Description of the condition
Stroke is the second most common cause of death and the leading cause of disability in developed countries. It consumes about 5% of health services resources, and much of the cost is attributable to the care of disabled inpatients (Liu 2007). The acute phase of stroke is crucial for long-term outcome; indeed, the most common complications arise mainly in the first 24 hours after stroke onset and rarely after four days (Indredavik 2008). The approximate frequency of complications of different types is: neurological (6%), infective (21%), pressure sores (21%), deep venous thrombosis (1%), pulmonary embolism (2%), pain (22%) and psychological (11%) (Indredavik 2004). These complications, along with cardiac complications and mortality, may be related to the quality of care provided.
Description of the intervention
The intervention is the non-invasive intensive or continuous monitoring of vital functions such as blood pressure, pulse rate, oxygen saturation, temperature, and heart rhythm assessed by electrocardiogram (ECG). Intensive monitoring should be useful because acute medical interventions targeted at maintaining these essential physiological variables within a narrow physiological range should improve the outcome (Langhorne 2000).
However, continuous monitoring could both increase the rate of unnecessary medical interventions and, by limiting patients' mobility, favour bed-related complications such as infections, hypostatic pneumonia, deep venous thrombosis, thromboembolism and pain (Adams 2007). Moreover, intensive continuous monitoring for several days may lead to a psychological dependence on the monitoring equipment and delay mobilisation, even if the devices are removed during physiotherapy.
How the intervention might work
The use of continuous monitoring should favour prompt medical intervention aimed at obtaining an intensive control of physiological homeostasis and the early detection of cardiac arrhythmias. The importance of monitoring people with stroke is supported indirectly by the evidence that fever (Castillo 1998; Reith 1996), hypertension, hypotension (Leonardi-Bee 2002), hypoxia and cardiac arrhythmias (Oppenheimer 1992) play an important role in determining the outcome of stroke survivors. An elevated body temperature can increase metabolic demand in the ischaemic penumbra and enhance the release of cytotoxic excitatory amino acids and the formation of free radicals (Hajat 2000); mild to moderate hypotension may result in a critical fall in cerebral blood flow, worsening the degree of ischaemia in the penumbra and transforming areas of reversible injury into infarct (Powers 1992); hypoxia further depletes the energy stores and can influence ischaemic progression (Sulter 2000). People with acute stroke are at risk of developing ventricular arrhythmias (Oppenheimer 1992) and circulatory collapse, and congestive heart failure or cardiac arrest may occur in conditions such as myocardial infarction or atrial fibrillation. Studies conducted to evaluate the role of monitoring in the management of people with acute stroke showed a better outcome at both discharge and one-year follow up in people monitored continuously for at least 48 hours (Cavallini 2003; Silva 2005; Sulter 2003).
Why it is important to do this review
Stroke units significantly reduce mortality, institutionalisation and dependence of people with stroke (Stroke Unit Trialists' Collaboration 2007). However, the reasons for their effectiveness remain undetermined (Indredavik 1999; Rønning 1998). Despite the presence of standardised protocols for acute evaluation and medical treatment, strategies for early mobilisation and a strong focus on rehabilitation, multidisciplinary teams and integration between medical care, nursing and rehabilitation (Langhorne 2002), there is uncertainty about the most effective stroke unit model. One of the most controversial issues is the intensity of monitoring, which has a strong impact on stroke unit organisation in terms of the number of personnel, specialisation, structure and costs.
It is, therefore, important to gather all available evidence to determine whether continuous monitoring improves outcome. The questions are (1) how, when, and in which people should continuous monitoring start; and (2) for how long should it continue (Gladstone 2008)?
To assess whether continuous intensive monitoring compared with intermittent monitoring of physiological variables in people with acute stroke can change their prognosis in terms of mortality or disability.
Criteria for considering studies for this review
Types of studies
Our aim was to include all published or unpublished, unconfounded, randomised, cluster randomised or quasi-randomised controlled trials with or without blinding.
Types of participants
People who had an ischaemic or haemorrhagic stroke, according to the World Health Organization criteria (Hatano 1976), irrespective of age, gender, or social status whose symptom onset was less than three days from stroke onset and who were followed up at least until discharge. Confirmation of clinical diagnosis using imaging was not compulsory.
Types of interventions
Continuous monitoring, defined as non-invasive continuous (that is covering at least 12 hours) monitoring, within the first 72 hours of admission of at least one of the following variables: blood pressure, pulse rate, respiration rate, oxygenation, heart rhythm, or body temperature. We considered monitoring to be continuous even if it was discontinued for a period of less than one hour (so that the patient could be toileted, bathed or mobilised).
