The Myth of Carbohydrate Counting

The myth is simply this:

“If you can accurately count the grams of carbohydrates in your meals, you can control your blood glucose levels”

It is a nice idea and one that many hold on to, including health care professionals. When I needed to start taking insulin for meals, the parting words of my endocrinologist were “You know how to carb count, right?” I have heard tales of parents, caring for their type 1 child, taking a weighing scale wherever they go, weighing food to the gram to calculate the total carbohydrates.

An obsession with carbohydrates, while understandable, can set up an unhealthy relationship with food. I have spoken before on the risk of mental health issues, such as  orthorexia, which comes from unnecessarily strict diets.

The myth reflects a reality in diabetes that very little is straightforward and simple with this disease.

The Carb Counting Process

If we take the process of counting carbohydrates and then calculating how much insulin we need, it goes something like this:

  • We get served a meal we intend to eat
  • Based on the contents, and nutrition guides, we determine how many grams of carbohydrates there are in the meal
  • Using a IC ratio (the number of grams of carbohydrate needed to offset a specific number of units of insulin) we administer the right number of units of insulin to counter the carbohydrates.
  • Our blood glucose levels remain perfectly steady and never, ever, go too high or too low

Sadly, many people who follow this process do not achieve the last point. Here are some reasons why.

Problem 1: Many, Many Factors Affect Blood Glucose Levels

There is a good reason why a meal on one day can have a completely different effect on blood glucose levels than on another day. Diatribe identify 42 factors which affect blood glucose levels, the vast majority of which are independent of meals and their composition.

Problem 2: Not All Carbohydrates Are Created Equal, Not All Meals Are Created Equal

15 grams of sucrose will hit the bloodstream faster than the carbs in a slice of white bread (also, approximately 15 grams). Eating a slice of white bread with butter will hit the bloodstream differently than eating it without butter (even though the carb count is practically the same).

The glycaemic index tries to quantify “carbohydrate speed” but this only measures single items of food e.g. an apple but not mixed meals e.g. an apple with cheese. Combine this with pre-bolusing (depending on the insulin, we may need to administer the insulin well before the arrival of the food) trying to match the peak activity of the insulin with the emptying of the carbs into the blood and we can see there is a fair amount of art to the science.

Problem 3: Food Labels Are Not Perfectly Accurate

Putting aside the fact that some food items do not even carry nutrition guides e.g. beer bottles, fresh fruit, restaurant meals etc. even the items which do are not bulletproof.

NIST suggest the error margin for carbohydrates on nutrition labels is 2-5%. So that slice of bread, taking 15g as our “middle value” has between 14g and 16g of carbohydrates in it. The larger the number of carbs, the larger this range. So do we bolus for 14, 15, or 16g of carbohydrate? Do we need to measure to the gram when the labels are this inaccurate? Then there is the question of fibre…

Problem 4: Total Carbs vs Net Carbs

Here is a nutritional label for a popular brand of white bread in Australia (perhaps it is a personal bias but I find Australian food labels much easier to read than, say, US ones)

For two slices of bread (one serving), we have:

  • 31.1g of carbohydrate of which 2.2g are sugars
  • 5.2g of dietary fibre

On US food labels these two values are combined to form “Total Carbohydrates” so, in the US, two slices of this bread would have 36.3g of Total Carbohydrates as opposed to the 31.1g of “Net Carbs” we see here.

So which do we use for two slices of bread? 36.3g (error margin plus or minus 1.8g) or 31.1g (error margin plus or minus 1.5g). For me the answer is clear. Dietary fibre, while chemically a carbohydrate, cannot be broken down in the gut into glucose and passes through undigested. So bolusing for it makes no sense and it is Net Carbs we need to embrace. There is also the issue of sugar alcohols but let us assume, for simplicity, our meals do not contain significant amounts of these.

Problem 5: Calculating the IC Ratio is Problematic

To accurately work out the IC ratio, the only way I can think of to do this with any level of precision is to

  • Work out the carbohydrate sensitivity (how many grams of carbohydrate are needed to change blood glucose by a fixed amount)
  • Work out the insulin sensitivity (how many units of insulin needed to change blood glucose by a fixed amount)
  • Assuming it is the same fixed amount, divide the grams by the Units. So, for example, if I know 16g of glucose tablets raise my blood sugar by 1mmol/L (18 mg/dL) and I know it takes 2 units of insulin to lower my blood by the same amount, my IC ratio is 16/2 = 8.

