Leakage of cerebrospinal fluid (CSF) through the dural tear is considered the most plausible cause of postdural puncture headache (PDPH) [1,2]. This hypothesis has been substantiated in volunteers when headache was provoked by withdrawal of CSF, and relieved by replacement of CSF with an equal amount of saline . Further supporting this hypothesis is the fact that larger diameter spinal needles not only cause a higher rate of CSF leakage through the dural tear than needles of smaller size  but also a higher incidence of PDPH [1,5]. Based on these findings, one would expect that repeated dural punctures should also increase the incidence of PDPH by increasing leakage of CSF through the multiple dural tears. Indeed, Harrison and Langham  found an increased incidence of PDPH when more than one attempt at subarachnoid puncture was made; however, interpretation of their findings is difficult because of a stunning 24% incidence of PDPH in their elderly population. In contrast, a prospective study by Lybecker et al.  found a trend (P = 0.09) but failed to find a significant correlation between the number of dural punctures and the incidence of PDPH.
Because of the low incidences of both PDPH and failed spinal anesthetics, we calculated that insufficient statistical power of the study had caused this unexpected finding. This conclusion was supported by the finding that of 1021 spinal anesthetics only 75 patients had suffered from PDPH . Moreover, based on incidences of failed spinal anesthetics (4%-5%) reported in previous studies [8,9], the number of patients undergoing repeated dural punctures [not described by Lybecker et al. ] was probably even lower. We concluded that there was an insufficient sample size in the study of Lybecker et al.  for analyzing the correlation between repeated dural punctures and PDPH. Consequently, we designed the present study to prospectively evaluate whether repeated dural punctures increase the incidence of PDPH. We compared our patients who had uneventful spinal anesthetics with those who had a failed spinal anesthetic and underwent repeated dural puncture for a second subarachnoid injection of local anesthetics because there was insufficient anesthetic level for the scheduled surgical procedure.
The sample size calculation was based on a pilot study in our institution that suggested a 4% incidence of PDPH in patients undergoing a single dural puncture, and on published studies suggesting a 4% incidence of failed spinal anesthetics [8,9]. Because of the correlation between the size of the dural tear and the incidence of PDPH [1,5], we hypothesized a twofold increase in the incidence of PDPH (8%) in patients undergoing repeated dural puncture for a second subarachnoid injection of local anesthetics. To test this hypothesis at a 95% confidence level and 80% power, it was necessary to have a sample size of 7098 spinal anesthetics consisting of 6825 spinal anesthetics having a single dural puncture and 273 spinal anesthetics requiring two dural punctures.
The large number of patients required for this study placed into question the feasibility of such a study at our institution. However, in 1990, when Lybecker et al. , published their study, we began a quality assurance program which includes prospective collection and storage of data covering the preoperative, intraoperative, and postoperative period of all anesthetics performed at our institution . The institutional ethics committee approved use of the quality assurance database to address our study question.
The data, collected on an extra sheet added to the traditional anesthesia record, were automatically entered into the computer database by an optical scanner. Preoperative data included patient information such as age, sex, and ASA physical status. The intraoperative data included information regarding the anesthetic technique (e.g., spinal anesthesia), and anesthetic problems (e.g., insufficient level of spinal anesthesia, second subarachnoid injection of local anesthetics). Preoperative and intraoperative data were collected by the anesthesiologist in charge. The postoperative data included general information of the anesthetic experience (e.g., patient satisfaction) and specific information regarding side effects (e.g., PDPH, nonpostural headache). These postoperative data were collected by trained quality assurance nurses on Day 4 after surgery using a standardized questionnaire. For patients who had been discharged, these data were collected by a telephone interview on Day 4 or 5 after surgery using the same questionnaire. Patients who could not be interviewed postoperatively were excluded from the study. Side effects and complications detected in both groups were communicated to and treated by the anesthesiologists in charge and analyzed in monthly departmental meetings.
