Patient values and preferences regarding VTE disease: a systematic review to inform American Society of Hematology guidelines

The American Society of Hematology (ASH), together with the MacGRADE Centre at McMaster University, developed clinical practice guidelines for management of venous thromboembolism disease (VTE) (see the 10 topics in supplemental Material).^1-7  To develop these guidelines, we followed an evidence synthesis and guideline development approach based on the GIN-McMaster Guideline Development Checklist⁸  and the Handbook in Grading of Recommendations Assessment, Development and Evaluation's (GRADE) official application GRADEpro (www.gradepro.org). We used the GRADE Evidence to Decision frameworks,^9-11  which integrate, among other criteria, patients’ values and preferences in balancing desirable and undesirable consequences of investigated options to arrive at evidence-based recommendations.

The GRADE approach defines patients’ values and preferences as the relative importance that people place on health outcomes (eg, bleeding, having a deep venous thrombosis [DVT]).¹²  Considering this importance is essential when weighing benefits and harms in guideline recommendations. Incorporating values and preferences in the guideline processes also promotes the development of recommendations with greater acceptability and adherence by those intended to benefit from them.¹²  Such recommendations are also useful to inform individual patient decision making.^12-15 

To inform the judgments by the guideline panels of the ASH VTE guidelines, we conducted a systematic review focused on patients’ values and preferences for the health outcomes prioritized by the panels.

We developed a protocol for this systematic review (PROSPERO 2018 CRD42018094003). The review was part of the overall project coordinated by the McMaster GRADE Center to support the development of the ASH VTE guidelines.^1-7  We adopted an inclusive systematic review methodology to reach breadth and depth of understanding of the topic under review.¹⁶  To do so, we systematically reviewed findings from quantitative studies and aggregated results from qualitative primary studies.^6-8  Information was shared with the systematic review team and the panel members using an interactive approach during which the experts were asked for additional suggestions about the research evidence and synthesis of the findings prior to the in-person panel meetings. For the purpose of this article, 2 reviewers with qualitative research experience (F.B. and I.D.F.). independently extracted, assessed, analyzed, and interpreted qualitative data and regularly met with other authors to discuss and integrate the analytical findings.

We developed sensitive electronic search strategies based on previously published work and a search filter for values and preferences,¹⁷  which incorporates terms to capture content areas (utilities, direct choice, structured scales or questionnaires, and qualitative studies) in combination with terms related to VTE disease, the interventions prioritized across the different guidelines, and terms defining prioritized health outcomes rated as critical by the panels.

We first searched Medline, Embase, the Cochrane Central Register of Controlled Trials, PsycINFO, and the Cumulative Index to Nursing and Allied Health Literature from inception to October 2016 (see “Search strategies” in supplemental Material) and we set up and monitored search alerts until April 2018 to identify and incorporate newly published research.

We included original studies assessing the values and preferences of individuals at risk or with a VTE-related condition that was covered as a topic in the guidelines. Supplemental Table 1 defines the selection criteria.

First, we screened titles and abstracts of a set of references following a well-defined selection framework (supplemental Figure 1). We conducted a piloting exercise to ensure consistency among 22 screeners, with different levels of expertise in values and preferences research (Yuan Zhang, I.E.-I., H.B., C.A.C., Y. Roldan, R. Chen, C.D., R.L.M., J.J.R., Yuqing Zhang, R. Charide, A.A., S.B., G.P.M., J.J.Y.-N., Y. Rehman, I.N., N.S., T.B., C.B., and M.F.R.), who screened sets of references independently and in duplicate. A third reviewer resolved disagreements when needed. The screening results, together with the content of the search strategy, the selection criteria, and the framework, were used to develop and calibrate a machine-learning screening model in the Collaboratron platform. The screening model allowed us to predict the probability of a citation being relevant for full-text screening. We chose a ≤1% probability of relevance, specified by the machine-learning model, as the threshold to exclude irrelevant references.

