The subjects

• 55,525 cases
  • children born and diagnosed under the age of 15 in England and Wales from 1962-2008
  • taken from the National Registry of Childhood Tumours
  • each case has one or two individually matched controls selected from birth registers, matched for age, sex and birth registration sub-district.
• The resulting 116,815 subjects are largely the same as for the "follow on" study of OHLs, except:
  • confined to residency in England and Wales as they did not have the UGC records for Scotland.
  • some 3000 subjects fewer, largely cases diagnosed in 2008 and their controls together with a small number from earlier years, as the dataset for this analysis was frozen sooner than the one for the OHL analysis

The underground cables

Based on National Grid’s digital mapping of the locations of its UGCs, extracted for the purposes of this study as a single snapshot in 2011.

• High-voltage AC (400 kV, 275 kV and a very few 132 kV) and High-Voltage DC
• Records nearly 2000 km of underground cables, but the paper excluded UGCs entirely within substations, short lengths of UGC connecting a substation to the terminal tower of an OHL immediately outside it, and cables in deep bored tunnels (typically 20 m or more underground).
• The records are compiled for operational purposes and show all currently “live” cables, plus disconnected cables where the cable is still physically present in the ground.
  • They do not necessarily show all redundant cables or instances where a cable route has been changed at some point, which will have led to some errors
• recorded by 65,260 straight-line segments, of variable length as necessary to capture curves
  • generally accurate to 0.1 m or better.
• digital mapping information supplemented with relevant engineering information obtained from mainly internal National Grid records:
  • year of construction
  • spacing of the individual conductors
  • depth of burial
  • when two or more cables are in close proximity, relative phasing
• Some approximations were necessary:
  • cables that are buried directly in the ground (as opposed to being laid in concrete troughs) often have variable depths along their route, which were standardised as a single depth
  • as phasing often changes at joints, where joints between two different cables following the same route are staggered, the relative phasing will often be different for that length between joints, which was ignored.
• loads obtained from the same historical records as for studies of OHLs

Identifying location of addresses

• Grid reference of birth address obtained from Address-Point
  • For the 308 subjects where the Address-Point grid reference was within 40 m of the nearest UGC (the chosen criterion for when the spatial extent of the home is potentially significant compared to the distance), attempted to identify the home on large-scale mapping and to record the grid reference of the closest and farthest points of the building from the UGC.
  • also attempted to view each of these on Google StreetView to estimate the lowest and highest inhabited floor and to allow for sloping ground
  • site visits for 44 out of 51 addresses where either the spatial extent or the inhabited floors could not be estimated with acceptable confidence
    • This resolved 35 of these ambiguities.
    • The remaining 9 were mainly apartment blocks with no public access and were dropped from the analysis

The exposure metrics

• calculated the horizontal distance to the nearest UGC that had existed in the relevant birth year (including UGCs disconnected but still physically present).
• two distance estimates:
  • “closest”, the distance to the nearest point of the building, using only those subjects where this point was positively identified
  • “centre”, the distance to the midpoint of the building, this time including the distance to the Address-Point grid reference for subjects without robust nearest and farthest points.
• calculated magnetic fields for those subjects with successfully identified closest and farthest points and the occupied floors.
  • aim was to attempted a calculation for every subject where the field could possibly be >0.1 µT
  • attempted calculations for all subjects within 40 m of any UGC
  • For these subjects and cables, all cable segments within 100 m included in the calculation
• calculations at:
  • 1 m above floor level,
  • for each occupied floor (taking the height between floors as a constant 3 m),
  • at five points equally spaced from the nearest to the farthest point.
• two exposure estimates:
  • “closest”, the field at the nearest point on the lowest occupied floor
  • “average”, the mean of the field at all the points for that address.

As explained above, the paper calculates "closest" and "centre" distances, and "closest" and "average" magnetic fields.

Summary results

The summary results given in the abstract are:

• Trend estimates for leukaemia as shown by the odds ratio (and 95% confidence interval) per unit increase in exposure were:
  • reciprocal of distance 0.99 (0.95-1.03),
  • magnetic field 1.01 (0.76-1.33).

Full results

The full results tables given in the paper are:

 Table 1

Odds ratios (OR) and confidence intervals (CI) of distance to underground cables, based on “closest” distance estimate.  ORs are adjusted for socioeconomic status.  Trend is OR for each increase of 100/d. Bold indicates P<0.05. Italics indicates a combined, post hoc category (see text).

 Table 2

Odds ratios (OR) and confidence intervals (CI) of distance to underground cables, leukaemia only: “closest” distance estimate for earlier and later periods; “centre” distance estimate for whole period.  ORs are adjusted for socioeconomic status.  Trend is OR for each increase of 100/d.  0-9.9 and 10-19.99 categories are combined for “centre” distance estimate.