Overall, the present study indicates that a considerable proportion of sCJD may constitute a health care-related disorder, accidentally transmitted during surgery. While this has been suggested before [7
], the present study is unique because of the unbiased assessment of exposure histories for decades before disease onset, randomly chosen controls, and strict lag time measurement. The lack of surgical history data prior to the establishment of the National Hospital Discharge Registries in the early 1970s and low statistical power preclude the assessment of early-in-life surgery or specific infrequent procedures. The main findings were supported by analyses including both MCs and UMCs. The UMCs were not essential for this study, but it was reassuring to learn that a similar study could be undertaken in another setting where it was impossible to sample MCs as a reference group.
The validity of our results could be affected by selection bias if CJD detection rates were higher among surgical patients subjected to medical follow-up due to their underlying disease or to detect possible complications of surgery. Since most associations were only seen for long latency periods (i.e., 20 years or more prior to onset of CJD), such a mechanism appears highly unlikely. One exception, however, might be the case of coronary surgery, but would only be plausible, if there was considerable CJD underascertainment in persons where coronary surgery was indicated.
Another potential source of selection bias could be changes in diagnostic accuracy which inevitably occur over such a long study period, notably increased awareness of CJD after the report of vCJD in 1996 and availability of the test for protein 14-3-3 in cerebrospinal fluid since 1998. However, at least in Sweden, sCJD incidence for 1970–1998 [25
] was similar to that reported during active surveillance thereafter. Consequently, our low study accrual rate for the period 1987–1996 might best be attributed to difficulties in finding hospital records of these old cases, and would thus be unrelated to underascertainment or surgical history.
Differential misclassification of iatrogenic cases as sCJD cases might have occurred due to overlooked use of dura mater grafts in surgery. In Sweden, inspection of medical records revealed the presence of 1 case of iatrogenic CJD who died in 2002 and was removed from the study. While dura mater grafts may, in theory, have been used in gynecologic surgery for urinary incontinence and in other surgery, the lack of reported iatrogenic CJD in Sweden [25
], and the shorter incubation periods described for iatrogenic CJD [26
] suggest that overlooked use of heterologous dura mater grafts does not explain the excess risk of sCJD for individuals exposed to specific groups of SP in our study.
Surveillance bias, i.e., overascertainment of sCJD among the 54 probable sCJD cases without postmortem confirmation, cannot be ruled out. However, it seems unlikely that such misclassification would be associated with surgical history ≥20 years ago, and will therefore be nondifferential.
The geographically based referral system for surgery in both countries, with free and equal access to health care, registration of SPs prior to sCJD diagnosis, and similar expected-versus-registered times of residence should have minimized differential misclassification of exposure. Nondifferential misclassification of exposure to surgery, potentially resulting from our window design [27
], errors in person numbers, or coding of SPs should be low. Moreover, since their additional effect is dilution, i.e., OR towards unity [28
], true associations should be higher than those observed. Owing to our study design, recall bias can be ruled out.
Potential confounders are overlooked causes of CJD associated with surgery, not implicated in the same causal chain. Intra- or postoperative blood or blood component transfusion and skin incision are potential confounders which we did not control for. Neither surgery undergone in the 1970s nor body system groups with highest observed excess risk appear to be especially correlated with considerable loss of blood, blood transfusion or size of surgical incision. Potential confounding by variation in surgical practices or indications was controlled for by the study design with year-to-year entry of cases and controls into the study.
The fact that the highest ORs were observed for the earliest time window (≥20 years before CJD onset) fits well with epidemiological theory [24
] and with the long latency required for the process of neuroinvasion following surgical transmission [29
], thus making a causal association plausible. Since cognitive impairment and brain emboli following coronary artery bypass grafting have been reported [30
] and 2 cases exposed to coronary surgery were probable sCJD, bias due to misdiagnosis or indication of surgical treatment cannot be ruled out for heart and major vessel surgery.
Conflicting reports from previously published case-control studies regarding the association between surgery and sCJD may partially be explained by variations in the type of control subjects used and in exposure assessment [31
]. Even though we did not find a significant association between surgery performed at any time predating >1 year before onset, our results are consistent with positive results for lifetime surgical history found in recent, large studies using community controls [7
]. Like us, one study reports dose-response effects [7
]. Hence, in accordance with these studies and other reports suggesting that some types of surgery, e.g. cataract surgery [32
], might be performed as a consequence of early sCJD manifestations, it appears that surgery may constitute: (1) a risk factor of sCJD with a considerable time lag, and (2) a risk indicator with yet unknown causal links for specific types of (recent) surgery, possibly including coronary surgery.
What then are the potential implications for public health? Based on period-specific ORs of 1.44 (10–19 years) and 2.44 (≥20 years), and 22 and 19% exposed, unrepeated cases for each window, we estimate the population-attributable proportion to be 18%. This figure may constitute a considerable underestimation of the effect of lifetime surgical history on sCJD incidence, because in our study the ascertainment of exposure to surgery before the middle of the 1970s was compromised by the incompleteness of hospital discharge registries. In the UK, a positive lifetime surgical history was recorded from 1980 onwards in 58 and 59% of vCJD cases and controls, with median ages of 26 and 33 years [34
]. Since the majority of surgical interventions at such ages was not captured in our study, and our results suggest that the lower the age at surgery, the higher the risk, the figure of 35% proposed by Ward et al. [9
] for the population-attributable risk of sCJD due to lifetime surgery in the UK would appear to be a conservative estimate for Denmark and Sweden.
While negative results have been reported for clustered surgical chains [35
], studies supporting the hypothesis of a frequent surgical transmission of sCJD might be a French cluster of multioperated cases [36
]; the high sCJD incidence in regions with a high incidence of genetic transmissible spongiform encephalopathy, potentially acting as a focus for point source epidemics, such as in the Spanish Basque Country [37
), and occasional increases in CJD incidence in countries where iatrogenic transmission has been mentioned as a possible cause [38
To conclude, we provide evidence to indicate that surgery, acting with long incubation periods, has constituted a risk factor for sCJD in Sweden and Denmark. The associations may have implications for precautionary measures and surveillance.