Type | Cohort | Cohort (extension of Rothman et al, 1996) | Hospital based case-control | Hospital based case-control | Retrospective cohort | Hospital based case-control |
Endpoint(s) | Overall mortality | Brain cancer, leukaemia, motor vehicle accidents, circulatory deaths | Primary brain cancer | Brain tumours (primary brain cancer, meningioma, acoustic neurinoma) | Cancer incidence. Brain and nervous system cancer, salivary gland cancer, leukaemia of a priori interest | Acoustic neurinoma |
No. cases/controls; size of cohorts | 59245 portable bag/car; 48932 hand-held; 147340 unknown phone type | 152138 portable bag/car; 133423 hand-held | 469 cases; 422 controls | 489 malignant; 293 benign tumours; 799 controls | 420095 subscribers (154 brain and nervous tumours, 84 cases of leukaemia) | 90 cases; 86 controls |
Exposure assessment | Company records | Company records | Interview | Interview | Company records | Interview |
Outcome assessment | SSA Death Master File | National Death Index | Pathology and MRI | Histopathology, MRI/CT | Danish Cancer Registry | Pathology and MRI |
Telephone type(s) | Analogue | Analogue | 88% analogue | Not specified | 42% analogue; 58% digital | Not specified |
Duration of follow up/duration of phone use | 1 year follow up; 1.8 y av. phone use | 1 year follow up; 1.9 y av. phone use | 2.8 y av. phone use, 14% users among cases, of these 74% less than 4 y | 18% regular users in cases, of these 84% less than 5 y | 1–14 y of follow up. Overall 92% less than 5 y, digital phones: 93% less than 3 y | 20% users in cases, of these 61% 3–6 y |
Confounders considered* | None | Metropolitan area | Years of education, race, study centre, proxy interview, month and year of interview | Education, income, date of interview, proxy interview, race, hospital, distance residence to hospital | None | Education, study centre, occupation, date of interview |
Findings | No difference in overall mortality between users of different types of telephones | Increasing mortality from motor vehicle accident with increasing intensity of use. Higher mortality for brain tumours for longer duration of use (but only overall 2 deaths) | Overall no increased OR. Highest OR for neuroepitheliomatous cancer (2.1). Tumours occurred more frequently at the side the telephone was used (p = 0.06) | Overall no increased OR. No association with side of the head the phone was predominantly used | Overall cancer incidence reduced in men but not in women. Testicular cancer slightly increased (SIR 1.12). No overall increased brain tumour incidence. Highest SIRs for brain tumours and leukaemia for longest duration of use | Overall no increased risk. OR = 1.7 for 3–6 y of use. Cases used a mobile phone on av. 4.1 y compared to 2.2 y in controls. Tendency for greater proportion of contralateral use (p = 0.07) |
Evaluation
|
Selection of participants | Only 33% of subscriber records selected, in only 14% type of telephone ascertained | Same as Rothman et al, 1996 | Hospital controls including cancer patients. Except first year only prevalent cases. Response rate: cases 82%, controls 90% | Hospital controls. Response rate: cases 92%, controls 86% | Only 58% of subscribers selected | Hospital controls. Only prevalent cases. Response rate not specified |
Power† | 80% for overall survival | Negligible for cancer causes of death | 88% overall. For >4 y of use and excluding glioblastoma (52%) power less than 20% | 99% overall. For >3 y use and excluding glioblastoma (49%) power less than 40% | Overall brain tumours: 100%; leukaemia 100%. For ⩾3 use 62% for brain tumours, 39% for leukaemia | 50% overall. For >3 y 25% |
Exposure assessment | None | Intensity and duration of use only from company records | Interviewer not blinded to case status | Interviewer not blinded to case status | No data on intensity of use, duration of use only for digital phones from company records | Interviewer not blinded to case status |
Outcome assessment | None | None | Histopathology not unequivocal in all cases | State of the art | None | State of the art |
Confounding and bias | More than 70% of users excluded. Exposure misclassification. Assessment of laterality not possible. Healthy group effect | Same as Rothman et al, 1996 | Cases interviewed within 48 h after surgery. Response and recall bias possible | More proxy interviews in cases than controls | Comparison to general population containing users and non-users. More than 42% of users excluded. Exposure misclassification. Assessment of laterality not possible. Healthy group effect | Indication of reversal of cause and effect |
Latency considered | No | No | No | No | No | No |
Statistical methods | Standard | Standard | Standard, laterality by χ2 test | Standard, laterality by Fisher’s exact test | Standard | Standard, laterality by Fisher’s exact test |