Following a recent tweet from Partha Kar, and the subsequent conversation it created on Twitter, I thought a deeper investigation into this topic was worthwhile.
To investigate this further, I wanted to take accuracy as the measure of “quality” for glucose test strips, and the prescribing costs for a means of comparison.
Prescribing costs are easily found, thanks to the NHS BSA prescribing data being publicly available for the UK.
Accuracy, less so in a consolidated form, however, after some searching, I came across the following paper that investigated whether there was increased variability in Hba1C outcomes for people with T1D where test strips with a greater standard deviation were prescribed more frequently. The authors come from a range of universities and Diabetes and Endocrinology departments, including contributions from one of the National Diabetes Audit team. It was funded only from within the establishments involved.
It contains two important pieces of information.
- That prescribing test strips with greater Standard Deviation has a linear relationship with a wider variance in Hba1C for a population;
- A list of Standard Deviations for some of the most popular strips provided by NHS England & Wales.
So taking the first point, we can see that one measure of quality has to be accuracy, in the form of standard deviation. There are others, looking at ease of use of meters, ease of applying blood, etc, that I’ll not attempt to cover here. Instead, we’ll just look at accuracy, and the costs associated with it.
Test Strips: accuracy and cost
The below table documents a number of available test strips, with a column at the end, Variability, which is a reference to Standard Deviation that the researchers were able to find. We’ll use this a bit later on.
Given this data set, we can refer to the NHS BSA prescribing data and find out which of those listed appear in the most frequently prescribed strips. The data used for this was May 2019.
Of those with listed deviations, this is the table of the most prescribed:
|Strip Name||Standard Deviation (%)||Quantity|
|GlucoRx Nexus (Reagent)_Strips||9||8,860,350|
|Contour Next (Reagent)_Strips||4.6||4,481,350|
|WaveSense JAZZ (Reagent)_Strips||8.2||3,959,450|
|FreeStyle Optium (Reagent)_Strips||6.1||2,576,400|
|OneTouch Select Plus (Reagent)_Strips||7.1||1,625,600|
|FreeStyle Lite (Reagent)_Strips||7.4||1,546,850|
|OneTouch Verio (Reagent)_Strips||8.4||1,001,050|
This list accounts for 71% of the strips prescribed in England and Wales in May 2019. A further 51 brands were also prescribed. That’s not to say they are all for people with Type 1 or Type 2 on insulin, but I’m making the assumption that this is likely to be the case.
We can see that the standard deviation across this range is pretty large, stretching from 4.6% for the best (Contour Next) to the worst at 9% (GlucoRX Nexus). Sadly, the worst is also the most prescribed, so let’s look at pricing and see where this takes us.
Cost to the NHS
The following table identifies the cost to the NHS of each of the brands with an available Standard Deviation.
|Strip Name||Cost per unit (£)|
|Contour Next (Reagent)_Strips||0.30|
|FreeStyle Optium (Reagent)_Strips||0.32|
|OneTouch Select Plus (Reagent)_Strips||0.20|
|FreeStyle Lite (Reagent)_Strips||0.32|
|WaveSense JAZZ (Reagent)_Strips||0.17|
|OneTouch Verio (Reagent)_Strips||0.30|
|GlucoRx Nexus (Reagent)_Strips||0.18|
From this table, it’s immediately pretty obvious why GlucoRX Nexus test strips are the most widely prescribed. They are clearly amongst the cheapest per strip, whereas the second most heavily prescribed are the most expensive.