The technique used for spinal anesthetics at our teaching hospital is standardized. With the patient in the lateral decubitus or sitting position, a midline approach is performed in the vast majority of spinal anesthetics. Using an interspace between L2 and L5, the needle is advanced into the subarachnoid space at a 60-90 degrees angle. As a general rule, Quincke spinal needles are used in patients >65 yr of age, whereas noncutting spinal needles are used in younger patients. The relative use of the different needles in our institution during the study period is indicated by the percentages of the needles bought in this period: Quincke 22-gauge and 25-gauge needles (Spinocan; Braun Melsungen, Melsungen, Germany) 21% and 23%, respectively, noncutting 24-gauge needles (Sprotte; Pajunk, Geisingen, Germany) 29%, and noncutting 25-gauge needles (Polymedic; Temena SRL, Bondy, France) 27%. Isobaric or hyperbaric 0.5% bupivacaine, and occasionally (before 1993) 5% hyperbaric lidocaine were used as local anesthetics (Astra, Sodertalje, Sweden). Hypotension was treated with volume expansion, Trendelenburg position, and, if necessary, intravenous ephedrine. Intravenously administered midazolam was used for intraoperative sedation. If there was no surgical reason for recumbancy, early mobilization was recommended.
Based on the required sample size, approximately 2000 spinal anesthetics per year and approximately 80% of completed postoperative followups, we expected that a period of 5 yr had to be studied. In fact, between April 1, 1990 (when we started the quality assurance program), and March 31, 1995, a total of 10,077 spinal anesthetics in 8366 patients has been performed at our institution (every spinal anesthetic is counted as a new case if the postoperative followup of a previous anesthetic has been completed). Excluded from study evaluation were 171 anesthetics consisting of a combined spinal-epidural technique and 1872 anesthetics (19%) with incomplete or unobtainable postoperative followups. Thus, a total of 8034 spinal anesthetics were available for evaluation. Out of these 8034 anesthetics, 7869 (97.9%) were uneventful spinal anesthetics, whereas 165 (2.1%) were failed spinal anesthetics followed by repeated dural puncture for a second subarachnoid injection of local anesthetics. The two groups were similar with regard to age, sex, ASA physical status, and type of surgery Table 1.
chi squared test was used to analyze the influence of repeated dural punctures on PDPH, nonpostural headache, and future refusal of spinal anesthesia. Logistic regression was used to analyze the correlation between age and PDPH.
Repeated dural punctures significantly increased the incidence of PDPH Table 1. Moreover, "good" patient satisfaction with the anesthetic experience was significantly less frequent in the group who had undergone repeated dural punctures. In contrast, the incidence of nonpostural headache was similar in both study groups.
Our results in patients undergoing an uneventful spinal anesthetic (a single dural puncture) also showed an inverse correlation between age and incidence of PDPH Figure 1. In contrast, sex was not a predictor of PDPH, as PDPH was diagnosed in 63 of 4138 (1.5%) men and 60 of 3731 (1.6%) women.
Our study shows that repeated dural punctures increase the incidence of PDPH. In view of the hypothesis that leakage of CSF through the dural tear is the cause of PDPH, this finding is expected. It may appear surprising that the correlation between repeated dural punctures and incidence of PDPH was not found in previous large studies [1,7]. However, since the incidence of failed spinal anesthetics is very low, it was necessary to study large numbers of spinal anesthetics to get a sufficiently large study group of patients undergoing repeated dural punctures. The increased incidence in PDPH after repeated dural punctures is not only statistically significant but also clinically relevant.
Our study also confirms the previous finding of an inverse correlation between age and incidence of PDPH [1,7] but fails to find a correlation between sex and incidence of PDPH.
The failure of a previous large study to detect repeated dural punctures as a predictor of PDPH  illustrates a problem faced in many studies. Although large enough for studying frequently occurring predictors of PDPH, that study was not designed for analyzing infrequently occurring events as predictors, such as repeated dural punctures. Communicating such negative findings that are problematic because of insufficient statistical power may be justified when the limited statistical power is pointed out, and there is a detailed description of the results (number of patients and outcomes in each group) instead of only mentioning a P value.
