Combined oral contraceptives: venous thrombosis
Abstract
This is a protocol for a Cochrane Review (Intervention). The objectives are as follows:
The objectives of this review are:
-
to estimate venous thrombosis risk associated with COC use compared with non‐use;
-
to perform a direct comparison of the risk associated with the three generations of COCs;
-
to compare the effect of estrogen doses and types of progestogen on venous thrombosis risk.
Background
Description of the condition
Venous thrombosis comprises deep‐vein thrombosis (DVT) and pulmonary embolism. DVT typically starts in the calf veins, from where it may extend to the proximal veins and subsequently cause pulmonary embolism (Kearon 2003). Approximately one‐third of patients with symptomatic venous thrombosis manifest pulmonary embolism (White 2003; Huerta 2007). Venous thrombosis is associated with genetic (i.e., carriers of thrombophilic disorders and a positive family history for venous thrombosis) and acquired risk factors (i.e., surgery, trauma, marked immobility, pregnancy, hormonal replacement therapy, previous venous thrombotic event, active cancer). In women of reproductive age, an important risk factor is oral contraceptive use. Oral contraceptives and inherited thrombophilic defects (i.e., factor V Leiden mutation, deficiency of protein C, protein S or antithrombin, high levels of factor VIII, and prothrombin mutation) interact synergistically to increase the risk of venous thrombosis (Bloemenkamp 2003; Huerta 2007; Naess 2007).
Venous thrombosis in women has an incidence of 1.6 per 1000 person‐years. Incidence rates increase with age: women aged 30 to 34 years show an incidence of 0.25 per 1000 person‐years and women aged 60 to 64 years, 0.93 per 1000 person‐years (Naess 2007). Others have estimated the incidence in women during the reproductive years to be in the range of 0.5 to 1.0 per 1000 person‐years (Heinemann 2007). Despite the low incidence of venous thrombosis among women of reproductive age, the impact of oral contraceptives on the risk is large since it is estimated that more than 100 million women worldwide use an oral contraceptive (WHO 1998). Moreover, venous thrombosis is associated with an increased mortality risk. Overall, the 30‐day case fatality rate is higher in patients with pulmonary embolism than in those with DVT (9.7% to 12% versus 4.6% to 6%) (White 2003; Huerta 2007; Naess 2007). In women from 15 to 44 years of age the venous thrombosis‐associated mortality rate is lower (0.6% to 1.7%) (Lidegaard 1998).
DVT may damage deep venous valves with venous reflux and venous hypertension in the lower limbs, resulting in a post‐thrombotic syndrome (PTS). PTS is characterized by pain, heaviness, and swelling of the leg aggravated by standing or walking (Kearon 2003). PTS may develop in half of all DVT patients within three months, with no further increase being seen up to two years of follow‐up (Tick 2010). Complete resolution of pulmonary embolism occurs in about two‐thirds of patients, with partial resolution in the remainder. However, chronic thromboembolic pulmonary hypertension may occur in up to 5% of pulmonary embolism patients (Kearon 2003).
Description of the intervention
The first combined oral contraceptive (COC) was introduced in 1960 (Enovid®). It consisted of 150 μg mestranol, an estrogen, and 985 mg norethynodrel, a progestogen. Shortly after, the first case of venous thrombosis associated with COC was reported (Jordan 1961). Since then many studies have established the association between COC use and occurrence of venous thrombosis (van Hylckama Vlieg 2011).
Several large studies in the 1990s confirmed a two‐ to four‐fold increase in the risk of venous thrombosis associated with COC use (Thorogood 1992; Vandenbroucke 1994; WHO 1995; Farmer 1997). Since the estrogen compound in COC was thought to cause the increased risk, the dose of estrogen has been gradually lowered from 150 to 100 μg to 20 μg in the 1970s (Stolley 1975; Wharton 1988; Thorogood 1993). The lower dose of ethinylestradiol in contraceptives was indeed associated with a reduction in the venous thrombosis risk (Inman 1970; Meade 1980; Vessey 1986; WHO 1995; Lidegaard 2002). The oral contraceptives currently prescribed which contain 30 μg of ethinylestradiol are associated with a higher risk of venous thrombosis than contraceptives containing 20 μg (Lidegaard 2009; van Hylckama Vlieg 2009).