Non-continuous monitoring or intermittent physiological monitoring, defined as monitoring for less than 12 hours within the first 72 hours of admission, of at least one of the above-mentioned variables.
In order to ensure a certain degree of reliability of the definitions of continuous and intermittent monitoring, we decided not to be very restrictive in the definitions for both as regards the timing of the continuous monitoring other than the type of monitored variables.
Since there is the potential for great heterogeneity in the definitions, both of the threshold levels of abnormality that triggered an intervention for each of the above-mentioned physiological variables, and of the specific intervention that was given to correct the abnormality, we decided to include details on these aspects from each study.
Types of outcome measures
We studied the following outcome measures in both treatment groups.
- Death and disability (modified Rankin Scale (mRS) ≥ 3 or equivalent) at three months or longer, or at discharge if follow up was no longer available
- Death from any cause at discharge
- Disability, preferably at three months or longer
- Death from vascular causes
- Quality of life
- Length of stay
- Neurological complications (stroke progression, stroke recurrence, seizures)
- Cardiac complications (arrhythmias, heart failure, myocardial infarction)
- Complications related to immobility (infections, deep vein thrombosis, pulmonary embolism, falls, pressure sores, pain) or to treatments triggered by the monitoring of listed variables.
The number of people who triggered an intervention for at least one of the following conditions:
- oxygen desaturation;
- atrial fibrillation;
We described the type of intervention and threshold level of abnormality considered in each study.
Search methods for identification of studies
See the 'Specialized register' section in the Cochrane Stroke Group module. We searched for trials in all languages and we planned to arrange the translation of relevant trial reports published in languages other than English.
We searched the Cochrane Stroke Group Trials Register (last searched November 2012), the Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library 2011, Issue 8), MEDLINE (1966 to November 2012) (Appendix 1), EMBASE (1980 to November 2012), CINAHL (Cumulative Index of Nursing and Allied Health Literature) (1982 to November 2012) and the British Nursing Index (BNI) (1985 to November 2012).
Searching other resources
In an effort to identify further published, unpublished and ongoing studies, we:
- searched the following clinical trials and research registers (November 2012): Stroke Trials Registry (www.strokecenter.org/trials/), Current Controlled Trials (www.controlled-trials.com), ClinicalTrials.gov (www.clinicaltrials.gov/), Chinese Clinical Trial Registry (http://www.chictr.org) (November 2012);
- searched the reference lists of all relevant articles obtained and previously published articles on the subject;
- contacted trial authors, colleagues and researchers in the field of study;
- handsearched the following conference proceedings not already handsearched on behalf of The Cochrane Collaboration: proceedings of the annual meetings of the Italian Neurological Society and the Annual Meetings of Neurology, Neurosurgery and Psychiatry since 1990 to 2012.
Data collection and analysis
Selection of studies
One review author (MGC) scrutinised titles and abstracts of records identified by the electronic searches and excluded obviously irrelevant studies. We obtained the full text of the remaining studies and three authors (MGC, AC and RC) selected studies for inclusion based on the review selection criteria. The review authors resolved any disagreements by discussion until a consensus was reached.
Data extraction and management
All the review authors independently extracted data from the selected studies using an ad hoc standard data extraction form specifically designed for this review and piloted on two trials (Appendix 2). We attempted to obtain any missing data by contacting trial authors. Where possible, we documented:
- participant details;
- inclusion and exclusion criteria;
- duration and frequency of intervention;
- drop-outs and losses to follow up;
Where possible, we extracted data on the threshold level of abnormality that triggered intervention for a given continuous physiological variable, the specific intervention given to correct the abnormality, and compliance with allocated therapy.
Assessment of risk of bias in included studies
All the review authors independently evaluated the risk of bias in included studies, according to Chapter 8 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011), by assessing the following items: (1) generation of randomisation sequences (selection bias), (2) concealment of allocation (selection bias), (3) blinding of participants and personnel (performance bias), (4) blinding of outcome assessment (detection bias), (5) incomplete outcome data (attrition bias), (6) selective reporting (reporting bias), and (7) other bias. The authors discussed discrepancies and disagreements until they reached a consensus.
Measures of treatment effect
We based the efficacy analyses on the results of the individual trials for death and disability at the end of the scheduled follow up. We have used the analyses on secondary outcomes and on less severe adverse events to support the data on the primary outcomes. We used Review Manager 5.1 (RevMan 2011) for all analyses. We used an intention-to-treat analysis where possible. We calculated the weighted estimate of the treatment effects across trials (odds ratio) using a random-effects model. For interpreting the results, we used 95% confidence intervals (CI).
