Pooled analyses of childhood leukaemia and magnetic fields

There have been more than twenty individual epidemiological studies of magnetic fields and childhood cancer.  They are best understood by pooling the results from the individual studies.  That was done in 2000 in the "Ahlbom" pooled analysis with an update in 2010 by Kheifets et al.

Ahlbom - the summary

In 2000, an international group, led by Professor Anders Ahlbom from Sweden, took all the recent separate epidemiological studies of childhood leukaemia and magnetic fields and pooled the results, so that they could perform one single re-analysis of all the available data. They found that, statistically, there was little or no suggestion of an increased risk at the levels of magnetic fields to which the overwhelming majority of children are exposed. The study did, however, find that in the category of homes with a field, averaged over 24 hours, of greater than 0.4 microteslas (which affects fewer than half a percent of children in the UK) there is a statistical suggestion of increased risk. Some of these homes are near power lines, but many are not. They concluded:

“The explanation for the elevated risk is unknown, but selection bias may have accounted for some of the increase.” Full abstract.

graph showing Ahlbom results

Ahlbom - the results table

From “A pooled analysis of magnetic fields and childhood leukaemia”, Ahlbom et al, British Journal of Cancer (2000) vol 83 pp 692-698

Based on Table 3 from the published paper.

Total leukaemia. Relative risks (95% confidence interval in parentheses) by exposure level with adjustment for age, sex and socioeconomic status (measurement studies) and East/West in Germany. Reference level: <0.1 µT. Observed and expected case numbers ≥ 0.4 µT, with expected numbers given by modelling probability of membership of each exposure category based on distribution of controls including covariates.

Type of study
0.1-<0.2 µT
0.2-<0.4 µT
≥ 0.4 µT
≥ 0.4 µT
Observed
Expected
Measurement studies

Canada

1.29 (0.84-1.99)

1.39 (0.78-2.48)

1.55 (0.65-3.68)

13

10.3

Germany
1.24 (0.58-2.62)
1.67 (0.48-5.83)
2.00 (0.26-15.17)
2
0.9
New Zealand
0.67 (0.20-2.20)
4 cases/0 ctrls
0 cases/0 ctrls
0
0
UK
0.84 (0.57-1.24)
0.98 (0.50-1.93)
1.00 (0.30-3.37)
4
4.4
USA
1.11 (0.81-1.53)
1.01 (0.65-1.57)
3.44 (1.24-9.54)
17
4.7
Calculated field studies

Denmark

2.68 (0.24-30.45)

0 cases/8 ctrls

2 cases/0 ctrls

2

0

Finland
0 cases/19 ctrls
4.11 (0.48-35.1)
6.21 (0.68-56.9)
1
0.2
Norway
1.75 (0.65-4.72)
1.06 (0.21-5.22)
0 cases/10 ctrls
0
2.7
Sweden
1.75 (0.48-6.37)
0.57 (0.07-4.65)
3.74 (1.23-11.37)
5
1.5
Summary

Measurements studies

1.05 (0.86-1.28)

1.15 (0.85-1.54)

1.87 (1.10-3.18)

36

20.1

Calculated field studies
1.58 (0.77-3.25)
0.79 (0.27-2.28)
2.13 (0.93-4.88)
8
4.4
All studies
1.08 (0.89-1.31)
1.11 (0.84-1.47)
2.00 (1.27-3.13)
44
24.2

Ahlbom - alternative interpretations

The Ahlbom pooled analysis is a highly influential analysis of the results from 9 epidemiological studies of magnetic fields and childhood leukaemia.   This section discusses alternative ways of presenting the results.

The paper uses four categories of magnetic field and the normal way of presenting the results is to show the risk ratios in these four categories:

 graph showing Ahlbom results by categories

This is suggestive of what has become the commonest interpretation of the results: that there is a raised relative risk in the top exposure category, fields greater than 0.4 μT, but little evidence of increased risk in the lower exposure categories.

This presentation gives the four categories equally spaced.  But the categories are not in fact equally spaced in terms of field.  The paper itself does not give the average field in each category, but we can make intelligent guesses, and plot the results against the average field in each category:

 graph showing Ahlbom results properly spaced

This presentation is more suggestive of a smoothly increasing risk.  In fact, the paper itself performs what is known as a "continuous analysis" and finds a relative risk of 1.15 per 0.2 μT increase in exposure.  We can plot this continuous estimate of risk:

graph showing Ahlbom results with trend line 

On this interpretation, there is a raised risk at low fields.  But also, the increase in risk at higher fields is less than suggested by the original presentation.  Instead of the risk being doubled for fields above 0.4 μT, the risk does not become doubled until 0.99 μT.

Finally, these presentations are all very dependent on the last point - the risk in the top category >0.4 μT.  This is a wide category - the average field in it is 0.7 μT.  What happens if we split it up?  The Ahlbom paper does contain some further results for risks in categories 0.3-0.4-0.5 and >0.5 μT (and again we have to make educated guesses as to the average field in each category), but unfortunately not with confidence intervals:

graph showing Ahlbom results with top category 

This presentation suggests there is more variability in the results and perhaps any attempt to see a pattern is going to have limited success.

