Discussion
External validity describes to which extent the results of RCTs are applicable to patients treated in general practice. Together with the primary publication, the focus of this analysis was to investigate further aspects relevant for the external validity of the INSECT trial mainly according to the issues raised by Rothwell. As the data presented here comprises a significant amount of new information, it seemed reasonable to publish this data as a separate article.
Because of prospective data acquisition, external monitoring and external data analysis the INSECT trail is an optimal source for analyzing external validity. The INSECT trial was a multicenter German RCT. Thus, there were no differences based on different healthcare systems or countries. The participation was voluntarily suggesting that only motivated surgeons got involved. Even though the number of participating TCC was the lowest, the mean number of randomized patients did not differ between PCC, SCC and TCC, respectively.
One main conclusion of this paper would be that smaller hospitals such as PCC (or SCC) should be involved in clinical trials whenever possible because these hospitals are important resources in terms of patient recruitment. Not only that, but the here presented analyses also revealed that the randomization process occurred at similar speeds in all three types of hospitals; the adherence to the protocol was significantly better in PCC and SCC; the number of included patients and the main characteristics of included patients did not differ between the three hospital types; and the number of patients lost to follow-up did not differ significantly between PCC, SCC and TCC. Thus, the message of the INSECT trial (i.e. the rate of incisional hernias after midline incision is not dependent on the used suture material/technique) is truly independent of the hospital type in which the patient is operated.
Reasons for not participating in a surgical study are multifactorial and were not assessed systematically within this study. However, they include well-known reasons such as the dislike of randomization procedures, strict protocols for performance of interventions and unwillingness for follow-up investigations. In order to detect potential selection bias, the screening processes at the Departments of Surgery at the Universities of Heidelberg and Erlangen were assessed in detail. The process from screening until randomization could be described in detail in 116 patients out of 625 randomized patients (18.5%). Even though this alone may not allow us to draw general conclusions about external validity, other parameters such as the progress of randomized patients between the different hospitals and the clinical data of included patients were also similar. This suggests that the process from screening to randomization may be also comparable in all participating centers.
The largest group of excluded patients had a prior laparotomy. Because this patient group may not benefit from the results achieved during the INSECT trial, further studies are needed to define the optimal method of abdominal wall closure for patients with prior laparotomy. Out of the remaining potentially eligible patients, another 119 patients had to be excluded because of two reasons: firstly, patients refused randomization; and secondly, patients were already included in another RCT, one of which was the POVATI trial.
Compared to the multicenter INSECT trial, the POVATI trial was conducted as a single center trial at the Department of Surgery, University of Heidelberg. Given the inclusion and exclusion criteria, patients with primary laparotomy were eligible to participate in both RCTs. The decision by patients to participate in one or the other study could have an impact on external validity. Because there were no differences in demographics and baseline characteristics between patients included in either trial, it is unlikely that patients enrolled in the POVATI trial will affect the external validity of the INSECT trial (Table 4). The comparison of the 130 potential eligible patients in Heidelberg revealed significant differences only in the applied incision and closure techniques in non-consenting patients (n = 44) and patients randomized to the POVATI trial (n = 26). For non-consenting patients, surgeons could choose freely their preferred techniques for opening and closing the abdomen. When given the choice, the vast majority of surgeons at our hospital chose midline laparotomy as the primary abdominal approach and continuous suturing as preferred closure technique. This may suggest that surgeons at our hospital were somehow biased towards midline laparotomy and continuous suturing (Table 4).
In general, the random ratio ("enrollment fraction" according to Gross et al.) describes the proportion of randomized patients to the primarily eligible patients. The random ratio per se does not say anything as to whether the results of a study are widely acceptable. It is more likely to be a marker of the efforts necessary to randomize a patient. Even though the random ratio at the Department of Surgery at the University of Erlangen was similar to that at the University of Erlangen, the recruitment period lasted longer in Erlangen (16 months versus 6 months at Heidelberg). One possible explanation could be that all surgical trainees in Heidelberg rotate into the clinical study centre and residents who already finished their rotation are part of an on-call program. This setting allows screening and inclusion of patients even at night or on weekends. In contrast, one clinical investigator was primarily responsible for patient recruitment in Erlangen. Considering missing days due to holidays, weekends, days off after being on call, etc., screening and recruiting of patients was only possible on 120 days.