Unit of analysis issues
In each study, we considered whether:
- groups of individuals were randomised together to the same intervention (i.e. cluster randomised trials);
- individuals underwent more than one intervention (e.g. in a cross-over trial, or simultaneous treatment of multiple sites on each individual); or
- there were multiple observations for the same outcome (e.g. repeated measurements, recurring events, measurements on different body parts).
Dealing with missing data
If information about excluded participants or participants lost to follow up after randomisation was unavailable from the publications, we decided to seek further information through correspondence with the trialists. Where the data about these participants were unavailable, we performed a sensitivity analysis.
Assessment of heterogeneity
We quantified inconsistency across studies using the I
Assessment of reporting biases
We used funnel plots to assess reporting bias. We assessed funnel plots qualitatively.
We calculated a weighted estimate of the treatment effects across trials (odds ratio) using a random-effects model. For interpreting the results we used 95% CI.
Subgroup analysis and investigation of heterogeneity
If we found substantial heterogeneity in the efficacy analysis, and the number of studies permitted it, we explored heterogeneity with the following subgroup analyses:
- comparison between trials with low and high risk of bias;
- comparison of the efficacy of treatment between participants followed up until discharge and those followed up for three months or longer;
- comparison of the effect of different variables monitored, and length of monitoring;
- comparison of the efficacy of treatment of transient ischaemic attack with ischaemic or haemorrhagic stroke.
We considered the overlap of the CIs of the summary estimates in the groups of comparison (non-overlap of CIs indicated statistical significance).
If information about participants excluded or lost to follow up after randomisation remained unavailable, we planned to provide a worst-case scenario analysis for the outcome of disability and death from any cause to ensure significance of the results. In this analysis, we assumed that those participants who were lost to follow up in the treatment group had the worst outcome while those participants who were lost to follow up in the control group had the best outcome. If the effects of primary and worst-case meta-analyses were in the same direction and of the same magnitude, we made a definite conclusion about the treatment effect; otherwise we drew no definite conclusion.
Description of studies
Results of the search
The search of the Cochrane Stroke Group Trials Register identified 220 references, MEDLINE 2577, EMBASE 3910, CINAHL 506 and BNI 497. After eliminating duplicates and non-relevant studies from the titles, we selected 66 possibly relevant studies and 18 reviews. We screened cited references and cross-checked electronically retrieved references and identified no further studies. After examining abstracts, or in some cases full papers, we excluded 55 of the 66 studies: 50 because the delivery of care was confounded in different settings, one because monitoring was carried out only during the embolization procedure, and four because they were not controlled trials. This process left 11 potentially relevant studies (Cavallini 2003; Davis 2000; Durastanti 2005; Higgins 2012; Langhorne 2000; Pappa 2009; Silva 2005; Smith 2009; Sulter 2003; VERITAS 2007; Weber 2007). We selected eight after viewing the full papers. A PRISMA flowchart of studies selection is shown in Figure 1.
|Figure 1. Search flow diagram.|
We did not find any cluster randomised trials.
We included three trials, involving a total of 354 participants, in this review: Cavallini 2003, which represented 77% of the overall sample; Sulter 2003; and VERITAS 2007. Two were randomised but one (Cavallini 2003) was quasi-randomised, as consecutively admitted people were allocated to the continuous or intermittent monitoring group on the basis of availability of beds. The three studies were open label and enrolled participants within 24 (Sulter 2003) or 36 hours (Cavallini 2003; VERITAS 2007) from stroke onset. In all the trials considered, blood pressure (BP), electrocardiography (ECG), oxygen saturation (OS) and body temperature (BT) were monitored. Cavallini 2003 also monitored respiratory frequency (RF) and electroencephalography (EEG); Sulter 2003 and VERITAS 2007 measured blood glucose. The duration of continuous monitoring lasted at least 72 hours in Cavallini 2003 and VERITAS 2007, and 48 hours or longer in Sulter 2003. For participants allocated to the intermittent monitoring group, the majority of physiological variables were monitored every four hours in Cavallini 2003 and VERITAS 2007 and four times a day in Sulter 2003. All the trials set thresholds for when the measures of physiological variables were considered abnormal (see Characteristics of included studies for details) but VERITAS 2007 did not specify the type of medical intervention triggered. Primary outcomes measured were death and disability, which were assessed at discharge by Cavallini 2003 (modified Rankin Scale (mRS) score 4 to 6) and at three months by Sulter 2003 (mRS score 4 to 6 or Bathel Index (BI) < 60 or institutional care and mortality at three months) and VERITAS 2007 (mRS score 3 to 6).