Besides adjustments in the dose of ethinylestradiol, the progestogen compound was changed to reduce the side effects of the COC. After the first‐generation progestogens, new progestogens were developed in the 1970s and 1980s(second and third‐generation progestogens, respectively). It was shown that third‐generation COC users had a higher risk of venous thrombosis than second‐generation users (Kemmeren 2001; Vandenbroucke 2001; Lidegaard 2009; van Hylckama Vlieg 2009). However, these results were disputed: it was reasoned that bias or confounding could explain the difference in venous thrombosis risk between the progestogen generations. These issues were addressed in an opinion article and a meta‐analysis in which it was shown that the presence of bias or confounding could not explain the observed results (Vandenbroucke 1997; Kemmeren 2001).
Other progestogens have been developed since the introduction of the third‐generation progestogens, i.e., drospirenone (2001) and dienogest (1995). The use of drospirenone in a COC has been shown to increase the risk of venous thrombosis (Lidegaard 2009; van Hylckama Vlieg 2009), compared with non‐use and compared with second‐generation contraceptives (Jick 2011; Parkin 2011). However, no information concerning the risk of venous thrombosis is available for the contraceptive containing dienogest, mainly used in Germany (Kuhl 1998).
How the intervention might work
The use of COCs affects hemostasis in many ways. It increases factors involved in coagulation or indicative of increased activity of this system (i.e., factor II, factor VII, and factor VIII, prothrombin fragment 1+2, D‐Dimer). Natural anticoagulant factors are also affected, for example, the anticoagulant protein C is increased whereas other anticoagulation factors are decreased (i.e., antithrombin and protein S) in COC users. This trend is more pronounced in third‐generation COC users than in second‐generation users (Vandenbroucke 2001; Kemmeren 2002a; Kemmeren 2002b; Kemmeren 2004).
Besides these individual coagulation factors, the measurement of activated protein C (APC) resistance provides insight into the overall balance of coagulation (Vandenbroucke 2001). There are two APC resistance assays for probing the plasma response to APC (the endogenous thrombin potential assay and the activated partial thromboplastin time (APTT)‐based assay). The two assays rely on different coagulation triggers and endpoints and they probe different coagulation reactions. In summary, APC resistance evaluates the relative inability of protein C to cleave activated factors V and VIII leading to a prothrombotic state (Vandenbroucke 2001; Castoldi 2010). APC resistance predicts venous thrombosis risk in men and in women, as well as in COC users and non‐users (Tans 2003). Several studies have confirmed that APC resistance is increased in COC users (Kemmeren 2004; Rad 2006; Kluft 2008) and the effect is more pronounced in users of a third‐generation progestogen than with a second‐generation progestogen (Kemmeren 2004).
Why it is important to do this review
Since the introduction of the third‐generation progestogens, new progestogens have been introduced, such as nestorone, dienogest, nomegestrol acetate and spirolactone derivates, trimegestone, and drospirenone (Sitruk‐Ware 2006). Many studies compare these new COCs to a COC containing levonorgestrel, which is assumed to have the lowest risk of venous thrombosis (Gomes 2004; Jick 2011; Lidegaard 2011). We will set out to review the association between COC and risk of venous thrombosis at the level of different COCs, including the potential risk associated with COCs containing new progestogens. Specifically, we will perform a network meta‐analysis to compare one COC to another or to non‐use. Network meta‐analysis allows not only the comparison of two treatments but also a simultaneous comparison of several competing treatments, even where few or no direct comparisons exist. In addition, assessment of effect may be more realistic because it is based on a much larger body of evidence than in conventional meta‐analysis (Jansen 2008; Thijs 2008). In this network analysis we will take into account not only the progestogen used in the COC but also the estrogen dose. The rationale of the present systematic review proposal is to provide an update on the venous thrombosis risk associated with COC formulations and to perform a network meta‐analysis on the estrogen dosage and progestogen component of COCs.
