Cancer Risk in Parents of Patients With Schizophrenia

To the Editor: Reassuring as it is to see that a recent record linkage study from Denmark by Susanne Oksbjerg Dalton, M.D., Ph.D., et al. (1) accurately replicated our previous finding of a reduced incidence of cancer in the parents of patients with schizophrenia in Finland compared with the general population (2), I was surprised to recognize that this difference vanished when parents whose children were free of schizophrenia were chosen as the comparison group instead. Thus, the authors made a point that our common finding with the general population as the reference group should be invalid, and they claimed a “healthy parenthood effect” as the critical source of bias. However, when testing our genetic hypothesis, they should have considered parents exposed to cancer risk throughout their lifetime. I am not sure whether their method of having follow-up for cancer in parents starts only at the birth of the first child or, alternatively, at the birth (or age at disease onset) of the first child with schizophrenia might have biased their finding of equal cancer risk (e.g., given that some studies find schizophrenia risk to depend on birth order, e.g., Kemppainen et al. [3]). Also, the “healthy parenthood effect” they introduced from a Danish study that found parents of children with cancer at no higher incidence than the general population (4) did not receive general support from several other population studies in parents of individuals with cancer at a younger age (5–7), and it seems counterintuitive at least since a significant proportion of cancer at a younger age is known to occur on a genetic basis. In fact, the only other retrievable large population study that compared cancer risk between the relatives of patients with cancer and the relatives of otherwise deceased persons from Utah (instead of the general population) still found familial cancer risk increased (8). Therefore, I doubt the general validity of the “healthy parenthood effect.” It would have been useful to see whether cancer risk in the Danish comparison group of parous individuals with no child affected by schizophrenia was indeed lower than in the general population, including parous and nonparous individuals. Unfortunately, the lack of resources does not currently permit me to replicate, in turn, selection of a parous comparison group from the Finnish population register and analyses similar to those of the Danish study. Meanwhile, I commend the colleagues from Denmark for drawing benefit from the excellent epidemiological material available in Nordic countries, and I hope that other appropriate registers (e.g., Hemminki et al. [9]) will contribute.