Blood sampled from existing peripheral IV cannulae yields results equivalent to venepuncture: a systematic review

Summary Objectives To establish whether blood samples taken from used peripheral intravenous cannulae are clinically interchangeable with venepuncture. Design Systematic review. PubMed, Web of Science and Embase were searched for relevant trials. Setting Trials which compared blood samples from used peripheral intravenous cannulae to venepuncture and provided limits of agreement or data which allowed calculation of limits of agreement. Participants Seven trials with 746 participants. Blood tests included 13 commonly ordered biochemistry, haematology and blood gas measurements. Main outcome measures 95% limits of agreement. Data were pooled using inverse variance weighting and compared to a clinically acceptable range estimated by expert opinion from previous trials. Results Limits of agreement for blood samples from used peripheral intravenous cannulae were within the clinically acceptable range for sodium, chloride, urea, creatinine and haematology samples. Limits of agreement for potassium were ±0.47 mmol/L which exceeded the clinically acceptable range. Peripheral intravenous cannula samples for blood gas analysis gave limits of agreement which far exceeded the clinically acceptable range. Conclusions Blood sampling from used peripheral intravenous cannulae is a reasonable clinical practice for haematology and biochemistry samples. Potassium samples from used peripheral intravenous cannulae can be used in situations where error up to ±0.47 mmol/L is acceptable. Peripheral intravenous cannula samples should not be used for blood gas analysis.


Introduction
Venepuncture carries a certain amount of pain and a small risk of complications. Given many patients have an in-situ peripheral intra-venous cannula, sampling blood from this obviates the need for repeated venepuncture if there are clinically equivalent and reliable results. In general, this is not common practice due to concerns regarding the validity of results. 1 Other considerations include the method used for obtaining samples, prior or concurrent use of the cannula for fluid administration, the aspiration volume needing to be discarded and any need for special equipment.
Whether assay results from peripheral intravenous cannula and venepuncture are clinically equivalent is quantified using limits of agreement, expressed as the range within which 95% of values will lie in comparison to a reference standard measurement. 2 The limits of agreement are then compared with a clinically acceptable range usually defined by consensus.
In order to assess the quantity and quality of studies comparing blood samples from existing PIV with venepuncture, we undertook a systematic review. We focused on studies of sufficient quality to influence clinical practice.

Search strategy
Relevant keywords and terms were developed through a scoping search in PubMed, eventually expanding to three databases (PubMed, ISI Web of Science and Embase). Two reviewers screened titles and abstracts (FL and DL). The abstracts were categorised into 'not relevant' and 'potentially relevant' and all 'potentially relevant' studies were reviewed in full. References of included studies were also hand searched ( Figure 1). The full search strategy for PubMed is shown in Appendix 1.

Data collection and extraction
Relevant data were extracted from included papers in duplicate (FL and DL), including publication year, patient population, number of patients, blood tests carried out, discard volume prior to sampling, wait time between stopping infusions and sampling, aspiration method and cannula gauge as well as the assay results. Tests measured in non-SI units were converted to SI units. Blood urea nitrogen results were converted to urea by multiplying by 2.14 and then converting to SI units. The 95% limits of agreement was the primary outcome of interest. If not reported directly, it was calculated Limits of agreement ¼ Mean difference AE 1:96 Â SD of differences from mean Limits of agreement values were then pooled using inverse variance weighting variance ðpooledÞ

Clinically acceptable limits of agreement
The clinically acceptable errors in blood sampling are not fully established and vary depending on patient situation and the clinicians' tolerance for error.
Four studies specified such ranges established through clinician survey. [3][4][5][6] We used a mean of these values to define clinically acceptable limits for this review.

Results
Literature search provided 1857 articles for abstract review with 130 duplicates ( Figure  1). Hand-searching identified a further six studies for abstract review. There were 21 papers which were excluded at full text review for failure to meet the inclusion criteria or meeting the exclusion criteria (Appendix 2). Ultimately, seven studies were included with total individuals n ¼ 746 from a combination of adult inpatient, adult emergency department, paediatric inpatient and healthy volunteers ( Table 1).

Characteristics of included studies
Studies comparing venepuncture to peripheral intravenous cannula used different methods and protocols ( Table 1). The minimum discard volume in the studies was 2 mL. Cannula sizes varied from 16 to 22 French. All cannulae had been used prior to sampling but there was variation in volume and contents of infusion prior to sampling. Sampling devices used were either syringe or vacutainer systems.

Biochemistry
Sodium, chloride, urea and creatinine pooled limits of agreement were all within the clinically acceptable error range (Table 2). In some cases, the pooled limits of agreement was substantially lower than the

Haematology and international normalised ratio
The pooled results for haematology and international normalised ratio were all within the clinically acceptable limits of agreement (Table 3).

Blood gases and pH
The pooled limits of agreement for pH was within clinically acceptable range (Table 4). However, the pooled limits of agreement for pCO 2 exceeded the clinically acceptable range (in kPa limits of agreement ¼ À0.9, þ1.3; clinically acceptable range ¼ AE0.7) and limits of agreement pO 2 dramatically exceeded the clinical acceptable range (in kPa limits of agreement ¼ À4.5, þ3.0; clinically acceptable range ¼ AE0.7).

Discussion
Across pooled studies, we showed that assays from used peripheral intravenous cannula were reliable and clinically consistent with fresh venepuncture samples, except in the case of potassium and blood gases. Taken together, our findings suggest that peripheral intravenous cannula sampling could be given greater consideration in clinical practice -at least for the tests described.
The results for potassium levels were not within clinically acceptable agreement limits for some patients. The 95% limits of agreement for potassium of À0.46 to þ 0.47 mmol/L shows that, for patients where a tight control of potassium is essential, samples from used peripheral intravenous cannula should be used with caution. For most other patients, a sample from a peripheral intravenous cannula would be sufficient and the level of error AE0.47 mmol/L is unlikely to affect patient outcomes. The cause for the higher level of error is not clear, though haemolysis or haemodilution was excluded as causes in the studies considered here. For blood gas analyses (pO 2 and pCO 2 ), the two studies reporting these suggested errors were due to contamination of the samples with atmospheric air post-collection. 6,7 There were some limitations to this study. There was heterogeneity in study populations, protocols and equipment. It is not clear whether our findings are generalisable to other sampling techniques, e.g. with narrower gauge cannula.
In terms of haemolysis degrading the sample quality, we show that for most blood tests it does not lead to significant errors. However, if the laboratory or analysers do not check for haemolysed samples it could lead to errors in results. 8 We did not assess some commonly ordered blood tests. Our findings relate to specific assays and may not be generalisable to other haematology and biochemistry investigations.
The clinical impact of these findings will be greatest in those situations in which patients require repeated blood tests where samples from peripheral intravenous cannula would be suitable. For example, if a patient were admitted with symptomatic anaemia and needed serial haemoglobin measurements samples from a cannula could be used. Peripheral intravenous cannula sampling can be an alternative for patients who find venepuncture intensely distressing. There are also patients in which venepuncture is technically difficult and peripheral intravenous cannula samples can provide easier access to blood.
Our findings do not explain why some blood tests are not reliable when taken from a used peripheral intravenous cannula and this could be the subject of further research. Further studies could also be considered to assess other assays which were not included in this paper.

Conclusions
Peripheral intravenous cannula samples are interchangeable with venepuncture for sodium, chloride, urea, creatinine and haematology tests. Peripheral