This protocol was established before we developed a review that was published in September 2013 (Stegeman 2013). Reasons for not publishing the protocol before publication of the review were publication rights and unity between the protocol and The Cochrane Library/BMJ review.
For abbreviations, we refer to Table 1.
| Specific abbreviations | Explanation |
| APC | Activated protein C |
| APTT | Activated partial thromboplastin time |
| CC | Case‐control study |
| COC | Combined oral contraceptive |
| CT | Computed axial tomography |
| DVT | Deep‐vein thrombosis |
| MRI | Magnetic resonance imaging |
| NCC | Nested case‐control study |
| PCS | Prospective cohort study |
| RCT | Randomized controlled trial |
| V/Q | Ventilation‐perfusion |
Objectives
The objectives of this review are:
-
to estimate venous thrombosis risk associated with COC use compared with non‐use;
-
to perform a direct comparison of the risk associated with the three generations of COCs;
-
to compare the effect of estrogen doses and types of progestogen on venous thrombosis risk.
Methods
Criteria for considering studies for this review
Types of studies
Observational studies on adverse effects may provide valid evidence on unintended effects of treatment as they are often unpredictable and not linked to indications for treatment (Vandenbroucke 2004; Vandenbroucke 2006; Vandenbroucke 2008). Empirical evidence suggests that there may be no difference on average in side effects risk estimates of an intervention derived from meta‐analyses of randomized controlled trials (RCT) and meta‐analyses of observational studies. Therefore it seems reasonable not to restrict systematic reviews of adverse effects only to a specific study type (Golder 2011) and also because there is a paucity of experimental data on side effects. Thus, systematic reviews on the harms of interventions often come from observational studies. Observational studies in this review will include case‐control, cohort, and nested case‐control designs. If available, RCTs will also be evaluated and included. Study design criteria are described in Table 2 and Table 3.
| Question and checklist | RCT | PCS | RCS | NCC | CC |
| Was there a comparison: |
|
|
|
|
|
| Between two or more groups of participants receiving different interventions? | Y | Y | Y | Y | Y |
| Within the same group of participants over time? | P | N | N | N | N |
| Were participants allocated to groups by: |
|
|
|
|
|
| Concealed randomization? | Y | N | N | N | N |
| Quasi‐randomization? | N | N | N | N | N |
| Other action of researchers? | N | N | N | N | N |
| Time differences? | N | N | N | N | N |
| Location differences? | N | P | P | NA | NA |
| Treatment decisions? | N | P | P | N | N |
| Participants' preferences? | N | P | P | N | N |
| On the basis of outcome? | N | N | N | Y | Y |
| Some other process? (specify) |
|
|
|
|
|
| Which parts of the study were prospective: |
|
|
|
|
|
| Identification of participants? | Y | Y | N | Y | N |
| Assessment of baseline and allocation to intervention? | Y | Y | N | Y | N |
| Assessment of outcomes? | Y | Y | P | Y | N |
| Generation of hypotheses? | Y | Y | Y | Y | P |
| On what variables was comparability between groups assessed: |
|
|
|
|
|
| Potential confounders? | P | P | P | P | P |
| Baseline assessment of outcome variables? | P | P | P | N | N |
RCT = randomized clinical trial
PCS = prospective cohort study
RCS = retrospective cohort study
NCC = nested case‐control study
CC = case‐control study
Y = yes
N = no
P = possibly
NA = not applicable
| Note: Users need to be very clear about the way in which the terms 'group' and 'cluster' are used in these tables. The above table only refers to groups, which is used in its conventional sense to mean a number of individual participants. With the exception of allocation on the basis of outcome, 'group' can be interpreted synonymously with 'intervention group'. Although individuals are nested in clusters, a cluster does not necessarily represent a fixed collection of individuals. For instance, in cluster‐allocated studies, clusters are often studied at two or more time points (periods) with different collections of individuals contributing to the data collected at each time point. Was there a comparison? Typically, researchers compare two or more groups that receive different interventions; the groups may be studied over the same time period, or over different time periods (see below). Sometimes researchers compare outcomes in just one group but at two time points. It is also possible that researchers may have done both, i.e., studying two or more groups and measuring outcomes at more than one time point. How were participants/clusters allocated to groups? These items aim to describe how groups were formed. None will apply if the study does not compare two or