Second, the model was used in the remaining set of references, and the titles and abstracts of relevant records (those with >1% probability of relevance, determined by machine-learning model) were screened. Finally, the full texts of included records were independently assessed in duplicate, and disagreements were resolved by consensus or by consulting a third reviewer.

Seven reviewers (Yuan Zhang, I.E.-I., H.B., C.A.C., Y. Roldan, R. Chen, and C.D.) independently, and in duplicate, extracted data from studies using quantitative methods, and 2 other reviewers (F.B. and I.D.F.) extracted qualitative research study data. Using structured abstraction forms, we extracted 2 types of information for qualitative data: descriptive characteristics of studies (identification data, objectives, participant characteristics, study design and methodology, and country) and study findings, including original data excerpts (participant quotes, stories, or incidents). Disagreement was resolved through discussion and consensus. If necessary, a third party was consulted. QSR NVivo v.11 and Microsoft Excel software were used to extract and analyze the qualitative data.

For the synthesis of findings from quantitative-method studies, we present the raw data, ranges, and narrative descriptions, because, as expected, there was high heterogeneity in methods and measurement of outcomes that limited the ability to conduct any meta-analysis. We categorized the results into relative importance of outcomes (RIOs) or utilities and nonutility results. Utilities represent the strength of an individual’s preferences for different health outcomes. They are measured on an interval scale, with 0 reflecting states of health equivalent to being dead and 1 reflecting perfect health. The nonutility results were summarized in predetermined subgroups of patients who were the target population of the guidelines, such as medical patients, surgical patients, patients with cancer, or pregnant women.

For studies using qualitative methods, we followed an analytical approach adapted from grounded theory to extract and analyze qualitative data.^18,19  This inductive analytical technique, based on a constant comparison of data, suits the aims of aggregating qualitative evidence and providing new conceptual interpretations that integrate findings across studies.¹⁹  We did not apply an overall certainty of evidence framework like GRADE-CERQual,²⁰  because our search strategy was not sensitive enough to ensure a comprehensive identification of all of the qualitative evidence, and only the qualitative studies identified with our search strategy were included.

To assess the risk of bias (RoB) for individual quantitative studies, we used a priori defined items, including selection of study population, completeness of data, selection of measurement instruments, administration of measurement instruments, presentation of outcomes, understanding of participants, and data analysis.²¹  To assess the certainty of the body of evidence from quantitative studies, we applied the GRADE approach to assess the certainty of evidence for RIOs or values and preferences.^12,21,22  The approach includes the following domains: RoB, inconsistency, indirectness, imprecision, and publication bias. For qualitative research studies, we used the CASP Qualitative Checklist tool²³  to appraise the RoB of individual studies.

Our search strategy identified 72 814 records from the electronic database searches (61 130 nonduplicate records). After applying the machine learning (which required manual prescreening of a set of 10 196 records and its performance on the remaining 50 937 references), we selected and manually screened the title and abstract of 5259 records classified as having >1% probability of being relevant; 422 publications were retrieved for full-text screening. A set of 56 records identified from electronic database searches was selected for inclusion,^24-79  and 3 additional studies^80-82  were identified by checking references included in a previously published guideline on the same topic.⁸³  Overall, we included 44 quantitative studies,^{24-53,55-65,80-82}  14 qualitative studies,^66-79  and 1 study with both types of evidence⁵⁴  (Figure 1). Fourteen of the included studies reported patients’ values or utilities,^24-37  34 nonutility studies used quantitative methods,^{24,32,35,38-65,80-82}  and another 15 studies used qualitative methods.^54,66-79  Supplemental Table 2 lists the studies that did not fulfill the inclusion criteria with the reasons for exclusion at full-text screening.

Supplemental Table 3 summarizes the individual study characteristics, results, and the overall RoB assessment. Supplemental Table 4 summarizes the individual RoB criteria for utility and nonutility of the 45 quantitative studies.

