The InterConnect consortium investigated the association between fish intake and new-onset type 2 diabetes [1]. This required the following target variables (all in grams/day (g/d)) to be harmonised across participating studies:
- Total fish
- Fatty/oily fish (fat content > 4%)
- Lean fish
- Shellfish (crustaceans and molluscs)
- Saltwater fish
- Fresh water fish
- Fried fish
- Smoked or salted fish
Of 12 studies in the exemplar phase of this project (28 in total [1]), two studies assessed fish intake using diet history while the rest used food frequency questionnaires (FFQs). The number of questions on fish intake in the FFQs ranged between 1 and 22. At a basic level, all studies could create total fish; while not all studies contributed to analyses of different types of fish (see Table D.19.1).
Table D.19.1 Harmonisation potential across 12 studies.
| Target variables | |||||||
| Participating cohorts |
Total fish | Fatty fish | Lean fish | Shellfish | Fresh/ Salt water fish | Fried fish | Smoked/ salted fish |
| Study 1 (Australia) | Y | |
|
|
|
Y | |
| Study 2 (Europe) | Y | Y | Y | Y | Y | |
Y |
| Study 3 (Europe) | Y | Y | |
|
|
|
|
| Study 4 (Europe) | Y | Y | Y | Y | |
Y | |
| Study 5 (Asia) | Y | Y | Y | Y | Y | |
Y |
| Study 6 (Europe) | Y | Y | Y | Y | Y | |
|
| Study 7 (Asia) | Y | |
|
Y | Y | |
Y |
| Study 8 (Europe) | Y | Y | Y | Y | |
Y | Y |
| Study 9 (Europe) | Y | Y | Y | Y | |
|
Y |
| Study 10 (Europe) | Y | Y | Y | Y | |
Y | |
| Study 11 (Europe) | Y | Y | Y | Y | Y | Y | Y |
| Study 12 (Asia) | Y | ||||||
Processing rules to transform the variables from each study into the common target variable format were created and agreed with each study team. Examples of some of the pre-existing study variables and the algorithms or creation rules that were developed to derive target variables are shown below in Figures D.19.1 and D.19.2.
Challenges
- For some types of fish it was less clear whether these should be classified as lean or fatty fish and often studies have classified these differently.
- The fat content of certain fishes is variable across the globe; therefore decisions on how to classify specific types of fish might depend on local knowledge.
- Portion sizes differ across populations. Evidence or knowledge of local portion size is helpful to identify grams per portion or serving and calculate consumption in grams per day, the universal unit.
|
Fish items in the original cohort (assessed by FFQ) |
Fish types (assumption of g/portion) • Cod (150 g) • Salted or smoked fish (50 g) • Clam, oyster, mussels (60 g) • Prawn, king prawn, crayfish (100 g) • Octopus, squid, cuttlefish (150 g) • Total seafood per day |
Frequency never/almost never; 1-3 times/month; once a week; 2-4 times/week; 5-6 times/week; once per day; 2-3 times/day; 4-6 times/day; more than 6 times/day |
|
Target variable (g/d) |
Harmonisation - |
Harmonisation - |
|
Total fish |
total fish and seafood per day |
For total fish and total seafood, source data are in g/d. For the remainder, source data are in portion/day. Information on average portion size in grams is available. For Lean fish: multiply portion/day*150 g For fatty fish: multiply portion/day*150 g For Salted/smoked/dried fish: multiply portion/day*50 g |
|
Lean fish |
White fish/day + Cod/day |
|
|
Fatty fish |
Blue fish/day |
|
|
Seafood other than fish |
total seafood per day |
|
|
Fried fish |
Not assessed |
|
|
Salted/smoked/dried |
Salted or smoked fish/day |
|
|
Freshwater/saltwater fish |
Not assessed |
Figure D.19.1 Pre-existing data and algorithms used to derive target variables from a study using a food frequency questionnaire.
|
Fish items in the original cohort (assessed by diet history) |
Fish types • Cod • Baltic herring with bones; • Baltic herring; • Salmon; • Salmon salted; • Baltic herring salted with bones; • Herring slated; • Smoked Baltic herring with bones; • Sardine; • Smoked redfish • Perch; • Pike; • Flounder; • Bream; • Vendance with bones; • Fish frozen saithe; • Whitefish; • Fish average; • Fish in soup, average; • Roe; • Stockfish; • Vendance, salted with bones; • Smoked vendance with bones; • Smoked lamprey; • Smoked whitefish; • Smoked fish average; • Tuna • Shrimps |
Frequency g/d |
|
Target variable |
Harmonisation - |
Harmonisation - |
|
Total fish |
Total fish (sum of all available variables) - variables are mutually exclusive |
Variables already in g/d |
|
Lean fish |
• Cod • Stockfish • Fresh frozen saithe • Perch • Pike • Flounder • fish, average • fish in soup, average |
|
|
Fatty fish |
• Baltic herring with bones • Baltic herring • Salmon • Salmon salted • Baltic herring salted with bones • Herring slated • Smoked Baltic herring with bones • Sardine; • Smoked redfish • Whitefish • Vendance, salted with bones • Smoked vendance with bone • Vendance, with bones • Smoked fish average |
|
|
Seafood other than fish |
• shrimps |
|
|
Fried fish |
Not assessed |
|
|
Salted/smoked/dried fish |
• Salmon salted • Baltic herring salted with bones • Herring slalted • Smoked Baltic herring with bones • Smoked redfish • Vendance, salted with bones |
|
|
Freshwater fish |
• Perch • Pike • Bream • fish, average • fish in soup, average • Salmon • Salmon salted • Vendance, salted with bones • Smoked vendance with bone • Vendance, with bones |
|
|
Saltwater fish |
• Baltic herring with bones; • Tuna • Baltic herring • Baltic herring salted with bones • Herring slalted • Cod • Stockfish • Fresh frozen saithe • Flounder • Smoked redfish • Sardine • Smoked fish average |
|
Figure D.19.2 Pre-existing data and algorithms used to derive target variables from a study using a diet history.