We know we have an error margin of 2-5% with the carbohydrate amount. So what about the other factors?

Assuming we take the perfect reading (clean hands, ideal temperature etc.), glucometers are considered accurate if 99% of readings are within 15% of the lab result value.

Insulin pump delivery is accurate to within 5% and it is probably fair to assume injection by pen or syringe has a similar level of accuracy.

With all these error ranges, we can calculate how accurate a calculated IC ratio really is.

For our carbohydrate sensitivity we have 16g (error margin of about 0.5g) raising our blood glucose by 1mmol/L (error margin 0.15mmol/L). So the actual value lies somewhere between 15.5/1.15 = 13.5 and 16.5/0.85 = 19.5 (some rounding applied to keep numbers friendly).

For our insulin sensitivity we have 2 units of insulin (error margin 0.1 Units) lowering our blood glucose by 1mmol/L (error margin 0.15mmol/L). In this case the range for our insulin sensitivity is between 1.7 and 2.5.

Combining these, our IC ratio falls between 5.5 and 11.5. That is quite the range and means the amount of insulin required to cover a fixed amount of carbohydrate could literally be double the value we think it is and there is no way to know what is correct because of the inherent uncertainty in the measurements.

So Where To From Here?

Clearly, we need to keep using insulin so what do we do? The first step is to embrace the uncertainty and to accept, for all of the reasons above, sometimes there are going to be bad days where blood sugars misbehave.

We could go ultra-low carb but, for me, this is simply not practical, nor desirable. I enjoy eating at restaurants with family and friends even when no nutritional tables are available nor ultra-low carb options. I travel for work and have meals as part of that where keeping to, say, 32g of carb per day is almost impossible.

For someone without a Continuous Glucose Monitor (CGM), the best they can do is make the best guess for their IC ratio and periodically finger prick to see how it went. Courses like DAFNE can help with guessing the right amount of insulin to use.

For the most part, I stick to “lowish” carbohydrates i.e. I look for lower carb options when out and about but manage the spikes and understand the occasional high will NOT do me damage, it is simply part of having type 1 diabetes.

Because I do use a CGM, for the highs, I have a two-pronged attack. Firstly, I am running a loop (Android APS). This suspends glucose delivery when I am low and constantly adjusts the rate of insulin to try and keep me at my target glucose level (currently 6.0 mmol/L = 108 mg/dL). Android APS is very clever at automatically detecting meals so I generally do not declare carbohydrates when eating. This is not the recommended approach with Android APS but, so far, it seems to be working ok. This being said, insulins cannot always keep up with food so I also “Sugar Surf” with mini-bolusing to assist the loop. With this approach it is important to be mindful of the “insulin on board” levels as we do not want to combat the high only to induce a severe low through insulin stacking. This approach would be very dangerous without knowing the levels of insulin on board which, for me, is provided by Android APS.

For finding a workable value for the IC ratio which, those of us who use insulin need to do, it is a case of trial and error. Nightscout, my open source blood glucose tracker on the web, has an “AutoTune” feature where it can analyse your blood glucose results for a period of time and provide a “best guess” for values such as the IC ratio. Similarly, Android APS has graphs for sensitivity, insulin on board levels and departures from the expected insulin/carbohydrate behaviour to inform adjustments to values. Also, setting different IC values across the day, and periodically reviewing them to make sure the value I am using is still useful, keeps me from having too many highs and lows.


Carb counting is not bulletproof. While a useful tool to have in the toolkit, it is not the only one available, nor should it be seen as the only one that matters. There is an inherent level of inaccuracy in carb counting and this means, without other interventions, blood glucose will fluctuate and occasionally go where we do not want it to.

Other tools available to us are eating low carb, if practical, exercise such as a walk post-meal can help, and insulin intervention delivered manually using techniques such as Sugar Surfing, or automatically via a loop can also assist.

Find the tools that work for you and understand nothing is perfect, including blood glucose levels and accept that while diabetes cannot be perfectly controlled it can, in the long run, be very effectively managed.

One thought on “The Myth of Carbohydrate Counting

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s