We excluded three studies after reading the full published papers: two controlled studies were not randomised (Langhorne 2000; Silva 2005), and in a third study monitoring was tested in the particular situation of emergency flight rescue at less than 12 hours (Weber 2007). We contacted the authors of three other studies for further information where data were available only from conference proceeding abstracts. Two of the studies were non-random (Durastanti 2005; Pappa 2009) and we obtained no information for the third study (Davis 2000) as we did not receive a reply when the authors were contacted by email and telephone. As insufficient data were available to allow this study to be included, we classified it as 'awaiting classification' in Characteristics of studies awaiting classification.
Risk of bias in included studies
Overall, we considered Sulter 2003 and VERITAS 2007 to be at low risk of bias because the generation of allocation sequence, outcome data and the blinding of outcome assessment were adequate. We scored Cavallini 2003 at high risk of bias because it was not truly randomised and was devoid of adequate allocation concealment. Figure 2 and Figure 3 show a summary of the risk of bias in all included studies. A risk of bias table for each study is provided in the Characteristics of included studies section.
|Figure 2. Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.|
|Figure 3. Risk of bias summary: review authors' judgements about each risk of bias item for each included study.|
All the studies included were controlled and randomised, although participants in Cavallini 2003 were consecutively admitted to the Cerebrovascular Department and were allocated to the conventional stroke unit or continuous monitoring purely on the basis of availability of beds. The stroke unit beds were always filled first, where the allocation of participants was the responsibility of the ward administrator who was blind to the aims of the study. The allocation concealment was provided by an envelope system in Sulter 2003, although it was unclear whether the envelopes were opaque and sealed, and in VERITAS 2007 by telephoning a secretary in an independent office who logged the participant and opened the next in a series of sequentially numbered, opaque sealed envelopes.
There was no blinding of participants and personnel of the stroke units involved in the studies but we classified them at low risk of performance bias because we judged that the outcome was not likely to be influenced by lack of blinding. Blinding of outcome assessment was guaranteed in two studies (Sulter 2003; VERITAS 2007), while in Cavallini 2003 it was not specified whether outcome assessors were blind to allocation of participants.
Incomplete outcome data
In the three studies, all the randomised participants were accounted for in the analysis.
We obtained data on length of stay, institutionalisation, and poor outcome from the authors for Cavallini 2003 and VERITAS 2007, and we contacted the principal investigator of VERITAS 2007 several times for information about telemetry, death from vascular cause, dependency, infections, deep vein thrombosis, fever, hypoxia, hypotension and hypertension.
Different cut-offs for the mRS were used in the included studies (≥ 4 in Cavallini 2003 and Sulter 2003; ≥ 3 in VERITAS 2007) and different timing of assessment (at discharge in Cavallini 2003). See Characteristics of included studies for further details. A priori, we considered mRS ≥ 3 as the primary outcome. However, data were not available in Cavallini 2003 and Sulter 2003. We tried to contact the study authors for this information but the number of participants with mRS ≥ 3 was not available for Cavallini 2003 and we could not make contact with the authors of Sulter 2003.
There was little or no information about quality of life, delirium, depression, falls or infections due to bladder catheterization.
Reports of the studies were free from suggestions concerning selective outcome reporting.
Other potential sources of bias
There are many potential sources of bias and heterogeneity: the types and intensity of medical and rehabilitative interventions were not detailed in the studies and could differ, the frequency of variables assessed with monitoring was variable and the frequency of intermittent detection varied from four hours (Cavallini 2003, VERITAS 2007) to six hours (Sulter 2003). Some outcomes, such as institutionalisation, as well as the availability of home care for disabled persons can vary across countries.
Effects of interventions
Continuous monitoring was compared with intermittent monitoring on primary outcomes, secondary outcomes, adverse effects and surrogate outcomes.
We extracted data on death and disability (that is mRS ≥ 3 in VERITAS 2007, while only data on mRS ≥ 4 were available for Sulter 2003 and Cavallini 2003 ) at three months (Sulter 2003; VERITAS 2007) or at discharge (Cavallini 2003).
- Death and disability at three months, or at discharge if follow up was no longer available: continuous monitoring significantly reduced the composite endpoint of death and disability at three months or discharge (OR 0.27, 95% CI 0.13 to 0.56) ( Analysis 1.1). Cavallini 2003 contributed most of the weight to the results due to its sample size. If Cavallini 2003 was removed from the analysis, death and disability reduction would no longer be statistically significant (OR 0.32, 95% CI from 0.06 to 1.63). However, there was no significant inconsistency across the three studies on this outcome (I
2= 37%, 95% CI 80% to 0%).
- Death from any cause at discharge: continuous monitoring was associated with a non-significant reduction in death from any cause at discharge (OR 0.72, 95% CI 0.28 to 1.85) ( Analysis 1.2).
