
tion
accu
racy
Ran
dom
ized
cro
ss-s
ectio
nal,
betw
een-
subj
ects
Verb
al d
escr
ipto
rs, (
abso
lute
) fr
eque
ncy,
com
bina
tion
of v
erba
l de
scrip
tors
and
freq
uenc
y ba
nd
Parti
cipa
nts i
n th
e fr
eque
ncy
cond
i-tio
n, d
emon
strat
ed si
gnifi
cant
ly
grea
ter a
ccur
acy
whe
n es
timat
ing
the
likel
ihoo
d of
them
selv
es o
r the
av
erag
e pe
rson
hav
ing
any
side
eff
ect f
rom
taki
ng ta
mox
ifen.
The
y w
ere
also
sign
ifica
ntly
mor
e ac
cu-
rate
in e
stim
atin
g th
e lik
elih
ood
of h
avin
g (tw
o of
four
que
stion
ed)
side
effe
cts t
han
the
othe
r tw
o fo
rmat
s.
Hig
h
Kna
pp 2
013a
#33
Inte
rpre
tatio
n ac
cura
cyR
ando
miz
ed c
ross
-sec
tiona
l, be
twee
n-su
bjec
tsFr
eque
ncy,
per
cent
age,
freq
uenc
y an
d pe
rcen
tage
com
bine
dTh
ere
was
no
sign
ifica
nt d
iffer
ence
in
acc
urac
y be
twee
n th
e th
ree
cond
ition
s, ex
ept f
or o
ne o
f six
ac
cura
cy q
uesti
ons;
par
ticip
ants
in
the
perc
enta
ge c
ondi
tion
wer
e m
ore
accu
rate
whe
n es
timat
ing
pers
onal
ch
ance
of c
atar
acts
then
in th
e fr
eque
ncy
cond
ition
.
Low
Affe
ctiv
e ou
tcom
es C
arey
201
8b#2
8Pr
efer
ence
Obs
erva
tiona
l cro
ss-s
ectio
nal,
with
in-s
ubje
cts
Wor
ds, n
umbe
rs, b
oth
wor
ds a
nd
num
bers
Mos
t pat
ient
s pre
ferr
ed b
oth
wor
ds
and
num
bers
(38%
) ove
r onl
y w
ords
(28%
), or
onl
y nu
mbe
rs
(16%
). 16
% h
ad n
o pr
efer
ence
and
0.
5% d
id n
ot w
ant a
ny in
form
atio
n.
Low
Kna
pp 2
009
(exp
. 1)
#32
Satis
fact
ion
Ran
dom
ized
cro
ss-s
ectio
nal,
betw
een-
subj
ects
Verb
al d
escr
ipto
rs, p
erce
ntag
e,
freq
uenc
yTh
ere
wer
e no
sign
ifica
nt d
iffer
ence
s on
satis
fact
ion
betw
een
the
thre
e co
nditi
ons.
Hig
h
1761Quality of Life Research (2020) 29:1747–1766
1 3
a Fou
nd a
t lea
st on
e si
gnifi
cant
resu
lt be
twee
n fo
rmat
s, on
the
disc
usse
d ou
tcom
eb D
id n
ot re
port
any
stat
istic
al a
naly
sis o
n th
e di
scus
sed
outc
ome
Tabl
e 3
(con
tinue
d)
Stud
yO
utco
me
Des
ign
Com
pare
d m
etho
dsFi
ndin
gs o
f int
eres
tM
eth-
odol
ogic
al
qual
ity
Kna
pp 2
009I
Ia#3
4Sa
tisfa
ctio
nR
ando
miz
ed c
ross
-sec
tiona
l, be
twee
n-su
bjec
tsVe
rbal
des
crip
tors
, (ab
solu
te)
freq
uenc
y, c
ombi
natio
n of
ver
bal
desc
ripto
rs a
nd fr
eque
ncy
band
Parti
cipa
nts i
n th
e fr
eque
ncy
cond
i-tio
n an
d in
the
com
bine
d co
nditi
on
wer
e si
gnifi
cant
ly m
ore
satis
fied
with
the
info
rmat
ion
they
rece
ived
th
an th
ose
in th
e ve
rbal
con
ditio
n.