more groups of participants. The information is often not reported or is difficult to find in a paper. The items provided cover the main ways in which groups may be formed. More than one option may apply to a single study, although some options are mutually exclusive (i.e., a study is either randomized or not). ‐ Randomization: Allocation was carried out on the basis of truly random sequence. Check carefully whether allocation was adequately concealed until participants were definitively recruited. ‐ Quasi‐randomization: Allocation was done on the basis of a pseudo‐random sequence, e.g., odd/even hospital number or date of birth, alternation. Note: when such methods are used, the problem is that allocation is rarely concealed. ‐ By other action of researchers: this is a catch‐all category and further details should be noted if the researchers report them. Allocation happened as the result of some decision or system applied by the researchers. For example, participants managed in particular 'units' of provision (e.g. wards, general practices) were 'chosen' to receive the intervention and participants managed in other units to receive the control intervention. ‐ Time differences: Recruitment to groups did not occur contemporaneously. For example, in a historically controlled study participants in the control group are typically recruited earlier in time than participants in the intervention group; the intervention is then introduced and participants receiving the intervention are recruited. Both groups are usually recruited in the same setting. If the design was under the control of the researchers, both this option and 'other action of researchers' must be ticked for a single study. If the design 'came about' by the introduction of a new intervention, both this option and 'treatment decisions' must be ticked for a single study. ‐ Location differences: Two or more groups in different geographic areas were compared, and the choice of which area(s) received the intervention and control interventions was not made randomly. So, both this option and 'other action of researchers' could be ticked for a single study. ‐ Treatment decisions: Intervention and control groups were formed by naturally occurring variation in treatment decisions. This option is intended to reflect treatment decisions taken mainly by the clinicians responsible; the following option is intended to reflect treatment decisions made mainly on the basis of participants' preferences. If treatment preferences are uniform for particular provider 'units', or switch over time, both this option and 'location' or 'time' differences should be ticked. ‐ Patient preferences: Intervention and control groups were formed by naturally occurring variation in patients' preferences. This option is intended to reflect treatment decisions made mainly on the basis of patients' preferences; the previous option is intended to reflect treatment decisions taken mainly by the clinicians responsible. ‐ On the basis of outcome: A group of people who experienced a particular outcome of interest were compared with a group of people who did not, i.e., a case‐control study. Note: this option should be ticked for papers that report analyses of multiple risk factors for a particular outcome in a large series of participants, i.e. in which the total study population is divided into those who experienced the outcome and those who did not. These studies are much closer to nested case‐control studies than cohort studies, even when longitudinal data are collected prospectively for consecutive patients. Which parts of the study were prospective? These items aim to describe which parts of the study were conducted prospectively. In a randomized controlled trial, all four of these items would be prospective. For non‐randomized trials (NRS) it is also possible that all four are prospective, although inadequate detail may be presented to discern this, particularly for generation of hypotheses. In some cohort studies, participants may be identified, and have been allocated to treatment retrospectively, but outcomes are ascertained prospectively. On what variables was comparability of groups assessed? These questions should identify 'before‐and‐after' studies. Baseline assessment of outcome variables is particularly useful when outcomes are measured on continuous scales, e.g., health status or quality of life. Response options Try to use only 'Yes', 'No', and 'Can't tell' response options. 'NA' should be used if a study does not report a comparison between groups. |
Types of participants
Participants will be healthy women taking a COC. We will exclude studies of women on postmenopausal hormone replacement therapy, studies of women taking non‐oral or progestogen‐only contraceptives, and studies of women with venous thrombosis recurrence.