We extracted data on all the scheduled variables to assess secondary outcomes from the studies with the exception of: (1) dependency at three months or longer, for which no data were available from Cavallini 2003 as this study did not provide a follow up after discharge; and (2) quality of life, for which no data were available for any of the three studies.
- Dependency, preferably at three months or longer: continuous monitoring did not significantly reduce this outcome (OR 0.79, 95% CI 0.30 to 2.06) ( Analysis 1.3).
- Death from vascular causes: continuous monitoring was associated with a non-significant reduction in death from vascular causes (OR 0.48, 95% CI 0.10 to 2.39) ( Analysis 1.4).
- Quality of life: no data were available.
- Length of stay: continuous monitoring was associated with a non-significant reduction in the number of days of hospital stay (mean difference -5.24 days, 95% CI -10.51 to 0.03). There was substantial heterogeneity across trials for this outcome (I
2= 83%, 95% CI 94% to 49%; P = 0.003) ( Analysis 1.5) and if the VERITAS 2007 study was removed from the analysis the reduction in hospital stay with continuous monitoring was statistically significant (mean difference (MD) -8.15 days, 95% CI -9.85 to -6.44) without significant inconsistency (I 2= 0%, P = 0.54).
- Institutionalisation: continuous monitoring was associated with a non-significant reduction in institutionalisation (OR 0.83, 95% CI 0.04 to 15.72) ( Analysis 1.6). In this case there was wide variability in the effect estimates between the two trials considered, due to substantial heterogeneity (I
2= 90%, 95% CI 97% to 61%).
- Neurological complications (stroke progression, stroke recurrence, seizures): continuous monitoring did not significantly reduce this outcome (OR 0.81, 95% CI 0.46 to 1.43) ( Analysis 1.7).
- Cardiac complications (arrhythmias, heart failure, myocardial infarction): continuous monitoring was associated with a significant increase in the detection of cardiac complications (OR 8.65, 95% CI 2.52 to 29.66) ( Analysis 1.8).
Adverse effects (complications related to immobility)
For complications related to immobility, on the basis of the availability of data (VERITAS 2007 in the first five days, Sulter 2003 in the first 48 hours, Cavallini 2003 during hospitalisation), we separately extracted the number of participants with pulmonary infections, including aspiration related. Other infections were grouped together as they could not be considered separately; data were available for urinary infections only in Sulter 2003 and as "other infections" in Cavallini 2003. Information regarding deep vein thrombosis was available only from Cavallini 2003 and VERITAS 2007. No data were available for pulmonary embolism, falls, pressure sores and pain in any of the included studies.
We were able to separately extract the number of participants with hypotension (that is in Sulter 2003 mean blood pressure (BP) ≤ 80 mmHg, Cavallini 2003 systolic BP < 80 mmHg, VERITAS 2007 systolic BP < 110 mmHg), hypertension (that is Sulter 2003 systolic BP > 220 mmHg or mean BP > 130, or both; Cavallini 2003 systolic > 200 mmHg or diastolic > 105 mmHg, or both; VERITAS 2007 systolic > 220 mmHg), and hypoxia (that is Cavallini 2003 oxygen saturation (OS) < 91%, Sulter 2003 OS < 95%, VERITAS 2007 OS < 95%). For fever, data were available from Sulter 2003 as body temperature > 37.5 °C with a rectal thermometer, Cavallini 2003 as body temperature > 37.8 °C, and VERITAS 2007 as body temperature (no further specification) > 37.5 °C.
Only Sulter 2003 reported the number of participants with atrial fibrillation, while Cavallini 2003 reported the number of participants with adverse cardiac events (new or onset of worsening of arrhythmias, worsening of pre-existing cardiac disease, new changes on the ECG, heart failure), and there were no major cardiac complications in VERITAS 2007 (personal communication). As we could not consider arrhythmias separately from hearth failure and other cardiac diseases in Cavallini 2003, we resolved to consider these data together under the secondary outcome of cardiac complications.
Detected conditions that triggered an intervention were as follows.
- Fever: continuous monitoring was associated with a significant increase in the detection of fever (OR 2.17, 95% CI 1.27 to 3.70) ( Analysis 1.12).
- Oxygen desaturation: continuous monitoring was associated with a non-significant increase in the detection of hypoxia (OR 2.25, 95% CI 0.98 to 5.18) ( Analysis 1.13).
- Atrial fibrillation: atrial fibrillation and other cardiac complications were detected significantly more frequently in the continuous monitoring group (see 'Secondary outcomes') ( Analysis 1.8).
We found no substantial inconsistency across studies for any of the surrogate outcomes assessed.
Summary of main results
- Continuous monitoring significantly reduced death and disability at three months or at discharge and was associated with a non-significant reduction of death from any cause at discharge (primary outcomes), but these results depended on one study at high risk of bias.