Given what we’ve already seen regarding the impact of inaccurate strips on clinical outcomes, the value of the strips comes in being the most accurate. When we align prices with standard deviation, we can see that pricing doesn’t always reflect this.
|Strip Name||Cost per unit (£)||Standard Deviation (%)|
|Contour Next (Reagent)_Strips||0.30||4.6|
|FreeStyle Optium (Reagent)_Strips||0.32||6.1|
|OneTouch Select Plus (Reagent)_Strips||0.20||7.1|
|FreeStyle Lite (Reagent)_Strips||0.32||7.4|
|WaveSense JAZZ (Reagent)_Strips||0.17||8.2|
|OneTouch Verio (Reagent)_Strips||0.30||8.4|
|GlucoRx Nexus (Reagent)_Strips||0.18||9|
Costing strips based on accuracy, using standard deviation
In fact, some of those with some of the worst standard deviations are amongst the most expensive. For me that does not represent value for money, as it is likely to increase the risk of long term problems due to the variability of Hba1C as a result.
As a result of this data, I wanted to provide a mechanism to establish a way of pricing the strips based on Standard Deviation, in comparison to the best.
For this, I’ve compared accuracy of the test strips to the current leader, the Contour Next. I’ve created a cost per unit of deviation for this strip, which simply does the following:
CUD = Cost/SD
I’ve then calculated the difference between each brand’s SD and the Contour Next, and multiplied that difference by CUD to establish the cost of the difference.
Once the cost of the difference is established, I’ve subtracted that from the cost per unit of the Contour Next to arrive at an “Expected Cost per unit” based on discounting the best, using the worsening variance.
|Strip Name||Standard Deviation (%)||Difference in SD||Cost of difference (£)||Expected cost per unit (£)||Actual Cost per unit (£)||Over/Under-priced (%)|
|Contour Next (Reagent)_Strips||4.6||0||0.0000||0.30||0.30||0|
|FreeStyle Optium (Reagent)_Strips||6.1||1.5||0.0989||0.20||0.32||58|
|OneTouch Select Plus (Reagent)_Strips||7.1||2.5||0.1648||0.14||0.20||44|
|FreeStyle Lite (Reagent)_Strips||7.4||2.8||0.1846||0.12||0.32||174|
|WaveSense JAZZ (Reagent)_Strips||8.2||3.6||0.2373||0.07||0.17||165|
|OneTouch Verio (Reagent)_Strips||8.4||3.8||0.2505||0.05||0.30||473|
|GlucoRx Nexus (Reagent)_Strips||9||4.4||0.2900||0.01||0.18||1258|
From this, we can see that if we were to base the price we paid for test strips on accuracy alone, all those getting the cheaper GlucoRX are paying 12.6x more than the strips are worth. Those using either the Mobile of Performa systems are getting a bargain, and pretty much everything else is overpriced based on its standard deviation.
If a CCG has moved people using insulin from Contour Next to GlucoRX, I’d argue that they’ve done themselves a disservice. There’s a fairly high likelihood that they’re not helping either themselves or their T1Ds in the long run. On this analysis, even 18p a strip seems expensive.
What does this mean for the NHS?
This is a data point looking at a single part of the equation for what makes a good glucose testing meter, but perhaps, I’d argue, the most important.
As we referenced earlier, there is a strong relationship between meters that have higher variability and variability in Hba1C results, or in other words, if you don’t have CGM, you’ll have greater variability with a less accurate meter, so it’s false economics to provide many of the cheaper meters as in the long term they will cost the system more.
The price that the NHS is paying for some test strips is not necessarily in line with the value they provide based on accuracy, and while some of the higher priced options are noticeably less accurate there is also, generally, a negative correlation between cost and accuracy in these devices.
Whilst it’s clear that some strips seem over-priced, there may be additional circumstances where these take precedence, e.g, where they are used in a bolus calculating meter. Unfortunately, there are only a few self-contained meters available in the UK that provide this functionality, and the strips for these are all priced at the high end. In fact, what stands out is that the three strip brands that work with bolus calculator meters in this list all cost the same. Is that just a coincidence?
However you look at it, I’d expect pricing and provision of testing strips to be correlated to a number of factors. Accuracy would be one of the most important ones, given the research discussed earlier, followed by other factors such as accessibility and functionality.
It seems that many CCGs and test strip manufacturers disagree with me on that point!