Types of interventions
COC use will be compared with non‐use or with a reference COC (for example, levonorgestrel with 30 μg of ethinylestradiol). We define a woman as a COC non‐user when either she had never been exposed to a COC or she was a former/previous COC user.
As there is no generally accepted way to classify COC according to generation of progestogen, we will classify as 'first‐generation' COCs those including lynestrenol and norethisterone as progestogens. 'Second‐generation' COCs will include norgestrel and levonorgestrel, while 'third‐generation' COCs will include desogestrel, gestodene, or norgestimate as progestogens. Therefore, we will classify COCs by progestogen generation independently of ethinylestradiol dose. Whenever another COC generation classification is employed by the researchers, we will also kept the original generation classification data so we can evaluate the effect of COC generation classification on venous thrombosis risk (Henzl 2000; Sitruk‐Ware 2008). We will also categorize COCs according to estrogen dose and to progestogen type.
Types of outcome measures
The outcome will be fatal or non‐fatal first venous thrombosis events (DVT or pulmonary embolism). We will classify outcomes according to diagnostic criteria as:
-
strict diagnostic outcome and specified criteria for venous thrombosis;
-
discharge diagnoses from wards, but without a priori specified outcome criteria;
-
ad hoc outcome selection of venous thrombosis patients not specified in advance.
We will include these outcome measures in the data abstraction form and we will evaluate them in a sensitivity analysis. The outcome classification will be assessed independently by two review authors (MdB, BHS) and disagreements will be resolved by consensus.
Primary outcomes
The primary outcome will be fatal or non‐fatal first venous thrombosis event (DVT or pulmonary embolism).
Secondary outcomes
Not applicable.
Search methods for identification of studies
The search will be created in association with an expert librarian (JW Schoones, Walaeus Library, LUMC, Leiden, NL). The search strategy is shown in Appendix 1.
Electronic searches
We will search the following databases: the Cochrane Database of Systematic Reviews, MEDLINE, EMBASE, Web of Science, CINAHL, Academic Search Premier, and ScienceDirect. We will amend the search strategy for each database. We will set no language restriction on the study search.
Searching other resources
In addition, we will check the references of the selected studies and of any reviews identified.
Data collection and analysis
We will analyze the study results by comparing the venous thrombosis relative risk between COC users and non‐users and comparing different types and dosing of COC components based on a network meta‐analysis.
We will use standard piloted forms for study selection, 'Risk of bias' assessment, and data abstraction. Study selection forms will include study identification, inclusion/exclusion criteria, standard study design classification, intervention and outcome evaluation, exposure ascertainment, and completeness of results.
Selection of studies
Two review authors (MdB, BHS) will independently evaluate the title and abstract of each study in the study search for study retrieval using standard piloted forms and specific inclusion and exclusion criteria. Disagreements will be resolved by consensus and a third author (OMD) will be consulted if disagreement persists.
Data extraction and management
Two review authors (MdB, BHS) will independently perform data extraction using standard, piloted forms. We will extract details of methods (i.e., participants, age), intervention/exposure (i.e., hormone type, dosage, exposure ascertainment), study comparison, outcome criteria assessment (as defined in Types of outcome measures section), results (i.e., number of participants, sample size, number of events, adjusted and unadjusted measure of effect, absolute risk evaluation), and other variables (i.e., funding source, first time users). Any disagreements will be resolved by discussion and a third author (OMD) will be consulted if disagreement persists.