Hig
h
Kna
pp 2
016
#35
Satis
fact
ion
Ran
dom
ized
cro
ss-s
ectio
nal,
betw
een-
subj
ects
(2 ×
2)- R
isk
expr
essi
on: n
umer
ical
onl
y,
com
bine
d ve
rbal
and
num
eric
al- R
isk
qual
ifier
: "w
ill a
ffect
…",
"may
aff
ect…
"
Ther
e w
ere
no si
gnifi
cant
diff
eren
ces
conc
erni
ng sa
tisfa
ctio
n be
twee
n ne
ither
the
risk
expr
essi
on fo
rmat
s no
r the
risk
qua
lifier
s.
Hig
h
Kna
pp 2
013
#33
Pref
eren
ceR
ando
miz
ed c
ross
-sec
tiona
l, be
twee
n-su
bjec
tsFr
eque
ncy,
per
cent
age,
freq
uenc
y an
d pe
rcen
tage
com
bine
d53
% o
f the
par
ticip
ants
pre
ferr
ed th
e co
mbi
ned
form
at. H
owev
er th
ere
wer
e no
sign
ifica
nt d
iffer
ence
s be
twee
n co
nditi
ons.
Low
Zom
orod
bakh
sch
2018
b
#36
Pref
eren
ceO
bser
vatio
nal c
ross
-sec
tiona
l, w
ithin
-sub
ject
sW
ords
, bot
h w
ords
and
num
bers
Mos
t pat
ient
s pre
ferr
ed d
etai
led
info
rmat
ion
in w
ords
(42%
) or
deta
iled
info
rmat
ion
in w
ords
with
ad
ded
num
bers
(32%
) ove
r con
cise
, ge
nera
l inf
orm
atio
n in
wor
ds (5
%)
or a
refe
renc
e to
a b
ookl
et (1
6%).
Mod
erat
e
Beh
avio
ural
out
com
es G
uric
h 20
18a
#37
Trea
tmen
t cho
ice:
Lim
b am
puta
tion
vs. l
imb
salv
age
Ran
dom
ized
cro
ss-s
ectio
nal,
betw
een-
subj
ects
Posi
tive
and
nega
tive
fram
ing
Whe
n lim
b sa
lvag
e w
as fr
amed
neg
a-tiv
ely
(lim
b fu
nctio
ning
low
er th
an
gene
ral p
opul
atio
n), s
igni
fican
tly
mor
e pa
tient
s cho
se a
mpu
tatio
n th
an w
hen
limb
salv
age
was
fram
ed
posi
tivel
y (li
mb
func
tioni
ng h
ighe
r th
an w
ith a
mpu
tatio
n).
Hig
h
1762 Quality of Life Research (2020) 29:1747–1766
1 3
Tabl
e 4
Sum
mar
y of
find
ings
for q
ualit
y of
life
Stud
yO
utco
me
Des
ign
Com
pare
d m
etho
dsFi
ndin
gs o
f int
eres
tM
eth-
odol
ogic
al
qual
ity
Cog
nitiv
e ou
tcom
es B
rund
age
2005
a
#38
Inte
rpre
tatio
n ac
cura
cy +
subj
ectiv
e un
ders
tand
ing
Obs
erva
tiona
l cro
ss-s
ectio
nal,
with
in-
subj
ects
Line
gra
ph, l
ine
grap
h w
ith ra
nges
, te
xtua
l des
crip
tion,
side
-by-
side
ch
ange
(res
pons
e) b
ar c
hart,
stac
ked
chan
ge (r
espo
nse)
bar
cha
rt, st
acke
d ra
w d
ata
in b
ar c
hart
Bot
h ac
cura
cy sc
ores
and
eas
e-of
-un
ders
tand
ing
ratin
gs w
ere
high
est
on li
ne g
raph
s sho
win
g on
ly th
e m
ean
scor
es (n
o ra
nges
). Te
xtua
l de
scrip
tion
show
ed h
ighe
r acc
ura-
cies
than
the
bar c
hart
form
ats.
Bot
h re
sults
reac
hed
sign
ifica
nce.
Mod
erat
e
Bru
ndag
e 20
15b
#39
Subj
ectiv
e un
ders
tand
ing
Ran
dom
ized
cro
ss-s
ectio
nal,
mix
ed
met
hods
,, be
twee
n-su
bjec
tsSi
mpl
e lin
e gr
aph
of m
ean
scor
es o
ver
time,
line
gra
ph w
ith n
orm
s, lin
e gr
aph
with
con
fiden
ce in
terv
als,
bar
char
t of a
vera
ge c
hang
es, b
ar c
hart
base
d on
a re
spon
der d
efini
tion
(impr
oved
, sta
ble,
wor
sene
d), c
umu-
lativ
e di
strib
utio
n fu
nctio
n
With
in th
e gr
aphs
dis
play
ing
grou
p-le
vel d
ata,
ratin
gs o
n ea
se-o
f-un
der-
stan
ding
wer
e hi
ghes
t for
sim
ple
line
grap
hs o
f mea
n sc
ores
ove
r tim
e.