We identified 14 studies,^24-37  including 2465 participants, reporting utility or health state values of DVT, pulmonary embolism (PE), postthrombotic syndrome (PTS) (mild and severe), and bleeding events, as well as utilities of using treatments such as vitamin K antagonists (VKAs) and low molecular weight heparin (LMWH) (Figure 2; Tables 1 and 2).

A total of 1702 participants from 6 studies,^{27,28,31,32,34,37}  including the overall population, provided their estimates of RIOs for DVT, with a range from 0.61 to 0.99 across studies (Table 1). Five studies with 1474 participants^{27,28,31,32,36}  reported an RIO for PE in a range from 0.63 to 0.93. Across different techniques to elicit utilities, standard gamble appeared to yield higher utilities (or lower disutilities) for DVT (0.81-0.99) and PE (0.75-0.93). Meanwhile, EuroQol-5 Dimension (EQ-5D) (EQ-5D) and Short-Form Six-Dimension (SF-6D) utilities were lower. The EQ-5D utility ranged from 0.61 to 0.79 for DVT and from 0.62 to 0.80 for PE. The SF-6D utilities are 0.64 for DVT and 0.68 for PE. In general, people may probably find DVT and PE as having a small to moderate impact on their lives, although there is likely important variability for this assessment. Because of the inconsistency and the potential variability observed, we downgraded the certainty of evidence to moderate for the DVT and PE outcomes.

For PTS, 1 study³²  based on 124 participants using time trade-off techniques, reported a utility of 0.82 (moderate certainty of evidence), suggesting a probable small impact on people’s lives. Two other studies elicited the utility for mild and severe PTS^30,35  with the standard gamble technique (66 participants from 2 studies) and reported utilities from 0.99 to 1.00 for mild PTS and from 0.93 to 0.98 for severe PTS (Table 1). This suggests that people might possibly find PTS having a trivial to small impact on their lives; however, most of the participants were healthy volunteers, and studies included small sample sizes. Therefore, we downgraded the certainty of evidence to low for the importance of mild and severe PTS as a result of indirectness and imprecision.

The included studies reported a variety of utilities for bleeding events. The utilities ranged from 0.59 to 0.65 for gastrointestinal tract bleeding (1217 participants from 3 studies),^27,31,32  0.76 for muscular bleeding (124 participants from 1 study),³²  from 0.29 to 0.60 for central nervous system bleeding (66 participants from 2 studies),^30,35  0.75 for minor intracranial bleeding, and 0.15 for major intracranial bleeding (216 participants from 1 study).²⁷  The certainty was graded as moderate for gastrointestinal and muscular bleeding due to indirectness, as very low for central nervous system bleeding, and as high for minor and major intracranial bleeding (Table 1).

One of the studies²⁴  assessing thromboprophylaxis during pregnancy using a feeling thermometer (anchored at death [0] and full health [100]) reported a utility value of 46 (interquartile range [IQR], 30-65) for pregnancy-related DVT and 30 (IQR, 15-50) for pregnancy-related PE and obstetrical bleeding. The certainty was graded as low for the DVT, PE, and bleeding utilities due to inconsistency and imprecision (Table 2).

The utilities participants place on different treatment options are summarized in Table 1 (overall population) and Table 2 (pregnant women). People may probably find the inconvenience of medication treatment ranging from having a trivial to an important impact on their lives. Utility values ranged from 0.38 to 0.99 for VKA treatment (296 participants from 4 studies).^25,29,32,33  For LMWH treatment, a range of 0.66 to 0.99 was reported in the general population (72 participants from 2 studies)^25,33  and 83 (IQR, 70-90) was reported in pregnant women (123 participants from 1 study).²⁴  Overall, results showed considerable variability, and certainty was judged as moderate or low. The observed variability is explained, in part, by the different approaches used to determine the RIO and, in part, by the diversity of the respondents, including the general population, patients who recently started treatment, and patients with long-term treatment experience.