Assessment of risk of bias in included studies
Tools for assessing quality in clinical trials are well‐described but much less attention has been given to similar tools for observational studies. Although the Newcastle‐Ottawa tool is frequently used to assess observational studies, the reliability or validity is unknown (Deeks 2003; Sanderson 2007). Since the Newcastle‐Ottawa tool is not customized for case‐control study designs, and as many case‐control studies of COCs are available, we have customized a version of the Newcastle‐Ottawa tool for the research question (Higgins 2011). According to study design (case‐control or cohort designs), were have customized slightly different 'Risk of bias' assessment questions:
-
For participant selection in case‐control study designs and outcome assessment in cohort study designs, we will customize the following question: 'Was there a (pre)defined outcome assessment?' Possible options include 'Venous thrombosis objectively confirmed in all included cases'; 'Not all venous thrombosis objectively confirmed'; and 'Unclear'. The criteria for venous thrombosis objectively confirmed include DVT event diagnosed by plethysmography, ultrasound examination, computed tomographic scanning, magnetic resonance imaging (MRI), or venography; or when a pulmonary embolism event was diagnosed by ventilation‐perfusion (V/Q) scanning, multidetector helical computed axial tomography (CT), or pulmonary angiography (Goodacre 2006; Qaseem 2007), or by other strict diagnostic and specified criteria for venous thrombosis. Low risk of bias will be defined as venous thrombosis reported as objectively confirmed in all cases.
-
For participant selection in case‐control studies we will customize the question: 'Was the control sampling adequate?' Possible options include 'Yes, with controls truly representing the source population (community controls)'; 'No, with controls not representing the source population'; 'Unclear'. In cohort studies we will customize the question: 'Was the selection of the non‐exposed cohort adequately performed?' Possible options include 'Drawn from the same community as the exposed cohort'; 'Drawn from a different source'; 'No description of the derivation of the non‐exposed cohort'. This item will be assessed when the control or the non‐exposed participants were derived from the same population as the cases or the exposed participants. Low risk of bias will be defined as a study with controls or non‐exposed participants sampled from the source population or from the same community as exposed participants, respectively.
-
For both study designs, we will customize a question evaluating whether or not there were adjustments for confounding performed either in the analysis or by design (matching). Low risk of bias is defined as adjustment for age and calendar time.
-
Regarding exposure evaluation, the customized question for both study designs will be: 'Was COC utilization properly assessed?' Possible options for case‐control study designs include 'Database record' (i.e., drug deliverance records); 'Interview not blinded to case/control status'; 'Written self report or medical record only'; and 'No description'. For cohort study designs, the options include: 'Database record' (i.e., COC prescription deliverance records); 'Structured interview with interviewer blinded'; 'Written self report'; and 'No description'. Low risk of bias is defined as a database record selection or written self report in cohort design.
-
For cohort study designs, we will customize a further question regarding the possibility of loss to follow‐up. Possible options include 'complete follow‐up' (i.e., all participants accounted for); 'Participants lost to follow‐up unlikely to introduce bias' (i.e., less than 10% of the trial population lost to follow‐up); 'Follow‐up rate potentially leading to bias' (i.e., more than 10% of the trial population lost to follow‐up); and 'No statement'. So, for this question, the cut‐off point was 10% and low risk of bias is defined as studies with complete or over 90% follow‐up (Kristman 2005).
We will not use the 'Risk of bias' assessment to accept or reject studies. However, we will produce a table describing 'Risk of bias' assessment for the included studies. Two independent review authors (MdB, BHS) will assess risk of bias using a standard piloted form. Any persistent disagreement will be resolved by discussion with a third author (OMD).
Measures of treatment effect
We will extract effect estimates from observational studies or RCTs. Effect estimates can be either odds ratios (RCT, cohort studies, and case‐control studies) or risk ratios (RCT and cohort studies). We will extract or recalculate accompanying 95% confidence intervals based on standard errors or P values.
Unit of analysis issues
The unit of analysis will be a healthy women using COC specified by ethinylestradiol dose and progestogen type.
Dealing with missing data
The denominator for each outcome in each study will be the number of participants minus any participants whose outcomes were known to be missing.
Assessment of heterogeneity
For heterogeneity calculation we will use the standard deviation/variance of the effect between studies. We will explore possible reasons for heterogeneity (i.e., participants and intervention) whenever the number of studies allows. Study class with 0 (zero) events will be inflated to 0.5. Indirect comparisons will use a random‐effects model. We will consider results heterogeneous whenever homogeneity is unlikely, that is a low P value (< 0.10) in the Chi² test for heterogeneity.