Mod
erat
e
Tol
bert,
201
8a#4
0In
terp
reta
tion
accu
racy
+ su
bjec
tive
unde
rsta
ndin
gR
ando
miz
ed c
ross
-sec
tiona
l, be
twee
n-su
bjec
tsLi
ne g
raph
s with
line
s goi
ng u
p m
eani
ng (1
) "be
tter"
out
com
es, (
2)
"mor
e" o
f the
out
com
e or
(3) l
ines
th
at w
ere
norm
ed to
a g
ener
al p
opu-
latio
n av
erag
e
Patie
nts i
nter
pret
ed li
ne g
raph
s with
lin
es g
oing
up
mea
ning
"bet
ter"
out
-co
mes
sign
ifica
ntly
mor
e ac
cura
tely
th
an w
ith li
nes g
oing
up
mea
ning
"m
ore"
of t
he o
utco
me
or "n
orm
ed"
lines
. Gra
phs w
ith li
nes g
oing
up
mea
ning
"bet
ter"
wer
e m
ost o
ften
rate
d as
’ver
y cl
ear’
or ’s
omew
hat
clea
r’ (n
on-s
igni
fican
t).
Low
Tol
bert,
201
9b#4
1In
terp
reta
tion
accu
racy
+ su
bjec
tive
unde
rsta
ndin
gO
bser
vatio
nal c
ross
-sec
tiona
l, m
ixed
m
etho
ds, w
ithin
-sub
ject
sPi
e ch
art,
bar c
hart,
icon
arr
ayPa
tient
s’ in
terp
reta
tion
accu
racy
w
as h
ighe
st fo
r pie
cha
rts a
nd ic
on
arra
ys, c
ompa
red
to b
ar c
harts
. Pie
ch
arts
wer
e m
ost o
ften
rate
d as
‘v
ery
clea
r’ or
‘som
ewha
t cle
ar’
(no
stat
istic
al a
naly
sis a
vaila
ble
for
patie
nts o
nly)
.
Mod
erat
e
Affe
ctiv
e ou
tcom
es B
rund
age
2003
#42
Perc
eive
d us
eful
ness
Obs
erva
tiona
l cro
ss-s
ectio
nal,
with
in-
subj
ects
Gra
phic
al d
ispl
ays o
f: no
nnum
eric
al
trend
s, m
ean
scor
es, m
ean
scor
es
with
SD
’s, m
ean
scor
es d
escr
ibed
ve
rbal
ly, c
hang
e in
mea
n sc
ores
af
ter s
ix m
onth
s, in
divi
dual
cha
nge
scor
es o
f tw
o, th
ree
or fi
ve d
ivis
ions
, no
rmal
ized
/raw
scor
es o
f 20t
h an
d 80
th p
erce
ntile
(not
cle
ar w
hat
kind
of g
raph
s the
num
bers
wer
e di
spla
yed
in)
Parti
cipa
nts f
ound
gra
phic
al d
ispl
ays
of m
ean
scor
es th
e m
ost u
sefu
l, bu
t no
sign
ifica
nt d
iffer
ence
s wer
e fo
und.
Mod
erat
e
1763Quality of Life Research (2020) 29:1747–1766
1 3
a Fou
nd a
t lea
st on
e si
gnifi
cant
resu
lt be
twee
n fo
rmat
s, on
the
disc
usse
d ou
tcom
eb D
id n
ot re
port
any
stat
istic
al a
naly
sis o
n th
e di
scus
sed
outc
ome
Tabl
e 4
(con
tinue
d)
Stud
yO
utco
me
Des
ign
Com
pare
d m
etho
dsFi
ndin
gs o
f int
eres
tM
eth-
odol
ogic
al
qual
ity
Bru
ndag
e 20
05a
#38
Perc
eive
d us
eful
ness
Obs
erva
tiona
l cro
ss-s
ectio
nal,
with
in-
subj
ects
Line
gra
ph, l
ine
grap
h w
ith ra
nges
, te
xtua
l des
crip
tion,
side
-by-
side
ch
ange
(res
pons
e) b
ar c
hart,
stac
ked
chan
ge (r
espo
nse)
bar
cha
rt, st
acke
d ra
w d
ata
in b
ar c
hart
Line
gra
phs d
ispl
ayin
g m
ean
scor
es
(no
rang
es) w
ere
rate
d hi
ghes
t on
help
fuln
ess a
nd te
xtua
l des
crip
-tio
ns w
ere
rate
d lo
wes
t. H
elpf
ulne
ss
ratin
gs v
arie
d si
gnifi
cant
ly b
etw
een
form
ats.