We identified 34 quantitative studies,^{24,32,35,38-65,80-82}  including 6424 participants, reporting patients’ preferences or experiences with VTE management options. Tables 3 through 6 summarize the evidence and the certainty of RIOs for VTE management consequences (Table 3), burden of treatment (Table 4), different treatment alternatives (Table 5), and VTE diagnosis management (Table 6) classified by different subgroups of patients, such as medical or surgical patients receiving VTE prophylaxis, those receiving treatment of VTE, patients with cancer, or pregnant women. No evidence was identified for the population with heparin-induced thrombocytopenia or the pediatric population.

Studies assessing the RIOs that medical and surgical patients place on VTE prophylaxis benefits and adverse events showed that VTE risk reduction is a more important factor influencing their preferences than the potential harms of the treatment (510 participants from 3 studies; high certainty)^38,49,56  (Table 3). The same was true for patients treated for VTE, who held greater concerns for recurrent VTE risk (519 participants from 1 study; moderate certainty).⁵²  Studies also suggested that, although the majority of patients treated for VTE would like to avoid adverse events, only 21% to 25% are afraid of hemorrhagic events. Moreover, 87% of patients who had experienced a negative episode reported being “not afraid of” negative consequences (1019 participants from 4 studies; moderate certainty)^29,35,42,52  (Table 3). Similarly, studies including cancer patients and patients with cancer-related thrombosis suggest “the interference with cancer treatment” is the most important attribute influencing their preferences for the medication, followed by “efficacy of the VTE treatment” and “the risk of major bleeding” (509 participants from 2 conjoint analyses; moderate certainty)^46,54  (Table 3).

Sixteen studies evaluated patient RIOs for the burden associated with treatments.^{24,32,39,40,43-47,51-54,60,65,80}  One study, including patients taking long-term anticoagulation, reported that only some (12%) patients using direct oral anticoagulants (DOACs) may switch to VKAs because of complications and fear of adverse effects, as well as for not being reimbursed (subgroup of 19 participants from 1 study including 1001 total participants; low certainty)⁶⁵  (Table 4). With regard to the use of oral VKAs in general, although routine monitoring does not seem to represent a limitation for patients accepting it, many (58-64%) would switch to another anticoagulant if it were equally effective and required less monitoring or dietary restrictions (2070 participants from 5 studies; low certainty)^{40,45,47,52,65}  (Table 4). Two studies in cancer patients receiving LMWH treatment also showed lower RIOs for attributes such as monitoring through blood tests, frequency of administration, mistakes, and costs (509 participants from 2 studies; moderate certainty)^46,54  (Table 4). One study evaluating elastic compression stockings showed that participants would accept an increase in the risk of complications if they could put them on by themselves (300 participants from 1 study; moderate certainty)⁴³  (Table 4, treatment subgroup). Another study using a direct-choice exercise showed that, a majority of women with a previous VTE, who were at low or high risk for recurrent VTE (86% and 60%, respectively), and who were pregnant or planning a pregnancy, were willing to take LMWH (123 participants from 1 study; moderate certainty)²⁴  (Table 4).

Sixteen studies, including 3452 participants, assessed RIOs for different treatment alternatives (Table 5).^{41,42,49,50,52-57,59,60,62-64,82}  Six of those studies evaluated the preference for injections or oral administration routes in 1645 medical or surgical patients receiving prophylaxis.^{49,55,56,60,62,64}  There was a clear preference for oral medication (range, 60-86% of patients) over subcutaneous injection, because it is easier to take and to integrate into the daily routine, and it is also perceived as less painful and less expensive. For participants preferring the injection route, the stated reasons for the preferences included faster onset of action, avoidance of pill burden, and ease of use (1645 participants from 6 studies; moderate certainty).

Lastly, we found 1 study addressing patient preferences related to different diagnostic management options. A survey assessed preferences for the use of computed tomography pulmonary angiography (CTPA) for PE diagnosis in a hypothetical scenario of low pretest probability compared with not using it when results of D-dimer testing left patients with uncertainty. The majority of participants (63%) favored undergoing CTPA, regardless of its adverse effects. Most patients (85%) who accepted CTPA testing had concerns about missing a PE (203 participants from 1 study; high certainty)⁴⁸  (Table 6).