- With continuous compared with intermittent monitoring we found a non-significant reduction of dependency, death from vascular causes, neurological complications, length of hospital stay and institutionalisation (secondary outcomes). For the last two outcomes we detected a consistent heterogeneity across trials.
- Cardiac complications, fever and hypotension were significantly more often detected in participants who had continuous monitoring (surrogate outcomes).
- We did not detect any significant increase of adverse events due to immobility in participants monitored continuously compared with those monitored intermittently.
Overall completeness and applicability of evidence
The review indicated low-level evidence for the positive effects of continuous monitoring during early stroke treatment to reduce the number of dead or disabled people at three months or discharge from hospital. The evidence was low because the trial which contributed most to the primary outcome (Cavallini 2003) was not truly randomised as participants were allocated to a conventional stroke unit or to a stroke unit with continuous monitoring purely on the basis of bed availability, there was no long-term follow up and we are not certain that the assessment of outcomes was blinded. If this study is removed from the meta-analysis the result is no longer statistically significant (OR 0.32, 95% CI 0.06 to 1.63), with consistent heterogeneity between the two remaining studies (I
The long-term reduction in death or disability without a significant reduction of early deaths deserves further discussion. There are several possible explanations for this apparent contradiction: (1) the results could be affected by the small sample size; (2) continuous monitoring has a greater effect on functional outcomes (but this is contradicted by the absence of any reduction in long-term disability alone); (3) early detection of abnormalities of the physiological variables with continuous monitoring also prevents long-term case fatality but the phenomenon is not evident at discharge because events in this phase are still few (as is the case for paroxysmal atrial fibrillation, with a recurrence in the long-term in the absence of prophylactic anticoagulant therapy). If the latter explanation were true, continuous monitoring in the early phase of stroke should be an important time to detect previously unknown abnormalities, which could prevent a devastating stroke in the future. The non-statistically significant reduction in length of stay and institutionalisation associated with continuous monitoring supports the hypothesis that continuous monitoring prevents both early complications that can prolong hospital stay and the rate of stroke progression that can cause institutionalisation. The VERITAS 2007 results, compared with the two other trials, were different for these two latter outcomes and other outcomes of the study. The reason for such heterogeneity may be due to the small size of this study, the imbalance of prognostic factors at baseline and the different therapeutic approach when abnormalities of physiological variables were detected. Prompt multidisciplinary interventions to manage abnormal physiological variables in people with acute stroke may change their prognosis (Middleton 2011), and a difference in timing and modalities of therapeutic protocols might be an important issue to explain the heterogeneity between the studies.
The long-term reduction of death or disability, length of stay and institutionalisation observed in participants allocated to continuous monitoring is consistent with the greater chance of prompt detection of anomalies in physiological parameters triggering medical interventions, although just a few (cardiac complications, fever and hypotension) of many abnormalities of the physiological variables considered were significantly more often detected with continuous monitoring. Detection of abnormalities and the interventions were slightly heterogeneous, as were the clinical and instrumental endpoints used to measure them. Many factors could have contributed to the heterogeneity: the small number of samples, the differences in threshold considered, the frequency of assessment and the type and intensity of the interventions.
Continuous monitoring was not associated with significant adverse effects due to people's immobility compared with those who were monitored intermittently, although the monitoring was assessed for a short period, ranging from 24 to 72 hours, and many other conditions such as pressure sores, falls, and pain were not assessed. Moreover, in the VERITAS 2007 trial telemetry allowed the participants who were monitored continuously to move around, similar to those participants allocated to intermittent clinical monitoring, thus avoiding complications related to immobility.
The high standard of care with frequent assessment of physiological variables, also available in the non-continuous monitoring group, might limit the effect size of continuous monitoring. This may not correspond to clinical practice and, for this reason, could limit the application of the results.
There is no evidence for the type of person who needs continuous monitoring. A person's selection is an unavoidable choice to be made in the context of few resources and on the basis of our findings, preference for continuous monitoring should be given to people at higher risk of cardiac complications.
Quality of the evidence
The quality of the evidence was low. There was slight heterogeneity of the studies, a small number of studies, small samples sizes and a high risk of bias for the trial that contributed most in terms of the number of participants enrolled.