Assessment of reporting biases
We will investigate reporting biases (such as publication bias) using a funnel plot. After visual inspection for asymmetry we will use the linear regression test for asymmetry proposed by Egger (Egger 1997).
Data synthesis
We will calculate the meta‐analysis adjusted odds ratios by pooling adjusted odds ratios from individual studies, weighting individual study results by the inverse of their variance. For included studies, we will note levels of attrition.
If we cannot find explanations for heterogeneity, we will consider using a random‐effects model with appropriate cautious interpretation. We will used tables for graphical representation of the individual study point estimates and their associated 95% CI.
For the network meta‐analysis, we will select study categories for comparisons whenever there is at least one study with a specific comparison between estrogen dose or progestogen type. One can calculate indirect comparisons between two strategies by examining studies that contrast each strategy against a third 'reference' intervention. We will first derive pooled estimates from standard direct ('head‐to‐head') comparisons and then undertake indirect comparisons for estrogen dosing and progestogen type evaluations. We will estimate the comparisons in a pair‐wise manner combining all direct ('head‐to‐head') and indirect evidence in a single joint analysis (network meta‐analysis), using a log odds model with a random‐effects model. Graphic representation of the results will be made by a matrix representing each comparison.
The extent of disagreement between direct and indirect evidence will also be quantified by the incoherence of the network (Thijs 2008). We will also perform a conventional meta‐analysis comparing COCs by progestogen generation. We will perform most of the statistical analyses, including the network analysis, with a STATA package (Stata 2011).
Subgroup analysis and investigation of heterogeneity
To explore substantial heterogeneity, we will perform subgroup analysis and sensitivity analysis (study design and funding). Funding is defined as any study receiving money from pharmaceutical companies.
Sensitivity analysis
We will carry out sensitivity analyses to explore heterogeneity regarding study design, outcome certainty (venous thrombosis objectively confirmed), and source of funding. To determine the stability of the overall risk estimate, we will perform sensitivity analysis in which each design, outcome, and funding source category will be individually observed by progestogen generation.
| Specific abbreviations | Explanation |
| APC | Activated protein C |
| APTT | Activated partial thromboplastin time |
| CC | Case‐control study |
| COC | Combined oral contraceptive |
| CT | Computed axial tomography |
| DVT | Deep‐vein thrombosis |
| MRI | Magnetic resonance imaging |
| NCC | Nested case‐control study |
| PCS | Prospective cohort study |
| RCT | Randomized controlled trial |
| V/Q | Ventilation‐perfusion |
| Question and checklist | RCT | PCS | RCS | NCC | CC |
| Was there a comparison: |
|
|
|
|
|
| Between two or more groups of participants receiving different interventions? | Y | Y | Y | Y | Y |
| Within the same group of participants over time? | P | N | N | N | N |
| Were participants allocated to groups by: |
|
|
|
|
|
| Concealed randomization? | Y | N | N | N | N |
| Quasi‐randomization? | N | N | N | N | N |
| Other action of researchers? | N | N | N | N | N |
| Time differences? | N | N | N | N | N |
| Location differences? | N | P | P | NA | NA |
| Treatment decisions? | N | P | P | N | N |
| Participants' preferences? | N | P | P | N | N |
| On the basis of outcome? | N | N | N | Y | Y |
| Some other process? (specify) |
|
|
|
|
|
| Which parts of the study were prospective: |
|
|
|
|
|
| Identification of participants? | Y | Y | N | Y | N |
| Assessment of baseline and allocation to intervention? | Y | Y | N | Y | N |
| Assessment of outcomes? | Y | Y | P | Y | N |
| Generation of hypotheses? | Y | Y | Y | Y | P |
| On what variables was comparability between groups assessed: |
|
|
|
|
|
| Potential confounders? | P | P | P | P | P |
| Baseline assessment of outcome variables? | P | P | P | N | N |
| RCT = randomized clinical trial | |||||
| Note: Users need to be very clear about the way in which the terms 'group' and 'cluster' are used in these tables. The above table only refers to groups, which is used in its conventional sense to mean a number of individual participants. With the exception of allocation on the basis of outcome, 'group' can be interpreted synonymously with 'intervention group'. Although individuals are nested in clusters, a cluster does not necessarily represent a fixed collection of individuals. For instance, in cluster‐allocated studies, clusters are often studied at two or more time points (periods) with different collections of individuals contributing to the data collected at each time point. Was there a comparison? Typically, researchers compare two or more groups that receive