The 15 included qualitative studies involved individuals (570 participants) who had, or were at risk for having, VTE.^54,66-79  Supplemental Table 5 summarizes the individual study characteristics, and supplemental Table 6 provides a descriptive summary of the study designs, qualitative methodologies, and study locations. Supplemental Table 7 reports a narrative assessment of the major strengths and weaknesses of included studies, based on the CASP Qualitative Checklist tool.²³ 

Table 7 outlines the 4 analytic themes that summarize patients’ experiences and perspectives when thinking about VTE disease management. For each theme, we describe how patients understand and experience the intervention and its outcomes. In the following sections, we summarize the 2 analytic themes related to the importance of outcomes.

Diagnosis of VTE and treatment benefits brought feelings of relief, empowerment, and control^{54,70,71,74,75,78}  that increased patient motivation and self-discipline to initiate and pursue the treatment while acknowledging its inconveniences. VTE treatment provided a solution to live without fear and restrictions^68,70 ; patients described treatment as going “back to normal”^54,67,76-78  by learning to self-manage their condition and reprioritize their life goals and habits.^70,71,74,77  Patients expressed preferences for heparin injections, because these relieved them from the burdens of constant international normalized ratio monitoring.^54,75,76  For pregnant women, the diagnosis also provided reasons for past miscarriages and the treatment hopes for a healthy current pregnancy.⁷¹ 

In contrast, patients revealed feelings of distress, emotional discomfort, and hopelessness due to the unpredictability of the disease and recurrence of VTE episodes.^66,70,71,78  Most patients described the fear of the uncertainty of bleeding events, sudden death,^66,74,76  thrombophilia diagnosis, and the distressful feelings caused by misdiagnosis.^54,68,71  The burden of daily self-injections, the discomfort of stocking use, and the antithrombotic treatments’ side effects, such as the bleeding risk and bruising events, were concerns related to the treatment raised across the included studies.^67,68,74,79 

Personal risk assessment of VTE events influences patients’ treatment preferences and decisions for undergoing prophylaxis and treatment. When patients lack clarity in understanding the risks of the disease,^72,76  the need for treatment,^{54,66,67,71,72,74,76}  and the treatment’s side effects,^66,74,76  initiating treatment in the absence of symptoms might be confusing.^{54,66,67,74,76}  Overall, understanding the rationale and benefits for using anticoagulation treatment made patients more aware and accepting of the need for the intervention.^{66,70,72,74,76,78} 

We systematically reviewed quantitative and qualitative evidence of patients’ values and preferences related to VTE. The review directly informed ASH guidelines for management of VTE, aiming to inform guideline panel members’ decision making about the balance between benefits and harms in prophylaxis, treatment, and diagnosis of VTE.

This systematic review has a number of strengths. First, we used a comprehensive literature search strategy and set no time or language limit on the search, ensuring the comprehensiveness and directness of summarized evidence specific to VTE. We developed a machine-learning algorithm to balance efficiency and accuracy of screening a broad area of the literature. The training and calibration exercise, together with the 10 196 records initially screened for the algorithm development, gave us high confidence in the accuracy of the process. Second, we used the GRADE concept of outcome importance to determine the eligibility criteria, including studies reporting the importance that patients place on health outcomes, as well as studies about preferences for or against a treatment, because they implicitly provide us the RIOs.¹²  We also summarized the results within the GRADE Evidence-to-Decision framework for guideline decision making.^10,14 

Moreover, this review includes quantitative and qualitative research evidence on the topic, which enabled us to reach a wide breadth and depth of understanding and substantive knowledge of patients’ values and preferences. Finally, for the quantitative studies, we assessed the RoB of the included individual studies, as well as the certainty of the overall body of evidence. For qualitative research, we used the CASP Qualitative Checklist tool to appraise the quality of each included study to help readers assess the trustworthiness of our synthesis.²³ 