Potential biases in the review process
We aimed to include all relevant publications in this review by using multiple overlapping searches of a number of databases, contacting trial authors, colleagues and researchers in the field of study, and handsearching conference proceedings. However, the possibility of missing small trials and trials published in journals with a lower impact cannot be totally excluded. By performing funnel plots (Figure 4 and Figure 5) we aimed to analyse this influence but since the total number of included studies and the number of included participants were small, we could not reach definite conclusions.
|Figure 4. Funnel plot of comparison: 1 Continuous monitoring versus intermittent monitoring of physiological variables, outcome: 1.1 Death or dependency by the end of scheduled follow up.|
|Figure 5. Funnel plot of comparison: 1 Continuous monitoring versus intermittent monitoring of physiological variables, outcome: 1.2 Death at discharge.|
In order to get all the information for our prespecified analyses we contacted the authors with requests for additional information. We did not obtain additional information from the authors of one study (Davis 2000) that we excluded from the analyses because insufficient data were available. This large unpublished study, with 98 participants enrolled, could have an influence on our conclusions. For this reason, we have devoted a lot of effort to getting the data from this study and hope that it will become available for inclusion in a future version of this review.
Agreements and disagreements with other studies or reviews
Overall, other reviews have shown at least a marginal benefit of continuous monitoring in stroke units, although the studies and outcomes considered were slightly different. Langhorne 2009 (only an abstract from a conference proceeding was available) considered four randomised controlled trials (RCTs) involving 555 participants: the conclusion was that continuous automated physiological monitoring resulted in an increased detection and treatment of physiological complications and in a marginally lower rate of stroke progression. However, it did not significantly influence death or disability. Govan 2007 (only an abstract was available) considered three RCTs and one controlled non-randomised trial including a total of 785 participants: the conclusion was that routine monitoring might reduce the risk of stroke progression and length of stay. Foley 2007 examined three different forms of inpatient stroke care based on timing and duration of treatment. Five RCTs were included in the group 'acute stroke unit care' participants admitted within 36 hours of stroke onset and the results were similar to ours with continuous monitoring associated with a significant reduction of death and disability, and in length of hospital stay.
Implications for practice
Continuous monitoring of physiological variables in the first two or three days in a stroke unit may improve the outcome and prevent complications. Continuous monitoring improves attention to the change in physiological variables, which is one of the key features of the efficacy of stroke units.
Implications for research
Many questions remain unanswered and deserve further research, such as when to start continuous monitoring, when to interrupt it, which people should be given priority, which treatments are most appropriate, etc. To further strengthen the present weak evidence, well-designed high-quality studies and a core set of sensible outcome measures are needed. Research protocols should be prepared following multiprofessional and multidisciplinary discussions that consider the patients' views and opinions, and the costs of the disease.
The authors would like to thank Chiara Bassi, Brenda Thomas and Hazel Fraser for their help in developing the search strategy, Cinzia Del Giovane for help in calculating the I
Preliminary results of this review were presented in the workshop: 'Monitoring Stroke Units or Standard Stroke Units? Two different models in stroke care: does methodology matter?', organised by the European Association of Young Neurologists & Trainees and the Cochrane Neurological Field, and in a similar workshop organized with the Italian Society of Neurosciences (SNO). We would like to thank the participants of the two workshops for their insightful comments on the review.
Data and analyses
- Top of page
- Authors' conclusions
- Data and analyses
- Contributions of authors
- Declarations of interest
- Sources of support
- Differences between protocol and review
- Index terms
Appendix 1. MEDLINE search strategy
We used the following search strategy for MEDLINE (Ovid) and adapted it for the other databases.
1. cerebrovascular disorders/ or exp basal ganglia cerebrovascular disease/ or exp brain ischemia/ or exp carotid artery diseases/ or exp cerebrovascular trauma/ or exp intracranial arterial diseases/ or exp intracranial arteriovenous malformations/ or exp "intracranial embolism and thrombosis"/ or exp intracranial hemorrhages/ or stroke/ or exp brain infarction/ or vasospasm, intracranial/ or vertebral artery dissection/
2. (stroke$ or poststroke$ or cva$ or cerebrovascular$ or cerebral vascular).tw.
3. ((cerebral or cerebellar or brain$ or vertebrobasilar) adj5 (infarct$ or isch?emi$ or thrombo$ or apoplexy or emboli$)).tw.
4. ((cerebral or intracerebral or intracranial or brain or cerebellar or subarachnoid) adj5 (haemorrhage or hemorrhage or haematoma or hematoma or bleeding or aneurysm)).tw.
5. ((transi$ adj3 isch?em$ adj3 attack$) or TIA$1).tw.
6. 1 or 2 or 3 or 4 or 5
7. monitoring, physiologic/ or exp monitoring, ambulatory/
8. ((physiologic$ or continuous or noninvasive or non-invasive or patient$) adj5 monitor$).tw.
9. ((blood pressure or BP or pulse or heart or cardiac or oxygen$ or temperature or ECG or respirat$ or vital function$ or vital sign$) adj10 monitor$).tw.
