We started by downloading the MMT and by getting familiar with its use. To do so, we ran several of the pre-loaded databases from different countries worldwide and we learned how to input data using, at the beginning, series of mock breastfeeding rates. We then retrieved and used real data from three cohort studies. Table 1 shows the main features of these studies.

The first cohort study was carried out to compare the cost of health care between breastfed and non-breastfed infants [20]. Out of 9,100 live births recorded in 1999 in FVG, a total of 842 mothers with their healthy newborn infants complying with pre-established inclusion criteria were enrolled soon after hospital birth between January and August 1999, and were followed up monthly to age 12 months. The denominators for the computation of the rates of any breastfeeding were always lower than 842 due to missing or invalid data. However, the loss to follow up was generally slight; the lower rate of attrition was 1% at one month of age, and the higher was 7% at 11 months.

The second cohort of 400 mothers with their healthy newborn infants was enrolled between July 2007 and July 2008 (10,569 and 10,515 annual live births, respectively, in FVG) and was followed up to 24 months of age, with data on breastfeeding gathered at 1, 3, 6, 9, 12, 18 and 24 months [21, 22]. This was actually a sub-cohort of a larger one intended to study the effect of exposure to heavy metals during pregnancy on the neuro- and psychological development of their offspring [23]. This large study was very demanding, also because of the need to collect several biological samples, and many women refused to participate or withdrew after some time [24]. As a consequence, the loss to follow up was considerable, with data on breastfeeding available from 287 (72%) of mothers at one month, 235 (59%) at 6 months, 165 (41%) at 12 months, and only 132 (33%) at 24 months.

The third cohort consisted of 265 mothers and children registered from January 1st to December 31st, 2016, in a paediatric practice in the city of Trieste [25, 26]. This corresponded to 18% of all the infants born in Trieste in 2016; the total number of live births in FVG in that year was 8,492. Data on breastfeeding were gathered during well-child visits at 1, 3, 5, 8, 12, 18, 24 and 36 months of age. The status of breastfeeding at discharge from the maternity ward was retrieved from hospital records. Apart from few missing or invalid data, the prevalence of any breastfeeding was available for 95% of the mothers at one month and 85% at 36 months of age.

In the three cohorts, data on breastfeeding were collected using the standard definitions and methods recommended by the WHO [27]. Before entering the data into the MMT, we set the initial sample size of each cohort as reference population. Then we entered into a spreadsheet numerators and denominators from the three cohorts, at each available age. After double checking for possible transcription mistakes, we copied the prevalence rates into the “country data” section of the MMT. When data for given ages (in months) were not available, the MMT added the missing data using its internal predictor. A breastfeeding rate chart, from birth to 36 months of age, was then generated, followed by several tables on actual and potential breastmilk production (litres), value estimates (USD), and lost breastmilk (litres and percentage). We repeated the same procedure for each cohort.

As far as the value estimates are concerned, the MMT assigns a monetary value of 100 USD per litre of breastmilk produced, based on the price of fresh human milk exchanged among human milk banks in Norway [17]. The MMT would allow to generate values in different currencies, by adding a rate of exchange. However, exchange rates vary over time. To avoid using different exchange rates and different currencies for each cohort study (Italian Liras in 1999 and EUR in 2007 and 2016), we decided to accept the MMT output in USD. Also, the exchange rates between USD and EUR fluctuate around 1. USD values, therefore, are a good proxy of EUR values.

Being based on published data that had been gathered after clearance by a committee of ethics, the study did not require to carry out further ethical clearance.