Our study also has limitations. Because of time constraints, we used the machine-learning model to rule out irrelevant records, which was only validated in a set of the complete reference list. However, to overcome the risk of losing relevant information, we consulted our guideline panels and compared our included studies list with other systematic reviews on similar topics^84-87  to identify potentially missed relevant studies. Additionally, although we included quantitative and qualitative studies, the search strategy may not have been sensitive enough to ensure the identification of the full body of qualitative evidence, and the inclusion of these types of studies could be considered unsystematic. Nevertheless, our aim of aggregating identified qualitative studies’ findings was to complement quantitative findings rather than synthesizing the entire body of evidence. We consider our approach appropriate to summarize patient values and preferences related to VTE management, although we also acknowledge the opportunity for further improvement in the synthesis of the qualitative body of evidence.

Findings from quantitative research studies show high variability in the utility values for DVT, PE, and PTS, with a certainty of evidence varying from low to moderate, primarily due to the diverse methods used to obtain the RIOs and the diversity of the patient populations included in the studies. Overall, the studies suggest VTE-related health states have small to important impacts on patients’ lives. Qualitative findings also corroborate quantitative results regarding patients’ preferences for oral medication over subcutaneous medication. However, because all of the included qualitative studies focused on VKAs, which still require constant international normalized ratio monitoring, rather than on DOACs, as well as on subcutaneous injections, overall, they report similar preferences for treatment modalities that do not require monitoring or dietary changes. Finally, quantitative and qualitative findings reveal patients’ positive perceptions of treatment and diagnosis when informed and when included in the treatment decision-making process.

In this systematic review, we only included studies on VTE, excluding other diseases for which anticoagulation treatment might also be indicated. Results regarding RIOs were similar to another systematic review⁸⁸  that included, among other diseases, 3 studies on VTE disease, illustrating the significant variability in patients’ values and preferences regarding prophylaxis and treatment. In our systematic review, the highly sensitive search strategy allowed us to identify and include more studies, and the new machine-learning approach positively impacted the efficiency of the screening process.

In conclusion, the results of our review show high variability in how patients value the impact of VTE-related health outcomes, which is partially explained by the different approaches used to determine the RIOs and the diversity of the included participants. These findings highlight the necessity to explore methods-induced variability, which might be facilitated by the use of health outcome descriptors, patient scenarios, or hypothetical health states,^89,90  to differentiate true variability from methods-induced variability.

The authors thank Rachel Couban (McMaster University) for reviewing the search strategy and translating it into PsycINFO and Cumulative Index to Nursing and Allied Health Literature database languages and Daniel Perez Rada (Epistemonikos Foundation) and Aldo Canepa (Epistemonikos Foundation) for setting up and running the Collaboratron platform.

This work was supported by ASH to inform guideline recommendations in the ASH Guidelines for VTE Management.

Contribution: Yuan Zhang, I.E.-I., and H.J.S. designed the study; I.E.-I., Yuan Zhang, W.W., R.N., F.B., I.D.F., H.B., C.A.C., Y. Roldan, R. Chen, C.D., R.L.M., J.J.R., Yuqing Zhang, R. Charide, A.A., S.B., G.P.M., J.J.Y.-N., Y. Rehman, I.N., N.S., T.B., C.B., and M.F.R. screened the literature and/or abstracted the data; I.E.-I., Yuan Zhang, F.B., I.D.F., and H.J.S. drafted the manuscript; H.J.S., W.W., and R.N. coordinated the overall project; H.J.S. is cochair of the GRADE working group and was the principal investigator for this project; and all authors read and approved the final version of the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Itziar Etxeandia-Ikobaltzeta, McMaster University, 1280 Main St West, Hamilton, ON L8S 4K1, Canada; e-mail: itzi.etxe@gmail.com.