10. 7 or 8 or 9
11. Randomized Controlled Trials as Topic/
12. random allocation/
13. Controlled Clinical Trials as Topic/
14. control groups/
15. clinical trials as topic/ or clinical trials, phase i as topic/ or clinical trials, phase ii as topic/ or clinical trials, phase iii as topic/ or clinical trials, phase iv as topic/
16. double-blind method/
17. single-blind method/
18. Multicenter Studies as Topic/
19. Research Design/
20. Program Evaluation/
21. evaluation studies as topic/
22. randomised controlled trial.pt.
23. controlled clinical trial.pt.
24. (clinical trial or clinical trial phase i or clinical trial phase ii or clinical trial phase iii or clinical trial phase iv).pt.
25. multicenter study.pt.
26. (evaluation studies or comparative study).pt.
28. (controlled adj5 (trial$ or stud$)).tw.
29. (clinical$ adj5 trial$).tw.
30. ((control or treatment or experiment$ or intervention) adj5 (group$ or subject$ or patient$)).tw.
31. (quasi-random$ or quasi random$ or pseudo-random$ or pseudo random$).tw.
32. ((multicenter or multicentre or therapeutic) adj5 (trial$ or stud$)).tw.
33. ((control or experiment$ or conservative) adj5 (treatment or therapy or procedure or manage$)).tw.
34. ((singl$ or doubl$ or tripl$ or trebl$) adj5 (blind$ or mask$)).tw.
35. (coin adj5 (flip or flipped or toss$)).tw.
37. (assign$ or allocat$).tw.
40. 6 and 10 and 39
41. limit 40 to humans
Appendix 2. Data Collection Form
Contributions of authors
Dr Alfonso Ciccone, Dr Maria Grazia Celani, Registered Nurse Raimondo Chiaramonte, Dr Cristiana Rossi and Dr Enrico Righetti conceived the review, wrote the protocol and worked on the final review. All the authors approved the final version.
Declarations of interest
In the past 36 months Alfonso Ciccone had financial relationships with the following entities: PIERREL RESEARCH ITALY SPA (lectures); CONCENTRIC MEDICAL INC (activity in the Clinical Events Committee of the TREVO (Thrombectomy REvascularization of large Vessel Occlusion in acute ischemic stroke) study; Sanofi Aventis (activity in the Clinical Adjudication Committee of the CRESCENDO study); Istituto Italiano di Ricerche Cliniche ed Epidemiologiche (activity in the Scientific Committee). Neither he nor his institution at any time received payment or services from a third party for any aspect of the submitted work.
Sources of support
- No sources of support supplied
- Cochrane Neurological Field, Italy.The Cochrane Neurological Field provided financial support to allow Authors to meet and discuss about review in Zagabria, Perugia and Rome.Authors were helped with English by Ms Kathrin Mahan from the Cochrane Neurological Field.
Differences between protocol and review
The inclusion criteria have changed between the protocol and the review in that quasi-randomised trials and cluster randomised trials are now included. This criterion was changed after reading the Cavallini 2003 paper. We discussed this paper in depth as the method of allocation was not strictly random since consecutively admitted participants were allocated to continuous or intermittent monitoring purely on the basis of bed availability. The major issue of the discussion was that only two RCTs were identified, with a total of 86 participants, while in Cavallini 2003 three times more participants were included. Therefore, although at higher risk of bias, Cavallini 2003 would be much more informative for the sample considered. We also considered the difficulties of doing a truly randomised trial in this context due to organisational problems related to the availability of beds equipped with automatic monitors in an emergency. In the end, with the choice of being restrictive with little information at low risk of bias and more tolerant of any risk of bias with more information, we chose the latter option. In order to include as much information as possible, we also decided to include cluster randomised trials.
We changed the primary outcome of 'dependence, preferably at three months or longer' to 'death and disability at three months or longer or at discharge if follow up is no longer available', and we changed 'disability' to a secondary outcome. We still considered 'death from any cause at discharge' as a primary outcome. We considered 'death and disability' more appropriate than 'disability' alone because the number of disabled people depends on survival, and because there is a policy of considering the outcomes of death and disability cumulatively as a composite poor outcome in trials assessing the efficacy of interventions for acute stroke. In accordance with many trials of acute stroke, we used the modified Rankin Scale ≥ 3 or equivalent to identify disabled people.
Medical Subject Headings (MeSH)
Blood Pressure [physiology]; Body Temperature [physiology]; Heart Rate [physiology]; Hospital Units [standards]; Institutionalization; Length of Stay; Monitoring, Physiologic [methods]; Oxygen [metabolism]; Pulse; Quality of Health Care; Randomized Controlled Trials as Topic; Stroke [complications; mortality; *physiopathology]; Time Factors
MeSH check words
* Indicates the major publication for the study