Mod
erat
e
Bru
ndag
e 20
15b
#39
Perc
eive
d us
eful
ness
Ran
dom
ized
cro
ss-s
ectio
nal,
mix
ed
met
hods
, bet
wee
n-su
bjec
tsSi
mpl
e lin
e gr
aph
of m
ean
scor
es o
ver
time,
line
gra
ph w
ith n
orm
s, lin
e gr
aph
with
con
fiden
ce in
terv
als,
bar
char
t of a
vera
ge c
hang
es, b
ar c
hart
base
d on
a re
spon
der d
efini
tion
(impr
oved
, sta
ble,
wor
sene
d), c
umu-
lativ
e di
strib
utio
n fu
nctio
n
Whe
n gr
aphs
dis
play
ed g
roup
-leve
l da
ta, r
atin
gs o
n us
eful
ness
wer
e hi
ghes
t for
sim
ple
line
grap
hs o
f m
ean
scor
es o
ver t
ime.
Mod
erat
e
Tol
bert,
201
9b#4
1Pr
efer
ence
Obs
erva
tiona
l cro
ss-s
ectio
nal,
mix
ed
met
hods
, with
in-s
ubje
cts
Pie
char
t, ba
r cha
rt, ic
on a
rray
61%
of p
atie
nt’s
com
men
ts re
gard
ing
pie
char
ts w
ere
code
d as
pos
itive
, 21
% o
f com
men
ts re
gard
ing
bar
char
ts a
nd 3
5% o
f com
men
ts re
gard
-in
g ic
on a
rray
s.
Mod
erat
e
1764 Quality of Life Research (2020) 29:1747–1766
1 3
Discussion
In this review, we summarized the literature on methods of communicating probabilistic information in oncological treatment decision-making processes.
For communication of disease progression information, we found that the type of framing has an influence on treat-ment choice. This has also been observed in communica-tion in general health care [8]; positive framing (in terms of survival instead of mortality) may increase acceptance of treatments. However, in this review, we could not draw any conclusions regarding the direction of the effect due to the incomparability of treatment choices and to the low methodological quality of the studies.
Furthermore, we found that limiting the amount of sur-vival information that the patient has to process at once in a graphical display could benefit the patients’ understanding. There is growing evidence supporting the so-called ‘less is more’ approach in the field of decision-making; simpler forms of communication can make it easier for patients to use the information during decision making [24]. Addition-ally, it has been argued that the complexity of information may contribute to patients’ experience of uncertainty [43].
We found that precise and defined risk information (e.g., percentages) about side effects was better understood than verbal information. Here, however, a potential source of bias could exist in the way that outcomes were measured. To prevent recall bias, the risk format used in the outcome assessment should not be similar to the format used as an intervention. For example, comparing percentages to another format would not be completely ‘fair’ when the answer to an accuracy question has to be stated in percentages, as is the case in at least two of three accuracy studies. However, find-ing a suitable assessment format that will not be influenced by recall might be challenging in these cases.
For display of HRQL information, we did not find con-sensus among the included studies on which type of graphs to use. Whereas line graphs and pie charts seemed to result in better cognitive and affective outcomes, these results were only based on one significant result each. While a previous review in general health care [8] recommends using icon arrays or bar charts to display outcome information, recent studies on the display of HRQL information specifically, suggested using pie charts [44, 45]. Another issue is the direction of display of HRQL data when using bar or line graphs. Where other literature in HRQL research—beyond the scope of this review—recommends the graphical display direction of better = higher [45, 46], we did not find enough evidence—one study—to recommend on the direction of HRQL graphs [40].
Great heterogeneity was found not only in the study results but also in methodology and in the compared
communication formats. This precludes us from stating separate practice recommendations for the three different types of risk information (disease progression, side effects and complications and HRQL). Most importantly, several studies investigated the same communication methods and found different effects on outcomes. Therefore, a meta-anal-ysis would be of great help. However, for meta-analysis to be possible, the heterogeneity