different interventions; the groups may be studied over the same time period, or over different time periods (see below). Sometimes researchers compare outcomes in just one group but at two time points. It is also possible that researchers may have done both, i.e., studying two or more groups and measuring outcomes at more than one time point. How were participants/clusters allocated to groups? These items aim to describe how groups were formed. None will apply if the study does not compare two or more groups of participants. The information is often not reported or is difficult to find in a paper. The items provided cover the main ways in which groups may be formed. More than one option may apply to a single study, although some options are mutually exclusive (i.e., a study is either randomized or not). ‐ Randomization: Allocation was carried out on the basis of truly random sequence. Check carefully whether allocation was adequately concealed until participants were definitively recruited. ‐ Quasi‐randomization: Allocation was done on the basis of a pseudo‐random sequence, e.g., odd/even hospital number or date of birth, alternation. Note: when such methods are used, the problem is that allocation is rarely concealed. ‐ By other action of researchers: this is a catch‐all category and further details should be noted if the researchers report them. Allocation happened as the result of some decision or system applied by the researchers. For example, participants managed in particular 'units' of provision (e.g. wards, general practices) were 'chosen' to receive the intervention and participants managed in other units to receive the control intervention. ‐ Time differences: Recruitment to groups did not occur contemporaneously. For example, in a historically controlled study participants in the control group are typically recruited earlier in time than participants in the intervention group; the intervention is then introduced and participants receiving the intervention are recruited. Both groups are usually recruited in the same setting. If the design was under the control of the researchers, both this option and 'other action of researchers' must be ticked for a single study. If the design 'came about' by the introduction of a new intervention, both this option and 'treatment decisions' must be ticked for a single study. ‐ Location differences: Two or more groups in different geographic areas were compared, and the choice of which area(s) received the intervention and control interventions was not made randomly. So, both this option and 'other action of researchers' could be ticked for a single study. ‐ Treatment decisions: Intervention and control groups were formed by naturally occurring variation in treatment decisions. This option is intended to reflect treatment decisions taken mainly by the clinicians responsible; the following option is intended to reflect treatment decisions made mainly on the basis of participants' preferences. If treatment preferences are uniform for particular provider 'units', or switch over time, both this option and 'location' or 'time' differences should be ticked. ‐ Patient preferences: Intervention and control groups were formed by naturally occurring variation in patients' preferences. This option is intended to reflect treatment decisions made mainly on the basis of patients' preferences; the previous option is intended to reflect treatment decisions taken mainly by the clinicians responsible. ‐ On the basis of outcome: A group of people who experienced a particular outcome of interest were compared with a group of people who did not, i.e., a case‐control study. Note: this option should be ticked for papers that report analyses of multiple risk factors for a particular outcome in a large series of participants, i.e. in which the total study population is divided into those who experienced the outcome and those who did not. These studies are much closer to nested case‐control studies than cohort studies, even when longitudinal data are collected prospectively for consecutive patients. Which parts of the study were prospective? These items aim to describe which parts of the study were conducted prospectively. In a randomized controlled trial, all four of these items would be prospective. For non‐randomized trials (NRS) it is also possible that all four are prospective, although inadequate detail may be presented to discern this, particularly for generation of hypotheses. In some cohort studies, participants may be identified, and have been allocated to treatment retrospectively, but outcomes are ascertained prospectively. On what variables was comparability of groups assessed? These questions should identify 'before‐and‐after' studies. Baseline assessment of outcome variables is particularly useful when outcomes are measured on continuous scales, e.g., health status or quality of life. Response options Try to use only 'Yes', 'No', and 'Can't tell' response options. 'NA' should be used if a study does not report a comparison between groups. |