Kheifets update

The epidemiology of magnetic fields and childhood leukaemia is usually summarised by the Ahlbom et al pooled analysis from 2000 - see above - which combined the results from all the individual studies that met certain criteria.

Since then, further studies have been published.  A new pooled analysis, by Kheifets et al in 2010, looked at those more recent studies.  It concluded that the studies since 2000 broadly confirm the Ahlbom finding.

If all the studies are included, the results compared to Ahlbom look like this:

graph of Kheifets results all studies

That suggests the association has got slightly weaker.  But one of the studies included, from Brazil, is believed to have a particular bias that may make its results less reliable.  If the Brazil study is excluded and all the remaining studies compared, this is the result:

Graph showing Kheifets results without Brazil

Note that the more recent studies together have only about a third as many exposed subjects as the original Ahlbom analysis.  So they support rather than replace the original analysis.

British Journal of Cancer (2010) 103, 1128–1135.
Pooled analysis of recent studies on magnetic fields and childhood leukaemia
L Kheifets, A Ahlbom, C M Crespi, G Draper, J Hagihara, R M Lowenthal, G Mezei, S Oksuzyan, J Schüz, J Swanson, A Tittarelli, M Vinceti and V Wunsch Filho

Background:  Previous pooled analyses have reported an association between magnetic fields and childhood leukaemia. We present a pooled analysis based on primary data from studies on residential magnetic fields and childhood leukaemia published after 2000.

Methods:  Seven studies with a total of 10,865 cases and 12,853 controls were included. The main analysis focused on 24-h magnetic field measurements or calculated fields in residences.

Results:  In the combined results, risk increased with increase in exposure, but the estimates were imprecise. The odds ratios for exposure categories of 0.1–0.2 μT, 0.2–0.3 μT and 0.3 μT, compared with <0.1 μT, were 1.07 (95% CI 0.81–1.41), 1.16 (0.69–1.93) and 1.44 (0.88–2.36), respectively. Without the most influential study from Brazil, the odds ratios increased somewhat. An increasing trend was also suggested by a nonparametric analysis conducted using a generalised additive model.

Conclusions:  Our results are in line with previous pooled analyses showing an association between magnetic fields and childhood leukaemia. Overall, the association is weaker in the most recently conducted studies, but these studies are small and lack methodological improvements needed to resolve the apparent association. We conclude that recent studies on magnetic fields and childhood leukaemia do not alter the previous assessment that magnetic fields are possibly carcinogenic.

Subanalyses - age of child and type of leukaemia

Age of child

The Kheifets pooled analysis compared older and younger children at diagnosis:

  0.1-0.2 µT0.2-0.3 µT>0.3 µT
Kheifets (excluding Brazil)Age at diagnosis ≤81.21 (0.83-1.78)1.34 (0.63-2.86)1.63 (0.75-3.54)
Age at diagnosis >81.01 (0.50-2.02)1.21 (0.29-4.99)1.30 (0.30-5.63)

The paper says:

Risks were a little higher for ... a younger age group

The Ahlbom pooled analysis did not include this analysis.  The only one of its nine component papers to look at age was Feychting and Ahlbom from Sweden, where the results are too imprecise to allow conclusions:

Age at diagnosisOR (95% CI) for >0.2 µT
0-41.7 (0.4-5.9)
5-95.7 (1.0-26.0)
10-152,6 (0.1-22.2)

Type of leukaemia

The commonest type of leukaemia is Acute Lymphocytic leukaemia (ALL).  ALL comprises something like 80% of total leukaemia, so we expect little difference between findings for total leukaemia and ALL only, because they are mostly the same cases.  For what it's worth, both the Ahlbom and Kheifets pooled analyses found slightly higher risks for ALL alone:

  0.1-0.2 µT0.2-0.4 µT>0.4 µT
Ahlbom (all studies)total leukaemia1.08 (0.89-1.31)1.11 (0.84-1.47)2.00 (1.27-3.13)
ALL1.08 (0.88-1.32)1.12 (0.84-1.51)2.08 (1.30-3.33)
  0.1-0.2 µT0.2-0.3 µT>0.3 µT
Kheifets (all studies)total leukaemia1.07 (0.81-1.41)1.22 (0.78-1.89)1.46 (0.80-2.68)
ALL0.97 (0.71-1.31)1.19 (0.70-2.02)1.56 (0.93-2.60)

The Ahlbom paper says:

Since the ALL cases make up as much as 83% of all cases and since the controls are the same, the ALL results must be similar to the total leukaemia results. The results in Table 4 show that this is indeed the case, but in the highest exposure category the ALL relative risks are somewhat higher than for total leukaemia.

We also looked separately at other leukaemia to see whether the observed excess risk was restricted to the ALL group. The summary relative risk for other leukaemia was 1.42 in the highest exposure category, but based on only 4 exposed cases.

The Kheifets paper says:

Risks were a little higher for ALL...

 

Other pooled analyses

The Greenland pooled analysis was published in the same year, 2000, as Ahlbom.  It did not have quality criteria for including studies so included more studies but found similar results.

There was also a pooled analysis as part of the investigation of nighttime exposure.