Several limitations of our study need to be considered. Some patients suffering from PDPH were probably missed because the postoperative followup of our quality assurance program included only 4 postoperative days but PDPH may begin as late as several weeks after dural puncture . As this limitation would apply to both study groups, it does not detract from the validity of our conclusion that repeated dural punctures increased the risk of PDPH. Another limitation is that the nurses performing the postoperative followup had access to the intraoperative data, thus potentially knowing whether the patients had undergone one or repeated dural punctures. However, even if they were aware of the detailed intraoperative data, investigator bias is unlikely because the nurses were unaware of our hypothesis and the ongoing study. Another limitation of our database is that information about factors that might influence PDPH (e.g., previous history of PDPH, or technical aspects of dural puncture) was not available. This limitation did not prevent testing our study hypothesis but only restricted the scope of our study. Finally, we cannot exclude the possibility that inadvertent dural punctures occurred in some patients of both groups. Therefore, more than one dural puncture may have occurred in some patients of Group 1 (undergoing one subarachnoid injection of local anesthetics) and more than two dural punctures in some patients of Group 2 (undergoing two subarachnoid injections). One can speculate that this possibility may apply more to Group 2, because failed spinal anesthetics have been explained by technical problems [8,9,11], and technical problems may be related to inadvertent dural punctures. If true, this would explain the more than twofold increase in PDPH in Group 2.
In sum, these limitations mean that analyses of prospectively collected data derived from our quality assurance program do not have the strength of a randomized, blinded prospective study. However, analyses of prospectively collected data are clearly superior to those of retrospective studies [8,11] or inquiries , as they provide some of the advantages of a prospective study (a consistent and detailed prospective postoperative followup based on a standardized questionnaire). As an alternative, a prospective multicenter study might also have been used for testing our hypothesis. However, multicenter studies have limitations, such as inconsistencies between different centers in conducting the study. More importantly, the potential advantages of a prospective multicenter study have to be weighed against their major additional cost. We conclude that using our quality assurance database is an adequate scientific approach for studying some hypotheses that require huge sample sizes and, therefore, cannot be tested by a prospective, one-center study.
In conclusion, our study shows that repeated dural punctures increase the incidence of PDPH. Therefore, increased risk of PDPH is a disadvantage of performing a second subarachnoid injection of local anesthetics after a failed spinal anesthetic. Moreover, this result confirms the assumption that leakage of CSF through the dural tear is the most plausible cause of PDPH. Previous studies may have failed to find a correlation between repeated dural punctures and increased incidence of PDPH because the sample size was too small to evaluate low incidences. Finally, the present study shows that using a prospectively collected quality assurance database is a feasible approach for evaluating specific questions that require very large sample sizes.
We are grateful to Roberta Angelini, RN, and Sonja Gaiardo, RN, who performed the postoperative followups for all patients undergoing anesthetics in the Department of Anesthesia at the Kantons-spital Basel and thus not only made this study possible but are helping to improve the standard of care in our department. We also thank Bruno Wunderlin for database support, and Joan Etlinger for expert editorial assistance.
1. Vandam L, Dripps R. Long-term follow-up of patients who received 10,098 spinal anesthetics: syndrome of decreased intracranial pressure (headache and ocular and auditory difficulties). JAMA 1956;161:586-91.
2. Reid JA, Thorburn J. Headache after spinal anaesthesia [editorial]. Br J Anaesth 1991;67:674-7.
3. Tourtellate W, Haerer A, Heller G, Somers J. Postlumbar puncture headaches. Springfield, IL: Charles C Thomas, 1964:87-95.
4. Ready LB, Cuplin S, Haschke RH, Nessly M. Spinal needle determinants of rate of transdural fluid leak. Anesth Analg 1989;69:457-60.
5. Halpern S, Preston R. Postdural puncture headache and spinal needle design. Anesthesiology 1994;81:1376-83.
6. Harrison D, Langham B. Spinal anaesthesia for urological surgery. Anaesthesia 1992;47:902-3.
7. Lybecker H, Moller JT, May O, Nielsen HK. Incidence and prediction of postdural puncture headache: a prospective study of 1021 spinal anesthesias. Anesth Analg 1990;70:389-94.
8. Manchikanti L, Hadley C, Markwell SJ, Colliver JA. A retrospective analysis of failed spinal anesthetic attempts in a community hospital. Anesth Analg 1987;66:363-6.
9. Munhall RJ, Sukhani R, Winnie AP. Incidence and etiology of failed spinal anesthetics in a university hospital: a prospective study. Anesth Analg 1988;67:843-8.
10. Sykes M. Recognition of the anaesthetist. Anaesthesia 1995;50:381-2.
11. Levy JH, Islas JA, Ghia JN, Turnbull C. A retrospective study of the incidence and causes of failed spinal anesthetics in a university hospital. Anesth Analg 1985;64:705-10.
12. Scott D, Hibbard B. Serious non-fatal complications associated with extradural block in obstetric practice. Br J Anaesth 1990;64:537-41.